diff --git a/cts/cts-fencing.in b/cts/cts-fencing.in
index 2d9999ca01..239319cb16 100644
--- a/cts/cts-fencing.in
+++ b/cts/cts-fencing.in
@@ -1,1527 +1,1556 @@
#!@PYTHON@
""" Regression tests for Pacemaker's fencer
"""
# Pacemaker targets compatibility with Python 2.7 and 3.2+
from __future__ import print_function, unicode_literals, absolute_import, division
__copyright__ = "Copyright 2012-2019 the Pacemaker project contributors"
__license__ = "GNU General Public License version 2 or later (GPLv2+) WITHOUT ANY WARRANTY"
import io
import os
import re
import sys
import subprocess
import shlex
import time
import tempfile
import signal
# Where to find test binaries
# Prefer the source tree if available
BUILD_DIR = "@abs_top_builddir@"
SCHEMA_DIR = "@CRM_SCHEMA_DIRECTORY@"
TEST_DIR = sys.path[0]
AUTOGEN_COROSYNC_TEMPLATE = """
totem {
version: 2
cluster_name: cts-fencing
crypto_cipher: none
crypto_hash: none
transport: udp
}
nodelist {
node {
nodeid: 1
name: %s
ring0_addr: 127.0.0.1
}
}
logging {
debug: off
to_syslog: no
to_stderr: no
to_logfile: yes
logfile: %s
}
"""
# These values must be kept in sync with include/crm/crm.h
class CrmExit(object):
OK = 0
ERROR = 1
INVALID_PARAM = 2
UNIMPLEMENT_FEATURE = 3
INSUFFICIENT_PRIV = 4
NOT_INSTALLED = 5
NOT_CONFIGURED = 6
NOT_RUNNING = 7
USAGE = 64
DATAERR = 65
NOINPUT = 66
NOUSER = 67
NOHOST = 68
UNAVAILABLE = 69
SOFTWARE = 70
OSERR = 71
OSFILE = 72
CANTCREAT = 73
IOERR = 74
TEMPFAIL = 75
PROTOCOL = 76
NOPERM = 77
CONFIG = 78
FATAL = 100
PANIC = 101
DISCONNECT = 102
SOLO = 103
DIGEST = 104
NOSUCH = 105
QUORUM = 106
UNSAFE = 107
EXISTS = 108
MULTIPLE = 109
OLD = 110
TIMEOUT = 124
MAX = 255
def update_path():
""" Set the PATH environment variable appropriately for the tests """
new_path = os.environ['PATH']
if os.path.exists("%s/cts-fencing.in" % TEST_DIR):
print("Running tests from the source tree: %s (%s)" % (BUILD_DIR, TEST_DIR))
# For pacemaker-fenced and cts-fence-helper
new_path = "%s/daemons/fenced:%s" % (BUILD_DIR, new_path)
new_path = "%s/tools:%s" % (BUILD_DIR, new_path) # For stonith_admin
new_path = "%s/cts:%s" % (BUILD_DIR, new_path) # For cts-support
else:
print("Running tests from the install tree: @CRM_DAEMON_DIR@ (not %s)" % TEST_DIR)
# For pacemaker-fenced, cts-fence-helper, and cts-support
new_path = "@CRM_DAEMON_DIR@:%s" % (new_path)
print('Using PATH="{}"'.format(new_path))
os.environ['PATH'] = new_path
def find_validator(rng_file):
if os.access("/usr/bin/xmllint", os.X_OK):
return ["xmllint", "--relaxng", rng_file, "-"]
else:
return None
def rng_directory():
if "PCMK_schema_directory" in os.environ:
return os.environ["PCMK_schema_directory"]
elif os.path.exists("%s/cts-fencing.in" % TEST_DIR):
return "xml"
else:
return SCHEMA_DIR
def pipe_communicate(pipes, stdout=True, stderr=False, stdin=None):
""" Wrapper to get text output from pipes regardless of Python version """
output = ""
if stdin:
if sys.version_info < (3,):
pipe_outputs = pipes.communicate(input=stdin)
else:
pipe_outputs = pipes.communicate(input=stdin.encode())
else:
pipe_outputs = pipes.communicate()
if sys.version_info < (3,):
if stdout:
output = output + pipe_outputs[0]
if stderr:
output = output + pipe_outputs[1]
else:
if stdout:
output = output + pipe_outputs[0].decode(sys.stdout.encoding)
if stderr:
output = output + pipe_outputs[1].decode(sys.stderr.encoding)
return output
def output_from_command(command):
""" Execute command and return its standard output """
test = subprocess.Popen(shlex.split(command), stdout=subprocess.PIPE)
test.wait()
return pipe_communicate(test).split("\n")
def localname():
""" Return the uname of the local host """
our_uname = output_from_command("uname -n")
if our_uname:
our_uname = our_uname[0]
else:
our_uname = "localhost"
return our_uname
def killall(process):
""" Kill all instances of a process """
cmd = shlex.split("killall -9 -q %s" % process)
test = subprocess.Popen(cmd, stdout=subprocess.PIPE)
test.wait()
class TestError(Exception):
""" Base class for exceptions in this module """
pass
class ExitCodeError(TestError):
""" Exception raised when command exit status is unexpected """
def __init__(self, exit_code):
self.exit_code = exit_code
def __str__(self):
return repr(self.exit_code)
class OutputNotFoundError(TestError):
""" Exception raised when command output does not contain wanted string """
def __init__(self, output):
self.output = output
def __str__(self):
return repr(self.output)
class OutputFoundError(TestError):
""" Exception raised when command output contains unwanted string """
def __init__(self, output):
self.output = output
def __str__(self):
return repr(self.output)
class XmlValidationError(TestError):
""" Exception raised when xmllint fails """
def __init__(self, output):
self.output = output
def __str__(self):
return repr(self.output)
class Test(object):
""" Executor for a single test """
def __init__(self, name, description, verbose=0, with_cpg=0):
self.name = name
self.description = description
self.cmds = []
self.verbose = verbose
self.result_txt = ""
self.cmd_tool_output = ""
self.result_exitcode = CrmExit.OK
if with_cpg:
self.stonith_options = "-c"
self.enable_corosync = 1
else:
self.stonith_options = "-s"
self.enable_corosync = 0
self.stonith_process = None
self.stonith_output = ""
self.stonith_patterns = []
self.negative_stonith_patterns = []
self.executed = 0
def __new_cmd(self, cmd, args, exitcode, stdout_match="", no_wait=0, stdout_negative_match="", kill=None, validate=True):
""" Add a command to be executed as part of this test """
self.cmds.append(
{
"cmd" : cmd,
"kill" : kill,
"args" : args,
"expected_exitcode" : exitcode,
"stdout_match" : stdout_match,
"stdout_negative_match" : stdout_negative_match,
"no_wait" : no_wait,
"validate" : validate,
}
)
def start_environment(self):
""" Prepare the host for executing a test """
# Make sure we are in full control
killall("pacemakerd")
killall("pacemaker-fenced")
if self.verbose:
self.stonith_options = self.stonith_options + " -V"
print("Starting pacemaker-fenced with %s" % self.stonith_options)
if os.path.exists("/tmp/stonith-regression.log"):
os.remove('/tmp/stonith-regression.log')
cmd = "pacemaker-fenced %s -l /tmp/stonith-regression.log" % self.stonith_options
self.stonith_process = subprocess.Popen(shlex.split(cmd))
time.sleep(1)
def clean_environment(self):
""" Clean up the host after executing a test """
if self.stonith_process:
if self.stonith_process.poll() == None:
self.stonith_process.terminate()
self.stonith_process.wait()
else:
return_code = {
getattr(signal, _signame): _signame
for _signame in dir(signal)
if _signame.startswith('SIG') and not _signame.startswith("SIG_")
}.get(-self.stonith_process.returncode, "RET=%d" % (self.stonith_process.returncode))
msg = "FAILURE - '%s' failed. pacemaker-fenced abnormally exited during test (%s)."
self.result_txt = msg % (self.name, return_code)
self.result_exitcode = CrmExit.ERROR
self.stonith_output = ""
self.stonith_process = None
# the default for utf-8 encoding would error out if e.g. memory corruption
# makes fenced output any kind of 8 bit value - while still interesting
# for debugging and we'd still like the regression-test to go over the
# full set of test-cases
logfile = io.open('/tmp/stonith-regression.log', 'rt', encoding = "ISO-8859-1")
for line in logfile.readlines():
self.stonith_output = self.stonith_output + line
if self.verbose:
print("Daemon Output Start")
print(self.stonith_output)
print("Daemon Output End")
os.remove('/tmp/stonith-regression.log')
def add_stonith_log_pattern(self, pattern):
""" Add a log pattern to expect from this test """
self.stonith_patterns.append(pattern)
def add_stonith_neg_log_pattern(self, pattern):
""" Add a log pattern that should not occur with this test """
self.negative_stonith_patterns.append(pattern)
def add_cmd(self, cmd, args, validate=True):
""" Add a simple command to be executed as part of this test """
self.__new_cmd(cmd, args, CrmExit.OK, "", validate=validate)
def add_cmd_no_wait(self, cmd, args):
""" Add a simple command to be executed (without waiting) as part of this test """
self.__new_cmd(cmd, args, CrmExit.OK, "", 1)
def add_cmd_check_stdout(self, cmd, args, match, no_match=""):
""" Add a simple command with expected output to be executed as part of this test """
self.__new_cmd(cmd, args, CrmExit.OK, match, 0, no_match)
def add_expected_fail_cmd(self, cmd, args, exitcode=CrmExit.ERROR):
""" Add a command to be executed as part of this test and expected to fail """
self.__new_cmd(cmd, args, exitcode, "")
def get_exitcode(self):
""" Return the exit status of the last test execution """
return self.result_exitcode
def print_result(self, filler):
""" Print the result of the last test execution """
print("%s%s" % (filler, self.result_txt))
def run_cmd(self, args):
""" Execute a command as part of this test """
cmd = shlex.split(args['args'])
cmd.insert(0, args['cmd'])
if self.verbose:
print("\n\nRunning: "+" ".join(cmd))
test = subprocess.Popen(cmd, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
if args['kill']:
if self.verbose:
print("Also running: "+args['kill'])
subprocess.Popen(shlex.split(args['kill']))
if args['no_wait'] == 0:
test.wait()
else:
return CrmExit.OK
output = pipe_communicate(test, stderr=True)
if self.verbose:
print(output)
if test.returncode != args['expected_exitcode']:
raise ExitCodeError(test.returncode)
if (args['stdout_match'] != "" and
re.search(args['stdout_match'], output) is None):
raise OutputNotFoundError(output)
if (args['stdout_negative_match'] != "" and
re.search(args['stdout_negative_match'], output) is not None):
raise OutputFoundError(output)
if args['validate']:
rng_file = rng_directory() + "/api/api-result.rng"
cmd = find_validator(rng_file)
if not cmd:
return
if self.verbose:
print("\nRunning: "+" ".join(cmd))
validator = subprocess.Popen(cmd, stdin=subprocess.PIPE, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
output = pipe_communicate(validator, stderr=True, stdin=output)
if self.verbose:
print(output)
if validator.returncode != 0:
raise XmlValidationError(output)
def count_negative_matches(self, outline):
""" Return 1 if a line matches patterns that shouldn't have occurred """
count = 0
for line in self.negative_stonith_patterns:
if outline.count(line):
count = 1
if self.verbose:
print("This pattern should not have matched = '%s" % (line))
return count
def match_stonith_patterns(self):
""" Check test output for expected patterns """
negative_matches = 0
cur = 0
pats = self.stonith_patterns
total_patterns = len(self.stonith_patterns)
if len(self.stonith_patterns) == 0 and len(self.negative_stonith_patterns) == 0:
return
for line in self.stonith_output.split("\n"):
negative_matches = negative_matches + self.count_negative_matches(line)
if len(pats) == 0:
continue
cur = -1
for pat in pats:
cur = cur + 1
if line.count(pats[cur]):
del pats[cur]
break
if len(pats) > 0 or negative_matches:
if self.verbose:
for pat in pats:
print("Pattern Not Matched = '%s'" % pat)
msg = "FAILURE - '%s' failed. %d patterns out of %d not matched. %d negative matches."
self.result_txt = msg % (self.name, len(pats), total_patterns, negative_matches)
self.result_exitcode = CrmExit.ERROR
def set_error(self, step, cmd):
""" Record failure of this test """
msg = "FAILURE - '%s' failed at step %d. Command: %s %s"
self.result_txt = msg % (self.name, step, cmd['cmd'], cmd['args'])
self.result_exitcode = CrmExit.ERROR
def run(self):
""" Execute this test. """
res = 0
i = 1
self.start_environment()
if self.verbose:
print("\n--- START TEST - %s" % self.name)
self.result_txt = "SUCCESS - '%s'" % (self.name)
self.result_exitcode = CrmExit.OK
for cmd in self.cmds:
try:
self.run_cmd(cmd)
except ExitCodeError as e:
print("Step %d FAILED - command returned %s, expected %d" % (i, e, cmd['expected_exitcode']))
self.set_error(i, cmd);
break
except OutputNotFoundError as e:
print("Step %d FAILED - '%s' was not found in command output: %s" % (i, cmd['stdout_match'], e))
self.set_error(i, cmd);
break
except OutputFoundError as e:
print("Step %d FAILED - '%s' was found in command output: %s" % (i, cmd['stdout_negative_match'], e))
self.set_error(i, cmd);
break
if self.verbose:
print("Step %d SUCCESS" % (i))
i = i + 1
self.clean_environment()
if self.result_exitcode == CrmExit.OK:
self.match_stonith_patterns()
print(self.result_txt)
if self.verbose:
print("--- END TEST - %s\n" % self.name)
self.executed = 1
return res
class Tests(object):
""" Collection of all fencing regression tests """
def __init__(self, verbose=0):
self.tests = []
self.verbose = verbose
self.autogen_corosync_cfg = not os.path.exists("/etc/corosync/corosync.conf")
def new_test(self, name, description, with_cpg=0):
""" Create a named test """
test = Test(name, description, self.verbose, with_cpg)
self.tests.append(test)
return test
def print_list(self):
""" List all registered tests """
print("\n==== %d TESTS FOUND ====" % (len(self.tests)))
print("%35s - %s" % ("TEST NAME", "TEST DESCRIPTION"))
print("%35s - %s" % ("--------------------", "--------------------"))
for test in self.tests:
print("%35s - %s" % (test.name, test.description))
print("==== END OF LIST ====\n")
def start_corosync(self):
""" Start the corosync process """
if self.verbose:
print("Starting corosync")
test = subprocess.Popen("corosync", stdout=subprocess.PIPE)
test.wait()
time.sleep(10)
def run_single(self, name):
""" Run a single named test """
for test in self.tests:
if test.name == name:
test.run()
break
def run_tests_matching(self, pattern):
""" Run all tests whose name matches a pattern """
for test in self.tests:
if test.name.count(pattern) != 0:
test.run()
def run_cpg_only(self):
""" Run all corosync-enabled tests """
for test in self.tests:
if test.enable_corosync:
test.run()
def run_no_cpg(self):
""" Run all standalone tests """
for test in self.tests:
if not test.enable_corosync:
test.run()
def run_tests(self):
""" Run all tests """
for test in self.tests:
test.run()
def exit(self):
""" Exit (with error status code if any test failed) """
for test in self.tests:
if test.executed == 0:
continue
if test.get_exitcode() != CrmExit.OK:
sys.exit(CrmExit.ERROR)
sys.exit(CrmExit.OK)
def print_results(self):
""" Print summary of results of executed tests """
failures = 0
success = 0
print("\n\n======= FINAL RESULTS ==========")
print("\n--- FAILURE RESULTS:")
for test in self.tests:
if test.executed == 0:
continue
if test.get_exitcode() != CrmExit.OK:
failures = failures + 1
test.print_result(" ")
else:
success = success + 1
if failures == 0:
print(" None")
print("\n--- TOTALS\n Pass:%d\n Fail:%d\n" % (success, failures))
def build_api_sanity_tests(self):
""" Register tests to verify basic API usage """
verbose_arg = ""
if self.verbose:
verbose_arg = "-V"
test = self.new_test("standalone_low_level_api_test", "Sanity test client api in standalone mode.")
test.add_cmd("cts-fence-helper", "-t %s" % (verbose_arg), validate=False)
test = self.new_test("cpg_low_level_api_test", "Sanity test client api using mainloop and cpg.", 1)
test.add_cmd("cts-fence-helper", "-m %s" % (verbose_arg), validate=False)
def build_custom_timeout_tests(self):
""" Register tests to verify custom timeout usage """
# custom timeout without topology
test = self.new_test("cpg_custom_timeout_1",
"Verify per device timeouts work as expected without using topology.", 1)
test.add_cmd('stonith_admin',
'--output-as=xml -R false1 -a fence_dummy -o "mode=fail" -o "pcmk_host_list=node1 node2 node3"')
test.add_cmd('stonith_admin',
'--output-as=xml -R true1 -a fence_dummy -o "mode=pass" -o "pcmk_host_list=node3" -o "pcmk_off_timeout=1"')
test.add_cmd('stonith_admin',
'--output-as=xml -R false2 -a fence_dummy -o "mode=fail" -o "pcmk_host_list=node3" -o "pcmk_off_timeout=4"')
test.add_cmd("stonith_admin", "--output-as=xml -F node3 -t 5")
# timeout is 5+1+4 = 10
test.add_stonith_log_pattern("Total timeout set to 10")
# custom timeout _WITH_ topology
test = self.new_test("cpg_custom_timeout_2",
"Verify per device timeouts work as expected _WITH_ topology.", 1)
test.add_cmd('stonith_admin',
'--output-as=xml -R false1 -a fence_dummy -o "mode=fail" -o "pcmk_host_list=node1 node2 node3"')
test.add_cmd('stonith_admin',
'--output-as=xml -R true1 -a fence_dummy -o "mode=pass" -o "pcmk_host_list=node3" -o "pcmk_off_timeout=1"')
test.add_cmd('stonith_admin',
'--output-as=xml -R false2 -a fence_dummy -o "mode=fail" -o "pcmk_host_list=node3" -o "pcmk_off_timeout=4000"')
test.add_cmd("stonith_admin", "--output-as=xml -r node3 -i 1 -v false1")
test.add_cmd("stonith_admin", "--output-as=xml -r node3 -i 2 -v true1")
test.add_cmd("stonith_admin", "--output-as=xml -r node3 -i 3 -v false2")
test.add_cmd("stonith_admin", "--output-as=xml -F node3 -t 5")
# timeout is 5+1+4000 = 4006
test.add_stonith_log_pattern("Total timeout set to 4006")
def build_fence_merge_tests(self):
""" Register tests to verify when fence operations should be merged """
### Simple test that overlapping fencing operations get merged
test = self.new_test("cpg_custom_merge_single",
"Verify overlapping identical fencing operations are merged, no fencing levels used.", 1)
test.add_cmd("stonith_admin", "--output-as=xml -R false1 -a fence_dummy -o \"mode=fail\" -o \"pcmk_host_list=node3\"")
test.add_cmd("stonith_admin", "--output-as=xml -R true1 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=node3\" ")
test.add_cmd("stonith_admin", "--output-as=xml -R false2 -a fence_dummy -o \"mode=fail\" -o \"pcmk_host_list=node3\"")
test.add_cmd_no_wait("stonith_admin", "--output-as=xml -F node3 -t 10")
test.add_cmd("stonith_admin", "--output-as=xml -F node3 -t 10")
### one merger will happen
test.add_stonith_log_pattern("Merging stonith action 'off' targeting node3 originating from client")
### the pattern below signifies that both the original and duplicate operation completed
test.add_stonith_log_pattern("Operation 'off' targeting node3 on")
test.add_stonith_log_pattern("Operation 'off' targeting node3 on")
### Test that multiple mergers occur
test = self.new_test("cpg_custom_merge_multiple",
"Verify multiple overlapping identical fencing operations are merged", 1)
test.add_cmd("stonith_admin", "--output-as=xml -R false1 -a fence_dummy -o \"mode=fail\" -o \"pcmk_host_list=node3\"")
test.add_cmd("stonith_admin",
"--output-as=xml -R true1 -a fence_dummy -o \"mode=pass\" -o \"delay=2\" -o \"pcmk_host_list=node3\" ")
test.add_cmd("stonith_admin", "--output-as=xml -R false2 -a fence_dummy -o \"mode=fail\" -o \"pcmk_host_list=node3\"")
test.add_cmd_no_wait("stonith_admin", "--output-as=xml -F node3 -t 10")
test.add_cmd_no_wait("stonith_admin", "--output-as=xml -F node3 -t 10")
test.add_cmd_no_wait("stonith_admin", "--output-as=xml -F node3 -t 10")
test.add_cmd_no_wait("stonith_admin", "--output-as=xml -F node3 -t 10")
test.add_cmd("stonith_admin", "--output-as=xml -F node3 -t 10")
### 4 mergers should occur
test.add_stonith_log_pattern("Merging stonith action 'off' targeting node3 originating from client")
test.add_stonith_log_pattern("Merging stonith action 'off' targeting node3 originating from client")
test.add_stonith_log_pattern("Merging stonith action 'off' targeting node3 originating from client")
test.add_stonith_log_pattern("Merging stonith action 'off' targeting node3 originating from client")
### the pattern below signifies that both the original and duplicate operation completed
test.add_stonith_log_pattern("Operation 'off' targeting node3 on")
test.add_stonith_log_pattern("Operation 'off' targeting node3 on")
test.add_stonith_log_pattern("Operation 'off' targeting node3 on")
test.add_stonith_log_pattern("Operation 'off' targeting node3 on")
test.add_stonith_log_pattern("Operation 'off' targeting node3 on")
### Test that multiple mergers occur with topologies used
test = self.new_test("cpg_custom_merge_with_topology",
"Verify multiple overlapping identical fencing operations are merged with fencing levels.",
1)
test.add_cmd("stonith_admin", "--output-as=xml -R false1 -a fence_dummy -o \"mode=fail\" -o \"pcmk_host_list=node3\"")
test.add_cmd("stonith_admin", "--output-as=xml -R true1 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=node3\" ")
test.add_cmd("stonith_admin", "--output-as=xml -R false2 -a fence_dummy -o \"mode=fail\" -o \"pcmk_host_list=node3\"")
test.add_cmd("stonith_admin", "--output-as=xml -r node3 -i 1 -v false1")
test.add_cmd("stonith_admin", "--output-as=xml -r node3 -i 1 -v false2")
test.add_cmd("stonith_admin", "--output-as=xml -r node3 -i 2 -v true1")
test.add_cmd_no_wait("stonith_admin", "--output-as=xml -F node3 -t 10")
test.add_cmd_no_wait("stonith_admin", "--output-as=xml -F node3 -t 10")
test.add_cmd_no_wait("stonith_admin", "--output-as=xml -F node3 -t 10")
test.add_cmd_no_wait("stonith_admin", "--output-as=xml -F node3 -t 10")
test.add_cmd("stonith_admin", "--output-as=xml -F node3 -t 10")
### 4 mergers should occur
test.add_stonith_log_pattern("Merging stonith action 'off' targeting node3 originating from client")
test.add_stonith_log_pattern("Merging stonith action 'off' targeting node3 originating from client")
test.add_stonith_log_pattern("Merging stonith action 'off' targeting node3 originating from client")
test.add_stonith_log_pattern("Merging stonith action 'off' targeting node3 originating from client")
### the pattern below signifies that both the original and duplicate operation completed
test.add_stonith_log_pattern("Operation 'off' targeting node3 on")
test.add_stonith_log_pattern("Operation 'off' targeting node3 on")
test.add_stonith_log_pattern("Operation 'off' targeting node3 on")
test.add_stonith_log_pattern("Operation 'off' targeting node3 on")
test.add_stonith_log_pattern("Operation 'off' targeting node3 on")
def build_fence_no_merge_tests(self):
""" Register tests to verify when fence operations should not be merged """
test = self.new_test("cpg_custom_no_merge",
"Verify differing fencing operations are not merged", 1)
test.add_cmd("stonith_admin", "--output-as=xml -R false1 -a fence_dummy -o \"mode=fail\" -o \"pcmk_host_list=node3 node2\"")
test.add_cmd("stonith_admin", "--output-as=xml -R true1 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=node3 node2\" ")
test.add_cmd("stonith_admin", "--output-as=xml -R false2 -a fence_dummy -o \"mode=fail\" -o \"pcmk_host_list=node3 node2\"")
test.add_cmd("stonith_admin", "--output-as=xml -r node3 -i 1 -v false1")
test.add_cmd("stonith_admin", "--output-as=xml -r node3 -i 1 -v false2")
test.add_cmd("stonith_admin", "--output-as=xml -r node3 -i 2 -v true1")
test.add_cmd_no_wait("stonith_admin", "--output-as=xml -F node2 -t 10")
test.add_cmd("stonith_admin", "--output-as=xml -F node3 -t 10")
test.add_stonith_neg_log_pattern("Merging stonith action 'off' targeting node3 originating from client")
def build_standalone_tests(self):
""" Register a grab bag of tests that can be executed in standalone or corosync mode """
test_types = [
{
"prefix" : "standalone",
"use_cpg" : 0,
},
{
"prefix" : "cpg",
"use_cpg" : 1,
},
]
# test what happens when all devices timeout
for test_type in test_types:
test = self.new_test("%s_fence_multi_device_failure" % test_type["prefix"],
"Verify that all devices timeout, a fencing failure is returned.",
test_type["use_cpg"])
test.add_cmd("stonith_admin",
"--output-as=xml -R false1 -a fence_dummy -o \"mode=fail\" -o \"pcmk_host_list=node1 node2 node3\"")
test.add_cmd("stonith_admin",
"--output-as=xml -R false2 -a fence_dummy -o \"mode=fail\" -o \"pcmk_host_list=node1 node2 node3\"")
test.add_cmd("stonith_admin",
"--output-as=xml -R false3 -a fence_dummy -o \"mode=fail\" -o \"pcmk_host_list=node1 node2 node3\"")
if test_type["use_cpg"] == 1:
test.add_expected_fail_cmd("stonith_admin", "--output-as=xml -F node3 -t 2", CrmExit.TIMEOUT)
test.add_stonith_log_pattern("Total timeout set to 6")
else:
test.add_expected_fail_cmd("stonith_admin", "--output-as=xml -F node3 -t 2", CrmExit.ERROR)
test.add_stonith_log_pattern("for host 'node3' with device 'false1' returned: ")
test.add_stonith_log_pattern("for host 'node3' with device 'false2' returned: ")
test.add_stonith_log_pattern("for host 'node3' with device 'false3' returned: ")
# test what happens when multiple devices can fence a node, but the first device fails.
for test_type in test_types:
test = self.new_test("%s_fence_device_failure_rollover" % test_type["prefix"],
"Verify that when one fence device fails for a node, the others are tried.",
test_type["use_cpg"])
test.add_cmd("stonith_admin",
"--output-as=xml -R false1 -a fence_dummy -o \"mode=fail\" -o \"pcmk_host_list=node1 node2 node3\"")
test.add_cmd("stonith_admin",
"--output-as=xml -R true1 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=node1 node2 node3\"")
test.add_cmd("stonith_admin",
"--output-as=xml -R false2 -a fence_dummy -o \"mode=fail\" -o \"pcmk_host_list=node1 node2 node3\"")
test.add_cmd("stonith_admin", "--output-as=xml -F node3 -t 5")
if test_type["use_cpg"] == 1:
test.add_stonith_log_pattern("Total timeout set to 15")
# test what happens when we try to use a missing fence-agent.
for test_type in test_types:
test = self.new_test("%s_fence_missing_agent" % test_type["prefix"],
"Verify proper error-handling when using a non-existent fence-agent.",
test_type["use_cpg"])
test.add_cmd("stonith_admin",
"--output-as=xml -R true1 -a fence_missing -o \"mode=pass\" -o \"pcmk_host_list=node3\"")
test.add_cmd("stonith_admin",
"--output-as=xml -R true2 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=node2\"")
test.add_expected_fail_cmd("stonith_admin", "--output-as=xml -F node3 -t 5", CrmExit.ERROR)
test.add_cmd("stonith_admin", "--output-as=xml -F node2 -t 5")
# simple topology test for one device
for test_type in test_types:
if test_type["use_cpg"] == 0:
continue
test = self.new_test("%s_topology_simple" % test_type["prefix"],
"Verify all fencing devices at a level are used.", test_type["use_cpg"])
test.add_cmd("stonith_admin",
"--output-as=xml -R true -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=node1 node2 node3\"")
test.add_cmd("stonith_admin", "--output-as=xml -r node3 -i 1 -v true")
test.add_cmd("stonith_admin", "--output-as=xml -F node3 -t 5")
test.add_stonith_log_pattern("Total timeout set to 5")
test.add_stonith_log_pattern("for host 'node3' with device 'true' returned: 0")
# add topology, delete topology, verify fencing still works
for test_type in test_types:
if test_type["use_cpg"] == 0:
continue
test = self.new_test("%s_topology_add_remove" % test_type["prefix"],
"Verify fencing occurrs after all topology levels are removed",
test_type["use_cpg"])
test.add_cmd("stonith_admin",
"--output-as=xml -R true -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=node1 node2 node3\"")
test.add_cmd("stonith_admin", "--output-as=xml -r node3 -i 1 -v true")
test.add_cmd("stonith_admin", "--output-as=xml -d node3 -i 1")
test.add_cmd("stonith_admin", "--output-as=xml -F node3 -t 5")
test.add_stonith_log_pattern("Total timeout set to 5")
test.add_stonith_log_pattern("for host 'node3' with device 'true' returned: 0")
# test what happens when the first fencing level has multiple devices.
for test_type in test_types:
if test_type["use_cpg"] == 0:
continue
test = self.new_test("%s_topology_device_fails" % test_type["prefix"],
"Verify if one device in a level fails, the other is tried.",
test_type["use_cpg"])
test.add_cmd("stonith_admin",
"--output-as=xml -R false -a fence_dummy -o \"mode=fail\" -o \"pcmk_host_list=node1 node2 node3\"")
test.add_cmd("stonith_admin",
"--output-as=xml -R true -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=node1 node2 node3\"")
test.add_cmd("stonith_admin", "--output-as=xml -r node3 -i 1 -v false")
test.add_cmd("stonith_admin", "--output-as=xml -r node3 -i 2 -v true")
test.add_cmd("stonith_admin", "--output-as=xml -F node3 -t 20")
test.add_stonith_log_pattern("Total timeout set to 40")
test.add_stonith_log_pattern("for host 'node3' with device 'false' returned: -201")
test.add_stonith_log_pattern("for host 'node3' with device 'true' returned: 0")
# test what happens when the first fencing level fails.
for test_type in test_types:
if test_type["use_cpg"] == 0:
continue
test = self.new_test("%s_topology_multi_level_fails" % test_type["prefix"],
"Verify if one level fails, the next leve is tried.",
test_type["use_cpg"])
test.add_cmd("stonith_admin",
"--output-as=xml -R true1 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=node1 node2 node3\"")
test.add_cmd("stonith_admin",
"--output-as=xml -R true2 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=node1 node2 node3\"")
test.add_cmd("stonith_admin",
"--output-as=xml -R true3 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=node1 node2 node3\"")
test.add_cmd("stonith_admin",
"--output-as=xml -R true4 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=node1 node2 node3\"")
test.add_cmd("stonith_admin",
"--output-as=xml -R false1 -a fence_dummy -o \"mode=fail\" -o \"pcmk_host_list=node1 node2 node3\"")
test.add_cmd("stonith_admin",
"--output-as=xml -R false2 -a fence_dummy -o \"mode=fail\" -o \"pcmk_host_list=node1 node2 node3\"")
test.add_cmd("stonith_admin", "--output-as=xml -r node3 -i 1 -v false1")
test.add_cmd("stonith_admin", "--output-as=xml -r node3 -i 1 -v true1")
test.add_cmd("stonith_admin", "--output-as=xml -r node3 -i 2 -v true2")
test.add_cmd("stonith_admin", "--output-as=xml -r node3 -i 2 -v false2")
test.add_cmd("stonith_admin", "--output-as=xml -r node3 -i 3 -v true3")
test.add_cmd("stonith_admin", "--output-as=xml -r node3 -i 3 -v true4")
test.add_cmd("stonith_admin", "--output-as=xml -F node3 -t 3")
test.add_stonith_log_pattern("Total timeout set to 18")
test.add_stonith_log_pattern("for host 'node3' with device 'false1' returned: -201")
test.add_stonith_log_pattern("for host 'node3' with device 'false2' returned: -201")
test.add_stonith_log_pattern("for host 'node3' with device 'true3' returned: 0")
test.add_stonith_log_pattern("for host 'node3' with device 'true4' returned: 0")
# test what happens when the first fencing level had devices that no one has registered
for test_type in test_types:
if test_type["use_cpg"] == 0:
continue
test = self.new_test("%s_topology_missing_devices" % test_type["prefix"],
"Verify topology can continue with missing devices.",
test_type["use_cpg"])
test.add_cmd("stonith_admin",
"--output-as=xml -R true2 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=node1 node2 node3\"")
test.add_cmd("stonith_admin",
"--output-as=xml -R true3 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=node1 node2 node3\"")
test.add_cmd("stonith_admin",
"--output-as=xml -R true4 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=node1 node2 node3\"")
test.add_cmd("stonith_admin",
"--output-as=xml -R false2 -a fence_dummy -o \"mode=fail\" -o \"pcmk_host_list=node1 node2 node3\"")
test.add_cmd("stonith_admin", "--output-as=xml -r node3 -i 1 -v false1")
test.add_cmd("stonith_admin", "--output-as=xml -r node3 -i 1 -v true1")
test.add_cmd("stonith_admin", "--output-as=xml -r node3 -i 2 -v true2")
test.add_cmd("stonith_admin", "--output-as=xml -r node3 -i 2 -v false2")
test.add_cmd("stonith_admin", "--output-as=xml -r node3 -i 3 -v true3")
test.add_cmd("stonith_admin", "--output-as=xml -r node3 -i 3 -v true4")
test.add_cmd("stonith_admin", "--output-as=xml -F node3 -t 5")
# Test what happens if multiple fencing levels are defined, and then the first one is removed.
for test_type in test_types:
if test_type["use_cpg"] == 0:
continue
test = self.new_test("%s_topology_level_removal" % test_type["prefix"],
"Verify level removal works.", test_type["use_cpg"])
test.add_cmd("stonith_admin",
"--output-as=xml -R true1 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=node1 node2 node3\"")
test.add_cmd("stonith_admin",
"--output-as=xml -R true2 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=node1 node2 node3\"")
test.add_cmd("stonith_admin",
"--output-as=xml -R true3 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=node1 node2 node3\"")
test.add_cmd("stonith_admin",
"--output-as=xml -R true4 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=node1 node2 node3\"")
test.add_cmd("stonith_admin",
"--output-as=xml -R false1 -a fence_dummy -o \"mode=fail\" -o \"pcmk_host_list=node1 node2 node3\"")
test.add_cmd("stonith_admin",
"--output-as=xml -R false2 -a fence_dummy -o \"mode=fail\" -o \"pcmk_host_list=node1 node2 node3\"")
test.add_cmd("stonith_admin", "--output-as=xml -r node3 -i 1 -v false1")
test.add_cmd("stonith_admin", "--output-as=xml -r node3 -i 1 -v true1")
test.add_cmd("stonith_admin", "--output-as=xml -r node3 -i 2 -v true2")
test.add_cmd("stonith_admin", "--output-as=xml -r node3 -i 2 -v false2")
test.add_cmd("stonith_admin", "--output-as=xml -r node3 -i 3 -v true3")
test.add_cmd("stonith_admin", "--output-as=xml -r node3 -i 3 -v true4")
# Now remove level 2, verify none of the devices in level two are hit.
test.add_cmd("stonith_admin", "--output-as=xml -d node3 -i 2")
test.add_cmd("stonith_admin", "--output-as=xml -F node3 -t 20")
test.add_stonith_log_pattern("Total timeout set to 8")
test.add_stonith_log_pattern("for host 'node3' with device 'false1' returned: -201")
test.add_stonith_neg_log_pattern("for host 'node3' with device 'false2' returned: ")
test.add_stonith_log_pattern("for host 'node3' with device 'true3' returned: 0")
test.add_stonith_log_pattern("for host 'node3' with device 'true4' returned: 0")
# Test targeting a topology level by node name pattern.
for test_type in test_types:
if test_type["use_cpg"] == 0:
continue
test = self.new_test("%s_topology_level_pattern" % test_type["prefix"],
"Verify targeting topology by node name pattern works.",
test_type["use_cpg"])
test.add_cmd("stonith_admin",
"""--output-as=xml -R true -a fence_dummy -o "mode=pass" -o "pcmk_host_list=node1 node2 node3" """)
test.add_cmd("stonith_admin", """--output-as=xml -r '@node.*' -i 1 -v true""")
test.add_cmd("stonith_admin", "--output-as=xml -F node3 -t 5")
test.add_stonith_log_pattern("for host 'node3' with device 'true' returned: 0")
# test allowing commas and semicolons as delimiters in pcmk_host_list
for test_type in test_types:
test = self.new_test("%s_host_list_delimiters" % test_type["prefix"],
"Verify commas and semicolons can be used as pcmk_host_list delimiters",
test_type["use_cpg"])
test.add_cmd("stonith_admin",
"""--output-as=xml -R true1 -a fence_dummy -o "mode=pass" -o "pcmk_host_list=node1,node2,node3" """)
test.add_cmd("stonith_admin",
"""--output-as=xml -R true2 -a fence_dummy -o "mode=pass" -o "pcmk_host_list=pcmk1;pcmk2;pcmk3" """)
test.add_cmd("stonith_admin", "stonith_admin --output-as=xml -F node2 -t 5")
test.add_cmd("stonith_admin", "stonith_admin --output-as=xml -F pcmk3 -t 5")
test.add_stonith_log_pattern("for host 'node2' with device 'true1' returned: 0")
test.add_stonith_log_pattern("for host 'pcmk3' with device 'true2' returned: 0")
# test the stonith builds the correct list of devices that can fence a node.
for test_type in test_types:
test = self.new_test("%s_list_devices" % test_type["prefix"],
"Verify list of devices that can fence a node is correct",
test_type["use_cpg"])
test.add_cmd("stonith_admin",
"--output-as=xml -R true1 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=node3\"")
test.add_cmd("stonith_admin",
"--output-as=xml -R true2 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=node1 node2 node3\"")
test.add_cmd("stonith_admin",
"--output-as=xml -R true3 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=node1 node2 node3\"")
test.add_cmd_check_stdout("stonith_admin", "--output-as=xml -l node1 -V", "true2", "true1")
test.add_cmd_check_stdout("stonith_admin", "--output-as=xml -l node1 -V", "true3", "true1")
# simple test of device monitor
for test_type in test_types:
test = self.new_test("%s_monitor" % test_type["prefix"],
"Verify device is reachable", test_type["use_cpg"])
test.add_cmd("stonith_admin", "--output-as=xml -R true1 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=node3\"")
test.add_cmd("stonith_admin", "--output-as=xml -R false1 -a fence_dummy -o \"mode=fail\" -o \"pcmk_host_list=node3\"")
test.add_cmd("stonith_admin", "--output-as=xml -Q true1")
test.add_cmd("stonith_admin", "--output-as=xml -Q false1")
test.add_expected_fail_cmd("stonith_admin", "--output-as=xml -Q true2", CrmExit.ERROR)
# Verify monitor occurs for duration of timeout period on failure
for test_type in test_types:
test = self.new_test("%s_monitor_timeout" % test_type["prefix"],
"Verify monitor uses duration of timeout period given.",
test_type["use_cpg"])
test.add_cmd("stonith_admin",
'--output-as=xml -R true1 -a fence_dummy -o "mode=fail" -o "monitor_mode=fail" -o "pcmk_host_list=node3"')
test.add_expected_fail_cmd("stonith_admin", "--output-as=xml -Q true1 -t 5", CrmExit.ERROR)
test.add_stonith_log_pattern("Attempt 2 to execute")
# Verify monitor occurs for duration of timeout period on failure, but stops at max retries
for test_type in test_types:
test = self.new_test("%s_monitor_timeout_max_retries" % test_type["prefix"],
"Verify monitor retries until max retry value or timeout is hit.",
test_type["use_cpg"])
test.add_cmd("stonith_admin",
'--output-as=xml -R true1 -a fence_dummy -o "mode=fail" -o "monitor_mode=fail" -o "pcmk_host_list=node3"')
test.add_expected_fail_cmd("stonith_admin", "--output-as=xml -Q true1 -t 15", CrmExit.ERROR)
test.add_stonith_log_pattern("Attempted to execute agent fence_dummy (list) the maximum number of times")
# simple register test
for test_type in test_types:
test = self.new_test("%s_register" % test_type["prefix"],
"Verify devices can be registered and un-registered",
test_type["use_cpg"])
test.add_cmd("stonith_admin", "--output-as=xml -R true1 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=node3\"")
test.add_cmd("stonith_admin", "--output-as=xml -Q true1")
test.add_cmd("stonith_admin", "--output-as=xml -D true1")
test.add_expected_fail_cmd("stonith_admin", "--output-as=xml -Q true1", CrmExit.ERROR)
# simple reboot test
for test_type in test_types:
test = self.new_test("%s_reboot" % test_type["prefix"],
"Verify devices can be rebooted",
test_type["use_cpg"])
test.add_cmd("stonith_admin", "--output-as=xml -R true1 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=node3\"")
test.add_cmd("stonith_admin", "--output-as=xml -B node3 -t 5")
test.add_cmd("stonith_admin", "--output-as=xml -D true1")
test.add_expected_fail_cmd("stonith_admin", "--output-as=xml -Q true1", CrmExit.ERROR)
# test fencing history.
for test_type in test_types:
if test_type["use_cpg"] == 0:
continue
test = self.new_test("%s_fence_history" % test_type["prefix"],
"Verify last fencing operation is returned.",
test_type["use_cpg"])
test.add_cmd("stonith_admin", "--output-as=xml -R true1 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=node3\"")
test.add_cmd("stonith_admin", "--output-as=xml -F node3 -t 5 -V")
test.add_cmd_check_stdout("stonith_admin", "--output-as=xml -H node3", 'status="success" .* action="off" target="node3"')
# simple test of dynamic list query
for test_type in test_types:
test = self.new_test("%s_dynamic_list_query" % test_type["prefix"],
"Verify dynamic list of fencing devices can be retrieved.",
test_type["use_cpg"])
test.add_cmd("stonith_admin", "--output-as=xml -R true1 -a fence_dummy -o mode=pass -o mock_dynamic_hosts=fake_port_1")
test.add_cmd("stonith_admin", "--output-as=xml -R true2 -a fence_dummy -o mode=pass -o mock_dynamic_hosts=fake_port_1")
test.add_cmd("stonith_admin", "--output-as=xml -R true3 -a fence_dummy -o mode=pass -o mock_dynamic_hosts=fake_port_1")
test.add_cmd_check_stdout("stonith_admin", "--output-as=xml -l fake_port_1", 'count="3"')
# fence using dynamic list query
for test_type in test_types:
test = self.new_test("%s_fence_dynamic_list_query" % test_type["prefix"],
"Verify dynamic list of fencing devices can be retrieved.",
test_type["use_cpg"])
test.add_cmd("stonith_admin", "--output-as=xml -R true1 -a fence_dummy -o mode=pass -o mock_dynamic_hosts=fake_port_1")
test.add_cmd("stonith_admin", "--output-as=xml -R true2 -a fence_dummy -o mode=pass -o mock_dynamic_hosts=fake_port_1")
test.add_cmd("stonith_admin", "--output-as=xml -R true3 -a fence_dummy -o mode=pass -o mock_dynamic_hosts=fake_port_1")
test.add_cmd("stonith_admin", "--output-as=xml -F fake_port_1 -t 5 -V")
# simple test of query using status action
for test_type in test_types:
test = self.new_test("%s_status_query" % test_type["prefix"],
"Verify dynamic list of fencing devices can be retrieved.",
test_type["use_cpg"])
test.add_cmd("stonith_admin", "--output-as=xml -R true1 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_check=status\"")
test.add_cmd("stonith_admin", "--output-as=xml -R true2 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_check=status\"")
test.add_cmd("stonith_admin", "--output-as=xml -R true3 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_check=status\"")
test.add_cmd_check_stdout("stonith_admin", "--output-as=xml -l fake_port_1", 'count="3"')
# test what happens when no reboot action is advertised
for test_type in test_types:
test = self.new_test("%s_no_reboot_support" % test_type["prefix"],
"Verify reboot action defaults to off when no reboot action is advertised by agent.",
test_type["use_cpg"])
test.add_cmd("stonith_admin",
"--output-as=xml -R true1 -a fence_dummy_no_reboot -o \"mode=pass\" -o \"pcmk_host_list=node1 node2 node3\"")
test.add_cmd("stonith_admin", "--output-as=xml -B node1 -t 5 -V")
test.add_stonith_log_pattern("does not advertise support for 'reboot', performing 'off'")
test.add_stonith_log_pattern("with device 'true1' returned: 0 (OK)")
# make sure reboot is used when reboot action is advertised
for test_type in test_types:
test = self.new_test("%s_with_reboot_support" % test_type["prefix"],
"Verify reboot action can be used when metadata advertises it.",
test_type["use_cpg"])
test.add_cmd("stonith_admin",
"--output-as=xml -R true1 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=node1 node2 node3\"")
test.add_cmd("stonith_admin", "--output-as=xml -B node1 -t 5 -V")
test.add_stonith_neg_log_pattern("does not advertise support for 'reboot', performing 'off'")
test.add_stonith_log_pattern("with device 'true1' returned: 0 (OK)")
+ # make sure enforced fencing delay is applied only for the first device in the first level
+ for test_type in test_types:
+ if test_type["use_cpg"] == 0:
+ continue
+
+ test = self.new_test("%s_topology_delay" % test_type["prefix"],
+ "Verify enforced fencing delay is applied only for the first device in the first level.",
+ test_type["use_cpg"])
+ test.add_cmd("stonith_admin",
+ "--output-as=xml -R true1 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=node1 node2 node3\"")
+ test.add_cmd("stonith_admin",
+ "--output-as=xml -R false1 -a fence_dummy -o \"mode=fail\" -o \"pcmk_host_list=node1 node2 node3\"")
+ test.add_cmd("stonith_admin",
+ "--output-as=xml -R true2 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=node1 node2 node3\"")
+ test.add_cmd("stonith_admin",
+ "--output-as=xml -R true3 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=node1 node2 node3\"")
+
+ test.add_cmd("stonith_admin", "--output-as=xml -r node3 -i 1 -v true1")
+ test.add_cmd("stonith_admin", "--output-as=xml -r node3 -i 1 -v false1")
+ test.add_cmd("stonith_admin", "--output-as=xml -r node3 -i 2 -v true2")
+ test.add_cmd("stonith_admin", "--output-as=xml -r node3 -i 2 -v true3")
+
+ test.add_cmd("stonith_admin", "--output-as=xml -F node3 --delay 1")
+
+ test.add_stonith_log_pattern("Delaying 'off' action targeting node3 on true1 for enforced 1s")
+ test.add_stonith_neg_log_pattern("Delaying 'off' action targeting node3 on false1")
+ test.add_stonith_neg_log_pattern("Delaying 'off' action targeting node3 on true2")
+ test.add_stonith_neg_log_pattern("Delaying 'off' action targeting node3 on true3")
+
def build_nodeid_tests(self):
""" Register tests that use a corosync node id """
our_uname = localname()
### verify nodeid is supplied when nodeid is in the metadata parameters
test = self.new_test("cpg_supply_nodeid",
"Verify nodeid is given when fence agent has nodeid as parameter", 1)
test.add_cmd("stonith_admin",
"--output-as=xml -R true1 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=%s\"" % (our_uname))
test.add_cmd("stonith_admin", "--output-as=xml -F %s -t 3" % (our_uname))
test.add_stonith_log_pattern("For stonith action (off) for victim %s, adding nodeid" % (our_uname))
### verify nodeid is _NOT_ supplied when nodeid is not in the metadata parameters
test = self.new_test("cpg_do_not_supply_nodeid",
"Verify nodeid is _NOT_ given when fence agent does not have nodeid as parameter",
1)
# use a host name that won't be in corosync.conf
test.add_cmd("stonith_admin",
"--output-as=xml -R true1 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=regr-test\"")
test.add_cmd("stonith_admin", "--output-as=xml -F regr-test -t 3")
test.add_stonith_neg_log_pattern("For stonith action (off) for victim regr-test, adding nodeid")
### verify nodeid use doesn't explode standalone mode
test = self.new_test("standalone_do_not_supply_nodeid",
"Verify nodeid in metadata parameter list doesn't kill standalone mode",
0)
test.add_cmd("stonith_admin",
"--output-as=xml -R true1 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=%s\"" % (our_uname))
test.add_cmd("stonith_admin", "--output-as=xml -F %s -t 3" % (our_uname))
test.add_stonith_neg_log_pattern("For stonith action (off) for victim %s, adding nodeid" % (our_uname))
def build_unfence_tests(self):
""" Register tests that verify unfencing """
our_uname = localname()
### verify unfencing using automatic unfencing
test = self.new_test("cpg_unfence_required_1",
"Verify require unfencing on all devices when automatic=true in agent's metadata",
1)
test.add_cmd('stonith_admin',
'--output-as=xml -R true1 -a fence_dummy_auto_unfence -o "mode=pass" -o "pcmk_host_list=%s"' % (our_uname))
test.add_cmd('stonith_admin',
'--output-as=xml -R true2 -a fence_dummy_auto_unfence -o "mode=pass" -o "pcmk_host_list=%s"' % (our_uname))
test.add_cmd("stonith_admin", "--output-as=xml -U %s -t 3" % (our_uname))
# both devices should be executed
test.add_stonith_log_pattern("with device 'true1' returned: 0 (OK)")
test.add_stonith_log_pattern("with device 'true2' returned: 0 (OK)")
### verify unfencing using automatic unfencing fails if any of the required agents fail
test = self.new_test("cpg_unfence_required_2",
"Verify require unfencing on all devices when automatic=true in agent's metadata",
1)
test.add_cmd('stonith_admin',
'--output-as=xml -R true1 -a fence_dummy_auto_unfence -o "mode=pass" -o "pcmk_host_list=%s"' % (our_uname))
test.add_cmd('stonith_admin',
'--output-as=xml -R true2 -a fence_dummy_auto_unfence -o "mode=fail" -o "pcmk_host_list=%s"' % (our_uname))
test.add_expected_fail_cmd("stonith_admin", "--output-as=xml -U %s -t 6" % (our_uname), CrmExit.ERROR)
### verify unfencing using automatic devices with topology
test = self.new_test("cpg_unfence_required_3",
"Verify require unfencing on all devices even when at different topology levels",
1)
test.add_cmd('stonith_admin',
'--output-as=xml -R true1 -a fence_dummy_auto_unfence -o "mode=pass" -o "pcmk_host_list=%s node3"' % (our_uname))
test.add_cmd('stonith_admin',
'--output-as=xml -R true2 -a fence_dummy_auto_unfence -o "mode=pass" -o "pcmk_host_list=%s node3"' % (our_uname))
test.add_cmd("stonith_admin", "--output-as=xml -r %s -i 1 -v true1" % (our_uname))
test.add_cmd("stonith_admin", "--output-as=xml -r %s -i 2 -v true2" % (our_uname))
test.add_cmd("stonith_admin", "--output-as=xml -U %s -t 3" % (our_uname))
test.add_stonith_log_pattern("with device 'true1' returned: 0 (OK)")
test.add_stonith_log_pattern("with device 'true2' returned: 0 (OK)")
### verify unfencing using automatic devices with topology
test = self.new_test("cpg_unfence_required_4",
"Verify all required devices are executed even with topology levels fail.",
1)
test.add_cmd('stonith_admin',
'--output-as=xml -R true1 -a fence_dummy_auto_unfence -o "mode=pass" -o "pcmk_host_list=%s node3"' % (our_uname))
test.add_cmd('stonith_admin',
'--output-as=xml -R true2 -a fence_dummy_auto_unfence -o "mode=pass" -o "pcmk_host_list=%s node3"' % (our_uname))
test.add_cmd('stonith_admin',
'--output-as=xml -R true3 -a fence_dummy_auto_unfence -o "mode=pass" -o "pcmk_host_list=%s node3"' % (our_uname))
test.add_cmd('stonith_admin',
'--output-as=xml -R true4 -a fence_dummy_auto_unfence -o "mode=pass" -o "pcmk_host_list=%s node3"' % (our_uname))
test.add_cmd('stonith_admin',
'--output-as=xml -R false1 -a fence_dummy -o "mode=fail" -o "pcmk_host_list=%s node3"' % (our_uname))
test.add_cmd('stonith_admin',
'--output-as=xml -R false2 -a fence_dummy -o "mode=fail" -o "pcmk_host_list=%s node3"' % (our_uname))
test.add_cmd('stonith_admin',
'--output-as=xml -R false3 -a fence_dummy -o "mode=fail" -o "pcmk_host_list=%s node3"' % (our_uname))
test.add_cmd('stonith_admin',
'--output-as=xml -R false4 -a fence_dummy -o "mode=fail" -o "pcmk_host_list=%s node3"' % (our_uname))
test.add_cmd("stonith_admin", "--output-as=xml -r %s -i 1 -v true1" % (our_uname))
test.add_cmd("stonith_admin", "--output-as=xml -r %s -i 1 -v false1" % (our_uname))
test.add_cmd("stonith_admin", "--output-as=xml -r %s -i 2 -v false2" % (our_uname))
test.add_cmd("stonith_admin", "--output-as=xml -r %s -i 2 -v true2" % (our_uname))
test.add_cmd("stonith_admin", "--output-as=xml -r %s -i 2 -v false3" % (our_uname))
test.add_cmd("stonith_admin", "--output-as=xml -r %s -i 2 -v true3" % (our_uname))
test.add_cmd("stonith_admin", "--output-as=xml -r %s -i 3 -v false4" % (our_uname))
test.add_cmd("stonith_admin", "--output-as=xml -r %s -i 4 -v true4" % (our_uname))
test.add_cmd("stonith_admin", "--output-as=xml -U %s -t 3" % (our_uname))
test.add_stonith_log_pattern("with device 'true1' returned: 0 (OK)")
test.add_stonith_log_pattern("with device 'true2' returned: 0 (OK)")
test.add_stonith_log_pattern("with device 'true3' returned: 0 (OK)")
test.add_stonith_log_pattern("with device 'true4' returned: 0 (OK)")
def build_unfence_on_target_tests(self):
""" Register tests that verify unfencing that runs on the target """
our_uname = localname()
### verify unfencing using on_target device
test = self.new_test("cpg_unfence_on_target_1",
"Verify unfencing with on_target = true", 1)
test.add_cmd("stonith_admin",
"--output-as=xml -R true1 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=%s\"" % (our_uname))
test.add_cmd("stonith_admin", "--output-as=xml -U %s -t 3" % (our_uname))
test.add_stonith_log_pattern("(on) to be executed on the target node")
### verify failure of unfencing using on_target device
test = self.new_test("cpg_unfence_on_target_2",
"Verify failure unfencing with on_target = true",
1)
test.add_cmd("stonith_admin",
"--output-as=xml -R true1 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=%s node_fake_1234\"" % (our_uname))
test.add_expected_fail_cmd("stonith_admin", "--output-as=xml -U node_fake_1234 -t 3", CrmExit.ERROR)
test.add_stonith_log_pattern("(on) to be executed on the target node")
### verify unfencing using on_target device with topology
test = self.new_test("cpg_unfence_on_target_3",
"Verify unfencing with on_target = true using topology",
1)
test.add_cmd("stonith_admin",
"--output-as=xml -R true1 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=%s node3\"" % (our_uname))
test.add_cmd("stonith_admin",
"--output-as=xml -R true2 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=%s node3\"" % (our_uname))
test.add_cmd("stonith_admin", "--output-as=xml -r %s -i 1 -v true1" % (our_uname))
test.add_cmd("stonith_admin", "--output-as=xml -r %s -i 2 -v true2" % (our_uname))
test.add_cmd("stonith_admin", "--output-as=xml -U %s -t 3" % (our_uname))
test.add_stonith_log_pattern("(on) to be executed on the target node")
### verify unfencing using on_target device with topology fails when victim node doesn't exist
test = self.new_test("cpg_unfence_on_target_4",
"Verify unfencing failure with on_target = true using topology",
1)
test.add_cmd("stonith_admin",
"--output-as=xml -R true1 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=%s node_fake\"" % (our_uname))
test.add_cmd("stonith_admin",
"--output-as=xml -R true2 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=%s node_fake\"" % (our_uname))
test.add_cmd("stonith_admin", "--output-as=xml -r node_fake -i 1 -v true1")
test.add_cmd("stonith_admin", "--output-as=xml -r node_fake -i 2 -v true2")
test.add_expected_fail_cmd("stonith_admin", "--output-as=xml -U node_fake -t 3", CrmExit.ERROR)
test.add_stonith_log_pattern("(on) to be executed on the target node")
def build_remap_tests(self):
""" Register tests that verify remapping of reboots to off-on """
test = self.new_test("cpg_remap_simple",
"Verify sequential topology reboot is remapped to all-off-then-all-on", 1)
test.add_cmd("stonith_admin",
"""--output-as=xml -R true1 -a fence_dummy -o "mode=pass" -o "pcmk_host_list=node_fake" """
"""-o "pcmk_off_timeout=1" -o "pcmk_reboot_timeout=10" """)
test.add_cmd("stonith_admin",
"""--output-as=xml -R true2 -a fence_dummy -o "mode=pass" -o "pcmk_host_list=node_fake" """
"""-o "pcmk_off_timeout=2" -o "pcmk_reboot_timeout=20" """)
test.add_cmd("stonith_admin", "--output-as=xml -r node_fake -i 1 -v true1 -v true2")
test.add_cmd("stonith_admin", "--output-as=xml -B node_fake -t 5")
test.add_stonith_log_pattern("Remapping multiple-device reboot targeting node_fake")
# timeout should be sum of off timeouts (1+2=3), not reboot timeouts (10+20=30)
test.add_stonith_log_pattern("Total timeout set to 3 for peer's fencing targeting node_fake")
test.add_stonith_log_pattern("perform 'off' action targeting node_fake using 'true1'")
test.add_stonith_log_pattern("perform 'off' action targeting node_fake using 'true2'")
test.add_stonith_log_pattern("Remapped 'off' targeting node_fake complete, remapping to 'on'")
# fence_dummy sets "on" as an on_target action
test.add_stonith_log_pattern("Ignoring true1 'on' failure (no capable peers) targeting node_fake")
test.add_stonith_log_pattern("Ignoring true2 'on' failure (no capable peers) targeting node_fake")
test.add_stonith_log_pattern("Undoing remap of reboot targeting node_fake")
test = self.new_test("cpg_remap_automatic",
"Verify remapped topology reboot skips automatic 'on'", 1)
test.add_cmd("stonith_admin",
"""--output-as=xml -R true1 -a fence_dummy_auto_unfence """
"""-o "mode=pass" -o "pcmk_host_list=node_fake" """)
test.add_cmd("stonith_admin",
"""--output-as=xml -R true2 -a fence_dummy_auto_unfence """
"""-o "mode=pass" -o "pcmk_host_list=node_fake" """)
test.add_cmd("stonith_admin", "--output-as=xml -r node_fake -i 1 -v true1 -v true2")
test.add_cmd("stonith_admin", "--output-as=xml -B node_fake -t 5")
test.add_stonith_log_pattern("Remapping multiple-device reboot targeting node_fake")
test.add_stonith_log_pattern("perform 'off' action targeting node_fake using 'true1'")
test.add_stonith_log_pattern("perform 'off' action targeting node_fake using 'true2'")
test.add_stonith_log_pattern("Remapped 'off' targeting node_fake complete, remapping to 'on'")
test.add_stonith_log_pattern("Undoing remap of reboot targeting node_fake")
test.add_stonith_neg_log_pattern("perform 'on' action targeting node_fake using")
test.add_stonith_neg_log_pattern("'on' failure")
test = self.new_test("cpg_remap_complex_1",
"Verify remapped topology reboot in second level works if non-remapped first level fails",
1)
test.add_cmd("stonith_admin", """--output-as=xml -R false1 -a fence_dummy -o "mode=fail" -o "pcmk_host_list=node_fake" """)
test.add_cmd("stonith_admin", """--output-as=xml -R true1 -a fence_dummy -o "mode=pass" -o "pcmk_host_list=node_fake" """)
test.add_cmd("stonith_admin", """--output-as=xml -R true2 -a fence_dummy -o "mode=pass" -o "pcmk_host_list=node_fake" """)
test.add_cmd("stonith_admin", "--output-as=xml -r node_fake -i 1 -v false1")
test.add_cmd("stonith_admin", "--output-as=xml -r node_fake -i 2 -v true1 -v true2")
test.add_cmd("stonith_admin", "--output-as=xml -B node_fake -t 5")
test.add_stonith_log_pattern("perform 'reboot' action targeting node_fake using 'false1'")
test.add_stonith_log_pattern("Remapping multiple-device reboot targeting node_fake")
test.add_stonith_log_pattern("perform 'off' action targeting node_fake using 'true1'")
test.add_stonith_log_pattern("perform 'off' action targeting node_fake using 'true2'")
test.add_stonith_log_pattern("Remapped 'off' targeting node_fake complete, remapping to 'on'")
test.add_stonith_log_pattern("Ignoring true1 'on' failure (no capable peers) targeting node_fake")
test.add_stonith_log_pattern("Ignoring true2 'on' failure (no capable peers) targeting node_fake")
test.add_stonith_log_pattern("Undoing remap of reboot targeting node_fake")
test = self.new_test("cpg_remap_complex_2",
"Verify remapped topology reboot failure in second level proceeds to third level",
1)
test.add_cmd("stonith_admin", """--output-as=xml -R false1 -a fence_dummy -o "mode=fail" -o "pcmk_host_list=node_fake" """)
test.add_cmd("stonith_admin", """--output-as=xml -R false2 -a fence_dummy -o "mode=fail" -o "pcmk_host_list=node_fake" """)
test.add_cmd("stonith_admin", """--output-as=xml -R true1 -a fence_dummy -o "mode=pass" -o "pcmk_host_list=node_fake" """)
test.add_cmd("stonith_admin", """--output-as=xml -R true2 -a fence_dummy -o "mode=pass" -o "pcmk_host_list=node_fake" """)
test.add_cmd("stonith_admin", """--output-as=xml -R true3 -a fence_dummy -o "mode=pass" -o "pcmk_host_list=node_fake" """)
test.add_cmd("stonith_admin", "--output-as=xml -r node_fake -i 1 -v false1")
test.add_cmd("stonith_admin", "--output-as=xml -r node_fake -i 2 -v true1 -v false2 -v true3")
test.add_cmd("stonith_admin", "--output-as=xml -r node_fake -i 3 -v true2")
test.add_cmd("stonith_admin", "--output-as=xml -B node_fake -t 5")
test.add_stonith_log_pattern("perform 'reboot' action targeting node_fake using 'false1'")
test.add_stonith_log_pattern("Remapping multiple-device reboot targeting node_fake")
test.add_stonith_log_pattern("perform 'off' action targeting node_fake using 'true1'")
test.add_stonith_log_pattern("perform 'off' action targeting node_fake using 'false2'")
test.add_stonith_log_pattern("Attempted to execute agent fence_dummy (off) the maximum number of times")
test.add_stonith_log_pattern("Undoing remap of reboot targeting node_fake")
test.add_stonith_log_pattern("perform 'reboot' action targeting node_fake using 'true2'")
test.add_stonith_neg_log_pattern("node_fake with true3")
def setup_environment(self, use_corosync):
""" Prepare the host before executing any tests """
if use_corosync:
if self.autogen_corosync_cfg:
(handle, self.autogen_corosync_log) = tempfile.mkstemp(prefix="cts-fencing-",
suffix=".corosync.log")
os.close(handle)
corosync_cfg = io.open("/etc/corosync/corosync.conf", "w")
corosync_cfg.write(AUTOGEN_COROSYNC_TEMPLATE % (localname(), self.autogen_corosync_log))
corosync_cfg.close()
### make sure we are in control ###
killall("corosync")
self.start_corosync()
subprocess.call(["cts-support", "install"])
def cleanup_environment(self, use_corosync):
""" Clean up the host after executing desired tests """
if use_corosync:
killall("corosync")
if self.autogen_corosync_cfg:
if self.verbose:
print("Corosync output")
logfile = io.open(self.autogen_corosync_log, 'rt')
for line in logfile.readlines():
print(line.strip())
logfile.close()
os.remove(self.autogen_corosync_log)
os.remove("/etc/corosync/corosync.conf")
subprocess.call(["cts-support", "uninstall"])
class TestOptions(object):
""" Option handler """
def __init__(self):
self.options = {}
self.options['list-tests'] = 0
self.options['run-all'] = 1
self.options['run-only'] = ""
self.options['run-only-pattern'] = ""
self.options['verbose'] = 0
self.options['invalid-arg'] = ""
self.options['cpg-only'] = 0
self.options['no-cpg'] = 0
self.options['show-usage'] = 0
def build_options(self, argv):
""" Set options based on command-line arguments """
args = argv[1:]
skip = 0
for i in range(0, len(args)):
if skip:
skip = 0
continue
elif args[i] == "-h" or args[i] == "--help":
self.options['show-usage'] = 1
elif args[i] == "-l" or args[i] == "--list-tests":
self.options['list-tests'] = 1
elif args[i] == "-V" or args[i] == "--verbose":
self.options['verbose'] = 1
elif args[i] == "-n" or args[i] == "--no-cpg":
self.options['no-cpg'] = 1
elif args[i] == "-c" or args[i] == "--cpg-only":
self.options['cpg-only'] = 1
elif args[i] == "-r" or args[i] == "--run-only":
self.options['run-only'] = args[i+1]
skip = 1
elif args[i] == "-p" or args[i] == "--run-only-pattern":
self.options['run-only-pattern'] = args[i+1]
skip = 1
def show_usage(self):
""" Show command usage """
print("usage: " + sys.argv[0] + " [options]")
print("If no options are provided, all tests will run")
print("Options:")
print("\t [--help | -h] Show usage")
print("\t [--list-tests | -l] Print out all registered tests.")
print("\t [--cpg-only | -c] Only run tests that require corosync.")
print("\t [--no-cpg | -n] Only run tests that do not require corosync")
print("\t [--run-only | -r 'testname'] Run a specific test")
print("\t [--verbose | -V] Verbose output")
print("\t [--run-only-pattern | -p 'string'] Run only tests containing the string value")
print("\n\tExample: Run only the test 'start_stop'")
print("\t\t " + sys.argv[0] + " --run-only start_stop")
print("\n\tExample: Run only the tests with the string 'systemd' present in them")
print("\t\t " + sys.argv[0] + " --run-only-pattern systemd")
def main(argv):
""" Run fencing regression tests as specified by arguments """
update_path()
opts = TestOptions()
opts.build_options(argv)
use_corosync = 1
tests = Tests(opts.options['verbose'])
tests.build_standalone_tests()
tests.build_custom_timeout_tests()
tests.build_api_sanity_tests()
tests.build_fence_merge_tests()
tests.build_fence_no_merge_tests()
tests.build_unfence_tests()
tests.build_unfence_on_target_tests()
tests.build_nodeid_tests()
tests.build_remap_tests()
if opts.options['list-tests']:
tests.print_list()
sys.exit(CrmExit.OK)
elif opts.options['show-usage']:
opts.show_usage()
sys.exit(CrmExit.OK)
print("Starting ...")
if opts.options['no-cpg']:
use_corosync = 0
tests.setup_environment(use_corosync)
if opts.options['run-only-pattern'] != "":
tests.run_tests_matching(opts.options['run-only-pattern'])
tests.print_results()
elif opts.options['run-only'] != "":
tests.run_single(opts.options['run-only'])
tests.print_results()
elif opts.options['no-cpg']:
tests.run_no_cpg()
tests.print_results()
elif opts.options['cpg-only']:
tests.run_cpg_only()
tests.print_results()
else:
tests.run_tests()
tests.print_results()
tests.cleanup_environment(use_corosync)
tests.exit()
if __name__ == "__main__":
main(sys.argv)
diff --git a/cts/cts-scheduler.in b/cts/cts-scheduler.in
index f2957bad66..99436f1bb8 100644
--- a/cts/cts-scheduler.in
+++ b/cts/cts-scheduler.in
@@ -1,1461 +1,1462 @@
#!@PYTHON@
""" Regression tests for Pacemaker's scheduler
"""
# Pacemaker targets compatibility with Python 2.7 and 3.2+
from __future__ import print_function, unicode_literals, absolute_import, division
__copyright__ = "Copyright 2004-2020 the Pacemaker project contributors"
__license__ = "GNU General Public License version 2 or later (GPLv2+) WITHOUT ANY WARRANTY"
import io
import os
import re
import sys
import stat
import shlex
import shutil
import argparse
import subprocess
import platform
DESC = """Regression tests for Pacemaker's scheduler"""
# Each entry in TESTS is a group of tests, where each test consists of a
# test base name, test description, and additional test arguments.
# Test groups will be separated by newlines in output.
TESTS = [
[
[ "simple1", "Offline" ],
[ "simple2", "Start" ],
[ "simple3", "Start 2" ],
[ "simple4", "Start Failed" ],
[ "simple6", "Stop Start" ],
[ "simple7", "Shutdown" ],
#[ "simple8", "Stonith" ],
#[ "simple9", "Lower version" ],
#[ "simple10", "Higher version" ],
[ "simple11", "Priority (ne)" ],
[ "simple12", "Priority (eq)" ],
[ "simple8", "Stickiness" ],
],
[
[ "group1", "Group" ],
[ "group2", "Group + Native" ],
[ "group3", "Group + Group" ],
[ "group4", "Group + Native (nothing)" ],
[ "group5", "Group + Native (move)" ],
[ "group6", "Group + Group (move)" ],
[ "group7", "Group colocation" ],
[ "group13", "Group colocation (cant run)" ],
[ "group8", "Group anti-colocation" ],
[ "group9", "Group recovery" ],
[ "group10", "Group partial recovery" ],
[ "group11", "Group target_role" ],
[ "group14", "Group stop (graph terminated)" ],
[ "group15", "Negative group colocation" ],
[ "bug-1573", "Partial stop of a group with two children" ],
[ "bug-1718", "Mandatory group ordering - Stop group_FUN" ],
[ "bug-lf-2613", "Move group on failure" ],
[ "bug-lf-2619", "Move group on clone failure" ],
[ "group-fail", "Ensure stop order is preserved for partially active groups" ],
[ "group-unmanaged", "No need to restart r115 because r114 is unmanaged" ],
[ "group-unmanaged-stopped", "Make sure r115 is stopped when r114 fails" ],
[ "group-dependents", "Account for the location preferences of things colocated with a group" ],
[ "group-stop-ordering", "Ensure blocked group member stop does not force other member stops" ],
],
[
[ "rsc_dep1", "Must not" ],
[ "rsc_dep3", "Must" ],
[ "rsc_dep5", "Must not 3" ],
[ "rsc_dep7", "Must 3" ],
[ "rsc_dep10", "Must (but cant)" ],
[ "rsc_dep2", "Must (running)" ],
[ "rsc_dep8", "Must (running : alt)" ],
[ "rsc_dep4", "Must (running + move)" ],
[ "asymmetric", "Asymmetric - require explicit location constraints" ],
],
[
[ "orphan-0", "Orphan ignore" ],
[ "orphan-1", "Orphan stop" ],
[ "orphan-2", "Orphan stop, remove failcount" ],
],
[
[ "params-0", "Params: No change" ],
[ "params-1", "Params: Changed" ],
[ "params-2", "Params: Resource definition" ],
[ "params-4", "Params: Reload" ],
[ "params-5", "Params: Restart based on probe digest" ],
[ "novell-251689", "Resource definition change + target_role=stopped" ],
[ "bug-lf-2106", "Restart all anonymous clone instances after config change" ],
[ "params-6", "Params: Detect reload in previously migrated resource" ],
[ "nvpair-id-ref", "Support id-ref in nvpair with optional name" ],
[ "not-reschedule-unneeded-monitor",
"Do not reschedule unneeded monitors while resource definitions have changed" ],
[ "reload-becomes-restart", "Cancel reload if restart becomes required" ],
],
[
[ "target-0", "Target Role : baseline" ],
[ "target-1", "Target Role : master" ],
[ "target-2", "Target Role : invalid" ],
],
[
[ "base-score", "Set a node's default score for all nodes" ],
],
[
[ "date-1", "Dates", [ "-t", "2005-020" ] ],
[ "date-2", "Date Spec - Pass", [ "-t", "2005-020T12:30" ] ],
[ "date-3", "Date Spec - Fail", [ "-t", "2005-020T11:30" ] ],
[ "origin", "Timing of recurring operations", [ "-t", "2014-05-07 00:28:00" ] ],
[ "probe-0", "Probe (anon clone)" ],
[ "probe-1", "Pending Probe" ],
[ "probe-2", "Correctly re-probe cloned groups" ],
[ "probe-3", "Probe (pending node)" ],
[ "probe-4", "Probe (pending node + stopped resource)" ],
[ "standby", "Standby" ],
[ "comments", "Comments" ],
],
[
[ "one-or-more-0", "Everything starts" ],
[ "one-or-more-1", "Nothing starts because of A" ],
[ "one-or-more-2", "D can start because of C" ],
[ "one-or-more-3", "D cannot start because of B and C" ],
[ "one-or-more-4", "D cannot start because of target-role" ],
[ "one-or-more-5", "Start A and F even though C and D are stopped" ],
[ "one-or-more-6", "Leave A running even though B is stopped" ],
[ "one-or-more-7", "Leave A running even though C is stopped" ],
[ "bug-5140-require-all-false", "Allow basegrp:0 to stop" ],
[ "clone-require-all-1", "clone B starts node 3 and 4" ],
[ "clone-require-all-2", "clone B remains stopped everywhere" ],
[ "clone-require-all-3", "clone B stops everywhere because A stops everywhere" ],
[ "clone-require-all-4", "clone B remains on node 3 and 4 with only one instance of A remaining" ],
[ "clone-require-all-5", "clone B starts on node 1 3 and 4" ],
[ "clone-require-all-6", "clone B remains active after shutting down instances of A" ],
[ "clone-require-all-7",
"clone A and B both start at the same time. all instances of A start before B" ],
[ "clone-require-all-no-interleave-1", "C starts everywhere after A and B" ],
[ "clone-require-all-no-interleave-2",
"C starts on nodes 1, 2, and 4 with only one active instance of B" ],
[ "clone-require-all-no-interleave-3",
"C remains active when instance of B is stopped on one node and started on another" ],
[ "one-or-more-unrunnable-instances", "Avoid dependencies on instances that won't ever be started" ],
],
[
[ "location-date-rules-1", "Use location constraints with ineffective date-based rules" ],
[ "location-date-rules-2", "Use location constraints with effective date-based rules" ],
[ "nvpair-date-rules-1", "Use nvpair blocks with a variety of date-based rules" ],
],
[
[ "order1", "Order start 1" ],
[ "order2", "Order start 2" ],
[ "order3", "Order stop" ],
[ "order4", "Order (multiple)" ],
[ "order5", "Order (move)" ],
[ "order6", "Order (move w/ restart)" ],
[ "order7", "Order (mandatory)" ],
[ "order-optional", "Order (score=0)" ],
[ "order-required", "Order (score=INFINITY)" ],
[ "bug-lf-2171", "Prevent group start when clone is stopped" ],
[ "order-clone", "Clone ordering should be able to prevent startup of dependent clones" ],
[ "order-sets", "Ordering for resource sets" ],
[ "order-serialize", "Serialize resources without inhibiting migration" ],
[ "order-serialize-set", "Serialize a set of resources without inhibiting migration" ],
[ "clone-order-primitive", "Order clone start after a primitive" ],
[ "clone-order-16instances", "Verify ordering of 16 cloned resources" ],
[ "order-optional-keyword", "Order (optional keyword)" ],
[ "order-mandatory", "Order (mandatory keyword)" ],
[ "bug-lf-2493",
"Don't imply colocation requirements when applying ordering constraints with clones" ],
[ "ordered-set-basic-startup", "Constraint set with default order settings" ],
[ "ordered-set-natural", "Allow natural set ordering" ],
[ "order-wrong-kind", "Order (error)" ],
],
[
[ "coloc-loop", "Colocation - loop" ],
[ "coloc-many-one", "Colocation - many-to-one" ],
[ "coloc-list", "Colocation - many-to-one with list" ],
[ "coloc-group", "Colocation - groups" ],
[ "coloc-slave-anti", "Anti-colocation with slave shouldn't prevent master colocation" ],
[ "coloc-attr", "Colocation based on node attributes" ],
[ "coloc-negative-group", "Negative colocation with a group" ],
[ "coloc-intra-set", "Intra-set colocation" ],
[ "bug-lf-2435", "Colocation sets with a negative score" ],
[ "coloc-clone-stays-active",
"Ensure clones don't get stopped/demoted because a dependent must stop" ],
[ "coloc_fp_logic", "Verify floating point calculations in colocation are working" ],
[ "colo_master_w_native",
"cl#5070 - Verify promotion order is affected when colocating master to native rsc" ],
[ "colo_slave_w_native",
"cl#5070 - Verify promotion order is affected when colocating slave to native rsc" ],
[ "anti-colocation-order",
"cl#5187 - Prevent resources in an anti-colocation from even temporarily running on a same node" ],
[ "anti-colocation-master", "Organize order of actions for master resources in anti-colocations" ],
[ "anti-colocation-slave", "Organize order of actions for slave resources in anti-colocations" ],
[ "enforce-colo1", "Always enforce B with A INFINITY" ],
[ "complex_enforce_colo", "Always enforce B with A INFINITY. (make sure heat-engine stops)" ],
],
[
[ "rsc-sets-seq-true", "Resource Sets - sequential=false" ],
[ "rsc-sets-seq-false", "Resource Sets - sequential=true" ],
[ "rsc-sets-clone", "Resource Sets - Clone" ],
[ "rsc-sets-master", "Resource Sets - Master" ],
[ "rsc-sets-clone-1", "Resource Sets - Clone (lf#2404)" ],
],
[
[ "attrs1", "string: eq (and)" ],
[ "attrs2", "string: lt / gt (and)" ],
[ "attrs3", "string: ne (or)" ],
[ "attrs4", "string: exists" ],
[ "attrs5", "string: not_exists" ],
[ "attrs6", "is_dc: true" ],
[ "attrs7", "is_dc: false" ],
[ "attrs8", "score_attribute" ],
[ "per-node-attrs", "Per node resource parameters" ],
],
[
[ "mon-rsc-1", "Schedule Monitor - start" ],
[ "mon-rsc-2", "Schedule Monitor - move" ],
[ "mon-rsc-3", "Schedule Monitor - pending start" ],
[ "mon-rsc-4", "Schedule Monitor - move/pending start" ],
],
[
[ "rec-rsc-0", "Resource Recover - no start" ],
[ "rec-rsc-1", "Resource Recover - start" ],
[ "rec-rsc-2", "Resource Recover - monitor" ],
[ "rec-rsc-3", "Resource Recover - stop - ignore" ],
[ "rec-rsc-4", "Resource Recover - stop - block" ],
[ "rec-rsc-5", "Resource Recover - stop - fence" ],
[ "rec-rsc-6", "Resource Recover - multiple - restart" ],
[ "rec-rsc-7", "Resource Recover - multiple - stop" ],
[ "rec-rsc-8", "Resource Recover - multiple - block" ],
[ "rec-rsc-9", "Resource Recover - group/group" ],
[ "monitor-recovery", "on-fail=block + resource recovery detected by recurring monitor" ],
[ "stop-failure-no-quorum", "Stop failure without quorum" ],
[ "stop-failure-no-fencing", "Stop failure without fencing available" ],
[ "stop-failure-with-fencing", "Stop failure with fencing available" ],
[ "multiple-active-block-group", "Support of multiple-active=block for resource groups" ],
[ "multiple-monitor-one-failed",
"Consider resource failed if any of the configured monitor operations failed" ],
],
[
[ "quorum-1", "No quorum - ignore" ],
[ "quorum-2", "No quorum - freeze" ],
[ "quorum-3", "No quorum - stop" ],
[ "quorum-4", "No quorum - start anyway" ],
[ "quorum-5", "No quorum - start anyway (group)" ],
[ "quorum-6", "No quorum - start anyway (clone)" ],
[ "bug-cl-5212", "No promotion with no-quorum-policy=freeze" ],
[ "suicide-needed-inquorate", "no-quorum-policy=suicide: suicide necessary" ],
[ "suicide-not-needed-initial-quorum",
"no-quorum-policy=suicide: suicide not necessary at initial quorum" ],
[ "suicide-not-needed-never-quorate",
"no-quorum-policy=suicide: suicide not necessary if never quorate" ],
[ "suicide-not-needed-quorate", "no-quorum-policy=suicide: suicide necessary if quorate" ],
],
[
[ "rec-node-1", "Node Recover - Startup - no fence" ],
[ "rec-node-2", "Node Recover - Startup - fence" ],
[ "rec-node-3", "Node Recover - HA down - no fence" ],
[ "rec-node-4", "Node Recover - HA down - fence" ],
[ "rec-node-5", "Node Recover - CRM down - no fence" ],
[ "rec-node-6", "Node Recover - CRM down - fence" ],
[ "rec-node-7", "Node Recover - no quorum - ignore" ],
[ "rec-node-8", "Node Recover - no quorum - freeze" ],
[ "rec-node-9", "Node Recover - no quorum - stop" ],
[ "rec-node-10", "Node Recover - no quorum - stop w/fence" ],
[ "rec-node-11", "Node Recover - CRM down w/ group - fence" ],
[ "rec-node-12", "Node Recover - nothing active - fence" ],
[ "rec-node-13", "Node Recover - failed resource + shutdown - fence" ],
[ "rec-node-15", "Node Recover - unknown lrm section" ],
[ "rec-node-14", "Serialize all stonith's" ],
],
[
[ "multi1", "Multiple Active (stop/start)" ],
],
[
[ "migrate-begin", "Normal migration" ],
[ "migrate-success", "Completed migration" ],
[ "migrate-partial-1", "Completed migration, missing stop on source" ],
[ "migrate-partial-2", "Successful migrate_to only" ],
[ "migrate-partial-3", "Successful migrate_to only, target down" ],
[ "migrate-partial-4", "Migrate from the correct host after migrate_to+migrate_from" ],
[ "bug-5186-partial-migrate", "Handle partial migration when src node loses membership" ],
[ "migrate-fail-2", "Failed migrate_from" ],
[ "migrate-fail-3", "Failed migrate_from + stop on source" ],
[ "migrate-fail-4",
"Failed migrate_from + stop on target - ideally we wouldn't need to re-stop on target" ],
[ "migrate-fail-5", "Failed migrate_from + stop on source and target" ],
[ "migrate-fail-6", "Failed migrate_to" ],
[ "migrate-fail-7", "Failed migrate_to + stop on source" ],
[ "migrate-fail-8",
"Failed migrate_to + stop on target - ideally we wouldn't need to re-stop on target" ],
[ "migrate-fail-9", "Failed migrate_to + stop on source and target" ],
[ "migration-ping-pong", "Old migrate_to failure + successful migrate_from on same node" ],
[ "migrate-stop", "Migration in a stopping stack" ],
[ "migrate-start", "Migration in a starting stack" ],
[ "migrate-stop_start", "Migration in a restarting stack" ],
[ "migrate-stop-complex", "Migration in a complex stopping stack" ],
[ "migrate-start-complex", "Migration in a complex starting stack" ],
[ "migrate-stop-start-complex", "Migration in a complex moving stack" ],
[ "migrate-shutdown", "Order the post-migration 'stop' before node shutdown" ],
[ "migrate-1", "Migrate (migrate)" ],
[ "migrate-2", "Migrate (stable)" ],
[ "migrate-3", "Migrate (failed migrate_to)" ],
[ "migrate-4", "Migrate (failed migrate_from)" ],
[ "novell-252693", "Migration in a stopping stack" ],
[ "novell-252693-2", "Migration in a starting stack" ],
[ "novell-252693-3", "Non-Migration in a starting and stopping stack" ],
[ "bug-1820", "Migration in a group" ],
[ "bug-1820-1", "Non-migration in a group" ],
[ "migrate-5", "Primitive migration with a clone" ],
[ "migrate-fencing", "Migration after Fencing" ],
[ "migrate-both-vms", "Migrate two VMs that have no colocation" ],
[ "migration-behind-migrating-remote", "Migrate resource behind migrating remote connection" ],
[ "1-a-then-bm-move-b", "Advanced migrate logic. A then B. migrate B" ],
[ "2-am-then-b-move-a", "Advanced migrate logic, A then B, migrate A without stopping B" ],
[ "3-am-then-bm-both-migrate", "Advanced migrate logic. A then B. migrate both" ],
[ "4-am-then-bm-b-not-migratable", "Advanced migrate logic, A then B, B not migratable" ],
[ "5-am-then-bm-a-not-migratable", "Advanced migrate logic. A then B. move both, a not migratable" ],
[ "6-migrate-group", "Advanced migrate logic, migrate a group" ],
[ "7-migrate-group-one-unmigratable",
"Advanced migrate logic, migrate group mixed with allow-migrate true/false" ],
[ "8-am-then-bm-a-migrating-b-stopping",
"Advanced migrate logic, A then B, A migrating, B stopping" ],
[ "9-am-then-bm-b-migrating-a-stopping",
"Advanced migrate logic, A then B, B migrate, A stopping" ],
[ "10-a-then-bm-b-move-a-clone",
"Advanced migrate logic, A clone then B, migrate B while stopping A" ],
[ "11-a-then-bm-b-move-a-clone-starting",
"Advanced migrate logic, A clone then B, B moving while A is start/stopping" ],
[ "a-promote-then-b-migrate", "A promote then B start. migrate B" ],
[ "a-demote-then-b-migrate", "A demote then B stop. migrate B" ],
# @TODO: If pacemaker implements versioned attributes, uncomment this test
#[ "migrate-versioned", "Disable migration for versioned resources" ],
[ "bug-lf-2422", "Dependency on partially active group - stop ocfs:*" ],
],
[
[ "clone-anon-probe-1", "Probe the correct (anonymous) clone instance for each node" ],
[ "clone-anon-probe-2", "Avoid needless re-probing of anonymous clones" ],
[ "clone-anon-failcount", "Merge failcounts for anonymous clones" ],
[ "force-anon-clone-max", "Update clone-max properly when forcing a clone to be anonymous" ],
[ "anon-instance-pending", "Assign anonymous clone instance numbers properly when action pending" ],
[ "inc0", "Incarnation start" ],
[ "inc1", "Incarnation start order" ],
[ "inc2", "Incarnation silent restart, stop, move" ],
[ "inc3", "Inter-incarnation ordering, silent restart, stop, move" ],
[ "inc4", "Inter-incarnation ordering, silent restart, stop, move (ordered)" ],
[ "inc5", "Inter-incarnation ordering, silent restart, stop, move (restart 1)" ],
[ "inc6", "Inter-incarnation ordering, silent restart, stop, move (restart 2)" ],
[ "inc7", "Clone colocation" ],
[ "inc8", "Clone anti-colocation" ],
[ "inc9", "Non-unique clone" ],
[ "inc10", "Non-unique clone (stop)" ],
[ "inc11", "Primitive colocation with clones" ],
[ "inc12", "Clone shutdown" ],
[ "cloned-group", "Make sure only the correct number of cloned groups are started" ],
[ "cloned-group-stop", "Ensure stopping qpidd also stops glance and cinder" ],
[ "clone-no-shuffle", "Don't prioritize allocation of instances that must be moved" ],
[ "clone-max-zero", "Orphan processing with clone-max=0" ],
[ "clone-anon-dup",
"Bug LF#2087 - Correctly parse the state of anonymous clones that are active more than once per node" ],
[ "bug-lf-2160", "Don't shuffle clones due to colocation" ],
[ "bug-lf-2213", "clone-node-max enforcement for cloned groups" ],
[ "bug-lf-2153", "Clone ordering constraints" ],
[ "bug-lf-2361", "Ensure clones observe mandatory ordering constraints if the LHS is unrunnable" ],
[ "bug-lf-2317", "Avoid needless restart of primitive depending on a clone" ],
[ "clone-colocate-instance-1", "Colocation with a specific clone instance (negative example)" ],
[ "clone-colocate-instance-2", "Colocation with a specific clone instance" ],
[ "clone-order-instance", "Ordering with specific clone instances" ],
[ "bug-lf-2453", "Enforce mandatory clone ordering without colocation" ],
[ "bug-lf-2508", "Correctly reconstruct the status of anonymous cloned groups" ],
[ "bug-lf-2544", "Balanced clone placement" ],
[ "bug-lf-2445", "Redistribute clones with node-max > 1 and stickiness = 0" ],
[ "bug-lf-2574", "Avoid clone shuffle" ],
[ "bug-lf-2581", "Avoid group restart due to unrelated clone (re)start" ],
[ "bug-cl-5168", "Don't shuffle clones" ],
[ "bug-cl-5170", "Prevent clone from starting with on-fail=block" ],
[ "clone-fail-block-colocation", "Move colocated group when failed clone has on-fail=block" ],
[ "clone-interleave-1",
"Clone-3 cannot start on pcmk-1 due to interleaved ordering (no colocation)" ],
[ "clone-interleave-2", "Clone-3 must stop on pcmk-1 due to interleaved ordering (no colocation)" ],
[ "clone-interleave-3",
"Clone-3 must be recovered on pcmk-1 due to interleaved ordering (no colocation)" ],
[ "rebalance-unique-clones", "Rebalance unique clone instances with no stickiness" ],
[ "clone-requires-quorum-recovery", "Clone with requires=quorum on failed node needing recovery" ],
[ "clone-requires-quorum",
"Clone with requires=quorum with presumed-inactive instance on failed node" ],
],
[
[ "cloned_start_one", "order first clone then clone... first clone_min=2" ],
[ "cloned_start_two", "order first clone then clone... first clone_min=2" ],
[ "cloned_stop_one", "order first clone then clone... first clone_min=2" ],
[ "cloned_stop_two", "order first clone then clone... first clone_min=2" ],
[ "clone_min_interleave_start_one",
"order first clone then clone... first clone_min=2 and then has interleave=true" ],
[ "clone_min_interleave_start_two",
"order first clone then clone... first clone_min=2 and then has interleave=true" ],
[ "clone_min_interleave_stop_one",
"order first clone then clone... first clone_min=2 and then has interleave=true" ],
[ "clone_min_interleave_stop_two",
"order first clone then clone... first clone_min=2 and then has interleave=true" ],
[ "clone_min_start_one", "order first clone then primitive... first clone_min=2" ],
[ "clone_min_start_two", "order first clone then primitive... first clone_min=2" ],
[ "clone_min_stop_all", "order first clone then primitive... first clone_min=2" ],
[ "clone_min_stop_one", "order first clone then primitive... first clone_min=2" ],
[ "clone_min_stop_two", "order first clone then primitive... first clone_min=2" ],
],
[
[ "unfence-startup", "Clean unfencing" ],
[ "unfence-definition", "Unfencing when the agent changes" ],
[ "unfence-parameters", "Unfencing when the agent parameters changes" ],
[ "unfence-device", "Unfencing when a cluster has only fence devices" ],
],
[
[ "master-0", "Stopped -> Slave" ],
[ "master-1", "Stopped -> Promote" ],
[ "master-2", "Stopped -> Promote : notify" ],
[ "master-3", "Stopped -> Promote : master location" ],
[ "master-4", "Started -> Promote : master location" ],
[ "master-5", "Promoted -> Promoted" ],
[ "master-6", "Promoted -> Promoted (2)" ],
[ "master-7", "Promoted -> Fenced" ],
[ "master-8", "Promoted -> Fenced -> Moved" ],
[ "master-9", "Stopped + Promotable + No quorum" ],
[ "master-10", "Stopped -> Promotable : notify with monitor" ],
[ "master-11", "Stopped -> Promote : colocation" ],
[ "novell-239082", "Demote/Promote ordering" ],
[ "novell-239087", "Stable master placement" ],
[ "master-12", "Promotion based solely on rsc_location constraints" ],
[ "master-13", "Include preferences of colocated resources when placing master" ],
[ "master-demote", "Ordering when actions depends on demoting a slave resource" ],
[ "master-ordering", "Prevent resources from starting that need a master" ],
[ "bug-1765", "Master-Master Colocation (do not stop the slaves)" ],
[ "master-group", "Promotion of cloned groups" ],
[ "bug-lf-1852", "Don't shuffle master/slave instances unnecessarily" ],
[ "master-failed-demote", "Don't retry failed demote actions" ],
[ "master-failed-demote-2", "Don't retry failed demote actions (notify=false)" ],
[ "master-depend",
"Ensure resources that depend on the master don't get allocated until the master does" ],
[ "master-reattach", "Re-attach to a running master" ],
[ "master-allow-start", "Don't include master score if it would prevent allocation" ],
[ "master-colocation",
"Allow master instances placemaker to be influenced by colocation constraints" ],
[ "master-pseudo", "Make sure promote/demote pseudo actions are created correctly" ],
[ "master-role", "Prevent target-role from promoting more than master-max instances" ],
[ "bug-lf-2358", "Master-Master anti-colocation" ],
[ "master-promotion-constraint", "Mandatory master colocation constraints" ],
[ "unmanaged-master", "Ensure role is preserved for unmanaged resources" ],
[ "master-unmanaged-monitor", "Start the correct monitor operation for unmanaged masters" ],
[ "master-demote-2", "Demote does not clear past failure" ],
[ "master-move", "Move master based on failure of colocated group" ],
[ "master-probed-score", "Observe the promotion score of probed resources" ],
[ "colocation_constraint_stops_master",
"cl#5054 - Ensure master is demoted when stopped by colocation constraint" ],
[ "colocation_constraint_stops_slave",
"cl#5054 - Ensure slave is not demoted when stopped by colocation constraint" ],
[ "order_constraint_stops_master",
"cl#5054 - Ensure master is demoted when stopped by order constraint" ],
[ "order_constraint_stops_slave",
"cl#5054 - Ensure slave is not demoted when stopped by order constraint" ],
[ "master_monitor_restart", "cl#5072 - Ensure master monitor operation will start after promotion" ],
[ "bug-rh-880249", "Handle replacement of an m/s resource with a primitive" ],
[ "bug-5143-ms-shuffle", "Prevent master shuffling due to promotion score" ],
[ "master-demote-block", "Block promotion if demote fails with on-fail=block" ],
[ "master-dependent-ban",
"Don't stop instances from being active because a dependent is banned from that host" ],
[ "master-stop", "Stop instances due to location constraint with role=Started" ],
[ "master-partially-demoted-group", "Allow partially demoted group to finish demoting" ],
[ "bug-cl-5213", "Ensure role colocation with -INFINITY is enforced" ],
[ "bug-cl-5219", "Allow unrelated resources with a common colocation target to remain promoted" ],
[ "master-asymmetrical-order",
"Fix the behaviors of multi-state resources with asymmetrical ordering" ],
[ "master-notify", "Master promotion with notifies" ],
[ "master-score-startup", "Use permanent master scores without LRM history" ],
[ "failed-demote-recovery", "Recover resource in slave role after demote fails" ],
[ "failed-demote-recovery-master", "Recover resource in master role after demote fails" ],
],
[
[ "history-1", "Correctly parse stateful-1 resource state" ],
],
[
[ "managed-0", "Managed (reference)" ],
[ "managed-1", "Not managed - down" ],
[ "managed-2", "Not managed - up" ],
[ "bug-5028", "Shutdown should block if anything depends on an unmanaged resource" ],
[ "bug-5028-detach", "Ensure detach still works" ],
[ "bug-5028-bottom",
"Ensure shutdown still blocks if the blocked resource is at the bottom of the stack" ],
[ "unmanaged-stop-1",
"cl#5155 - Block the stop of resources if any depending resource is unmanaged" ],
[ "unmanaged-stop-2",
"cl#5155 - Block the stop of resources if the first resource in a mandatory stop order is unmanaged" ],
[ "unmanaged-stop-3",
"cl#5155 - Block the stop of resources if any depending resource in a group is unmanaged" ],
[ "unmanaged-stop-4",
"cl#5155 - Block the stop of resources if any depending resource in the middle of a group is unmanaged" ],
[ "unmanaged-block-restart",
"Block restart of resources if any dependent resource in a group is unmanaged" ],
],
[
[ "interleave-0", "Interleave (reference)" ],
[ "interleave-1", "coloc - not interleaved" ],
[ "interleave-2", "coloc - interleaved" ],
[ "interleave-3", "coloc - interleaved (2)" ],
[ "interleave-pseudo-stop", "Interleaved clone during stonith" ],
[ "interleave-stop", "Interleaved clone during stop" ],
[ "interleave-restart", "Interleaved clone during dependency restart" ],
],
[
[ "notify-0", "Notify reference" ],
[ "notify-1", "Notify simple" ],
[ "notify-2", "Notify simple, confirm" ],
[ "notify-3", "Notify move, confirm" ],
[ "novell-239079", "Notification priority" ],
#[ "notify-2", "Notify - 764" ],
[ "notifs-for-unrunnable", "Don't schedule notifications for an unrunnable action" ],
[ "route-remote-notify", "Route remote notify actions through correct cluster node" ],
[ "notify-behind-stopping-remote", "Don't schedule notifications behind stopped remote" ],
],
[
[ "594", "OSDL #594 - Unrunnable actions scheduled in transition" ],
[ "662", "OSDL #662 - Two resources start on one node when incarnation_node_max = 1" ],
[ "696", "OSDL #696 - CRM starts stonith RA without monitor" ],
[ "726", "OSDL #726 - Attempting to schedule rsc_posic041_monitor_5000 _after_ a stop" ],
[ "735", "OSDL #735 - Correctly detect that rsc_hadev1 is stopped on hadev3" ],
[ "764", "OSDL #764 - Missing monitor op for DoFencing:child_DoFencing:1" ],
[ "797", "OSDL #797 - Assert triggered: task_id_i > max_call_id" ],
[ "829", "OSDL #829" ],
[ "994",
"OSDL #994 - Stopping the last resource in a resource group causes the entire group to be restarted" ],
[ "994-2", "OSDL #994 - with a dependent resource" ],
[ "1360", "OSDL #1360 - Clone stickiness" ],
[ "1484", "OSDL #1484 - on_fail=stop" ],
[ "1494", "OSDL #1494 - Clone stability" ],
[ "unrunnable-1", "Unrunnable" ],
[ "unrunnable-2", "Unrunnable 2" ],
[ "stonith-0", "Stonith loop - 1" ],
[ "stonith-1", "Stonith loop - 2" ],
[ "stonith-2", "Stonith loop - 3" ],
[ "stonith-3", "Stonith startup" ],
[ "stonith-4", "Stonith node state" ],
[ "dc-fence-ordering", "DC needs fencing while other nodes are shutting down" ],
[ "bug-1572-1", "Recovery of groups depending on master/slave" ],
[ "bug-1572-2", "Recovery of groups depending on master/slave when the master is never re-promoted" ],
[ "bug-1685", "Depends-on-master ordering" ],
[ "bug-1822", "Don't promote partially active groups" ],
[ "bug-pm-11", "New resource added to a m/s group" ],
[ "bug-pm-12", "Recover only the failed portion of a cloned group" ],
[ "bug-n-387749", "Don't shuffle clone instances" ],
[ "bug-n-385265",
"Don't ignore the failure stickiness of group children - resource_idvscommon should stay stopped" ],
[ "bug-n-385265-2",
"Ensure groups are migrated instead of remaining partially active on the current node" ],
[ "bug-lf-1920", "Correctly handle probes that find active resources" ],
[ "bnc-515172", "Location constraint with multiple expressions" ],
[ "colocate-primitive-with-clone", "Optional colocation with a clone" ],
[ "use-after-free-merge", "Use-after-free in native_merge_weights" ],
[ "bug-lf-2551", "STONITH ordering for stop" ],
[ "bug-lf-2606", "Stonith implies demote" ],
[ "bug-lf-2474", "Ensure resource op timeout takes precedence over op_defaults" ],
[ "bug-suse-707150", "Prevent vm-01 from starting due to colocation/ordering" ],
[ "bug-5014-A-start-B-start", "Verify when A starts B starts using symmetrical=false" ],
[ "bug-5014-A-stop-B-started",
"Verify when A stops B does not stop if it has already started using symmetric=false" ],
[ "bug-5014-A-stopped-B-stopped",
"Verify when A is stopped and B has not started, B does not start before A using symmetric=false" ],
[ "bug-5014-CthenAthenB-C-stopped",
"Verify when C then A is symmetrical=true, A then B is symmetric=false, and C is stopped that nothing starts" ],
[ "bug-5014-CLONE-A-start-B-start",
"Verify when A starts B starts using clone resources with symmetric=false" ],
[ "bug-5014-CLONE-A-stop-B-started",
"Verify when A stops B does not stop if it has already started using clone resources with symmetric=false" ],
[ "bug-5014-GROUP-A-start-B-start",
"Verify when A starts B starts when using group resources with symmetric=false" ],
[ "bug-5014-GROUP-A-stopped-B-started",
"Verify when A stops B does not stop if it has already started using group resources with symmetric=false" ],
[ "bug-5014-GROUP-A-stopped-B-stopped",
"Verify when A is stopped and B has not started, B does not start before A using group resources with symmetric=false" ],
[ "bug-5014-ordered-set-symmetrical-false",
"Verify ordered sets work with symmetrical=false" ],
[ "bug-5014-ordered-set-symmetrical-true",
"Verify ordered sets work with symmetrical=true" ],
[ "bug-5007-masterslave_colocation",
"Verify use of colocation scores other than INFINITY and -INFINITY work on multi-state resources" ],
[ "bug-5038", "Prevent restart of anonymous clones when clone-max decreases" ],
[ "bug-5025-1", "Automatically clean up failcount after resource config change with reload" ],
[ "bug-5025-2", "Make sure clear failcount action isn't set when config does not change" ],
[ "bug-5025-3", "Automatically clean up failcount after resource config change with restart" ],
[ "bug-5025-4", "Clear failcount when last failure is a start op and rsc attributes changed" ],
[ "failcount", "Ensure failcounts are correctly expired" ],
[ "failcount-block", "Ensure failcounts are not expired when on-fail=block is present" ],
[ "per-op-failcount", "Ensure per-operation failcount is handled and not passed to fence agent" ],
[ "on-fail-ignore", "Ensure on-fail=ignore works even beyond migration-threshold" ],
[ "monitor-onfail-restart", "bug-5058 - Monitor failure with on-fail set to restart" ],
[ "monitor-onfail-stop", "bug-5058 - Monitor failure wiht on-fail set to stop" ],
[ "bug-5059", "No need to restart p_stateful1:*" ],
[ "bug-5069-op-enabled", "Test on-fail=ignore with failure when monitor is enabled" ],
[ "bug-5069-op-disabled", "Test on-fail-ignore with failure when monitor is disabled" ],
[ "obsolete-lrm-resource", "cl#5115 - Do not use obsolete lrm_resource sections" ],
[ "expire-non-blocked-failure",
"Ignore failure-timeout only if the failed operation has on-fail=block" ],
[ "asymmetrical-order-move", "Respect asymmetrical ordering when trying to move resources" ],
[ "asymmetrical-order-restart", "Respect asymmetrical ordering when restarting dependent resource" ],
[ "start-then-stop-with-unfence", "Avoid graph loop with start-then-stop constraint plus unfencing" ],
[ "order-expired-failure", "Order failcount cleanup after remote fencing" ],
[ "ignore_stonith_rsc_order1",
"cl#5056- Ignore order constraint between stonith and non-stonith rsc" ],
[ "ignore_stonith_rsc_order2",
"cl#5056- Ignore order constraint with group rsc containing mixed stonith and non-stonith" ],
[ "ignore_stonith_rsc_order3", "cl#5056- Ignore order constraint, stonith clone and mixed group" ],
[ "ignore_stonith_rsc_order4",
"cl#5056- Ignore order constraint, stonith clone and clone with nested mixed group" ],
[ "honor_stonith_rsc_order1",
"cl#5056- Honor order constraint, stonith clone and pure stonith group(single rsc)" ],
[ "honor_stonith_rsc_order2",
"cl#5056- Honor order constraint, stonith clone and pure stonith group(multiple rsc)" ],
[ "honor_stonith_rsc_order3",
"cl#5056- Honor order constraint, stonith clones with nested pure stonith group" ],
[ "honor_stonith_rsc_order4",
"cl#5056- Honor order constraint, between two native stonith rscs" ],
[ "multiply-active-stonith", "Multiply active stonith" ],
[ "probe-timeout", "cl#5099 - Default probe timeout" ],
[ "order-first-probes",
"cl#5301 - respect order constraints when relevant resources are being probed" ],
[ "concurrent-fencing", "Allow performing fencing operations in parallel" ],
+ [ "priority-fencing-delay", "Delay fencing targeting the more significant node" ],
],
[
[ "systemhealth1", "System Health () #1" ],
[ "systemhealth2", "System Health () #2" ],
[ "systemhealth3", "System Health () #3" ],
[ "systemhealthn1", "System Health (None) #1" ],
[ "systemhealthn2", "System Health (None) #2" ],
[ "systemhealthn3", "System Health (None) #3" ],
[ "systemhealthm1", "System Health (Migrate On Red) #1" ],
[ "systemhealthm2", "System Health (Migrate On Red) #2" ],
[ "systemhealthm3", "System Health (Migrate On Red) #3" ],
[ "systemhealtho1", "System Health (Only Green) #1" ],
[ "systemhealtho2", "System Health (Only Green) #2" ],
[ "systemhealtho3", "System Health (Only Green) #3" ],
[ "systemhealthp1", "System Health (Progessive) #1" ],
[ "systemhealthp2", "System Health (Progessive) #2" ],
[ "systemhealthp3", "System Health (Progessive) #3" ],
],
[
[ "utilization", "Placement Strategy - utilization" ],
[ "minimal", "Placement Strategy - minimal" ],
[ "balanced", "Placement Strategy - balanced" ],
],
[
[ "placement-stickiness", "Optimized Placement Strategy - stickiness" ],
[ "placement-priority", "Optimized Placement Strategy - priority" ],
[ "placement-location", "Optimized Placement Strategy - location" ],
[ "placement-capacity", "Optimized Placement Strategy - capacity" ],
],
[
[ "utilization-order1", "Utilization Order - Simple" ],
[ "utilization-order2", "Utilization Order - Complex" ],
[ "utilization-order3", "Utilization Order - Migrate" ],
[ "utilization-order4", "Utilization Order - Live Migration (bnc#695440)" ],
[ "utilization-shuffle",
"Don't displace prmExPostgreSQLDB2 on act2, Start prmExPostgreSQLDB1 on act3" ],
[ "load-stopped-loop", "Avoid transition loop due to load_stopped (cl#5044)" ],
[ "load-stopped-loop-2",
"cl#5235 - Prevent graph loops that can be introduced by load_stopped -> migrate_to ordering" ],
],
[
[ "colocated-utilization-primitive-1", "Colocated Utilization - Primitive" ],
[ "colocated-utilization-primitive-2", "Colocated Utilization - Choose the most capable node" ],
[ "colocated-utilization-group", "Colocated Utilization - Group" ],
[ "colocated-utilization-clone", "Colocated Utilization - Clone" ],
[ "utilization-check-allowed-nodes",
"Only check the capacities of the nodes that can run the resource" ],
],
[
[ "reprobe-target_rc", "Ensure correct target_rc for reprobe of inactive resources" ],
[ "node-maintenance-1", "cl#5128 - Node maintenance" ],
[ "node-maintenance-2", "cl#5128 - Node maintenance (coming out of maintenance mode)" ],
[ "shutdown-maintenance-node", "Do not fence a maintenance node if it shuts down cleanly" ],
[ "rsc-maintenance", "Per-resource maintenance" ],
],
[
[ "not-installed-agent", "The resource agent is missing" ],
[ "not-installed-tools", "Something the resource agent needs is missing" ],
],
[
[ "stopped-monitor-00", "Stopped Monitor - initial start" ],
[ "stopped-monitor-01", "Stopped Monitor - failed started" ],
[ "stopped-monitor-02", "Stopped Monitor - started multi-up" ],
[ "stopped-monitor-03", "Stopped Monitor - stop started" ],
[ "stopped-monitor-04", "Stopped Monitor - failed stop" ],
[ "stopped-monitor-05", "Stopped Monitor - start unmanaged" ],
[ "stopped-monitor-06", "Stopped Monitor - unmanaged multi-up" ],
[ "stopped-monitor-07", "Stopped Monitor - start unmanaged multi-up" ],
[ "stopped-monitor-08", "Stopped Monitor - migrate" ],
[ "stopped-monitor-09", "Stopped Monitor - unmanage started" ],
[ "stopped-monitor-10", "Stopped Monitor - unmanaged started multi-up" ],
[ "stopped-monitor-11", "Stopped Monitor - stop unmanaged started" ],
[ "stopped-monitor-12", "Stopped Monitor - unmanaged started multi-up (target-role=Stopped)" ],
[ "stopped-monitor-20", "Stopped Monitor - initial stop" ],
[ "stopped-monitor-21", "Stopped Monitor - stopped single-up" ],
[ "stopped-monitor-22", "Stopped Monitor - stopped multi-up" ],
[ "stopped-monitor-23", "Stopped Monitor - start stopped" ],
[ "stopped-monitor-24", "Stopped Monitor - unmanage stopped" ],
[ "stopped-monitor-25", "Stopped Monitor - unmanaged stopped multi-up" ],
[ "stopped-monitor-26", "Stopped Monitor - start unmanaged stopped" ],
[ "stopped-monitor-27", "Stopped Monitor - unmanaged stopped multi-up (target-role=Started)" ],
[ "stopped-monitor-30", "Stopped Monitor - new node started" ],
[ "stopped-monitor-31", "Stopped Monitor - new node stopped" ],
],
[
# This is a combo test to check:
# - probe timeout defaults to the minimum-interval monitor's
# - duplicate recurring operations are ignored
# - if timeout spec is bad, the default timeout is used
# - failure is blocked with on-fail=block even if ISO8601 interval is specified
# - started/stopped role monitors are started/stopped on right nodes
[ "intervals", "Recurring monitor interval handling" ],
],
[
[ "ticket-primitive-1", "Ticket - Primitive (loss-policy=stop, initial)" ],
[ "ticket-primitive-2", "Ticket - Primitive (loss-policy=stop, granted)" ],
[ "ticket-primitive-3", "Ticket - Primitive (loss-policy-stop, revoked)" ],
[ "ticket-primitive-4", "Ticket - Primitive (loss-policy=demote, initial)" ],
[ "ticket-primitive-5", "Ticket - Primitive (loss-policy=demote, granted)" ],
[ "ticket-primitive-6", "Ticket - Primitive (loss-policy=demote, revoked)" ],
[ "ticket-primitive-7", "Ticket - Primitive (loss-policy=fence, initial)" ],
[ "ticket-primitive-8", "Ticket - Primitive (loss-policy=fence, granted)" ],
[ "ticket-primitive-9", "Ticket - Primitive (loss-policy=fence, revoked)" ],
[ "ticket-primitive-10", "Ticket - Primitive (loss-policy=freeze, initial)" ],
[ "ticket-primitive-11", "Ticket - Primitive (loss-policy=freeze, granted)" ],
[ "ticket-primitive-12", "Ticket - Primitive (loss-policy=freeze, revoked)" ],
[ "ticket-primitive-13", "Ticket - Primitive (loss-policy=stop, standby, granted)" ],
[ "ticket-primitive-14", "Ticket - Primitive (loss-policy=stop, granted, standby)" ],
[ "ticket-primitive-15", "Ticket - Primitive (loss-policy=stop, standby, revoked)" ],
[ "ticket-primitive-16", "Ticket - Primitive (loss-policy=demote, standby, granted)" ],
[ "ticket-primitive-17", "Ticket - Primitive (loss-policy=demote, granted, standby)" ],
[ "ticket-primitive-18", "Ticket - Primitive (loss-policy=demote, standby, revoked)" ],
[ "ticket-primitive-19", "Ticket - Primitive (loss-policy=fence, standby, granted)" ],
[ "ticket-primitive-20", "Ticket - Primitive (loss-policy=fence, granted, standby)" ],
[ "ticket-primitive-21", "Ticket - Primitive (loss-policy=fence, standby, revoked)" ],
[ "ticket-primitive-22", "Ticket - Primitive (loss-policy=freeze, standby, granted)" ],
[ "ticket-primitive-23", "Ticket - Primitive (loss-policy=freeze, granted, standby)" ],
[ "ticket-primitive-24", "Ticket - Primitive (loss-policy=freeze, standby, revoked)" ],
],
[
[ "ticket-group-1", "Ticket - Group (loss-policy=stop, initial)" ],
[ "ticket-group-2", "Ticket - Group (loss-policy=stop, granted)" ],
[ "ticket-group-3", "Ticket - Group (loss-policy-stop, revoked)" ],
[ "ticket-group-4", "Ticket - Group (loss-policy=demote, initial)" ],
[ "ticket-group-5", "Ticket - Group (loss-policy=demote, granted)" ],
[ "ticket-group-6", "Ticket - Group (loss-policy=demote, revoked)" ],
[ "ticket-group-7", "Ticket - Group (loss-policy=fence, initial)" ],
[ "ticket-group-8", "Ticket - Group (loss-policy=fence, granted)" ],
[ "ticket-group-9", "Ticket - Group (loss-policy=fence, revoked)" ],
[ "ticket-group-10", "Ticket - Group (loss-policy=freeze, initial)" ],
[ "ticket-group-11", "Ticket - Group (loss-policy=freeze, granted)" ],
[ "ticket-group-12", "Ticket - Group (loss-policy=freeze, revoked)" ],
[ "ticket-group-13", "Ticket - Group (loss-policy=stop, standby, granted)" ],
[ "ticket-group-14", "Ticket - Group (loss-policy=stop, granted, standby)" ],
[ "ticket-group-15", "Ticket - Group (loss-policy=stop, standby, revoked)" ],
[ "ticket-group-16", "Ticket - Group (loss-policy=demote, standby, granted)" ],
[ "ticket-group-17", "Ticket - Group (loss-policy=demote, granted, standby)" ],
[ "ticket-group-18", "Ticket - Group (loss-policy=demote, standby, revoked)" ],
[ "ticket-group-19", "Ticket - Group (loss-policy=fence, standby, granted)" ],
[ "ticket-group-20", "Ticket - Group (loss-policy=fence, granted, standby)" ],
[ "ticket-group-21", "Ticket - Group (loss-policy=fence, standby, revoked)" ],
[ "ticket-group-22", "Ticket - Group (loss-policy=freeze, standby, granted)" ],
[ "ticket-group-23", "Ticket - Group (loss-policy=freeze, granted, standby)" ],
[ "ticket-group-24", "Ticket - Group (loss-policy=freeze, standby, revoked)" ],
],
[
[ "ticket-clone-1", "Ticket - Clone (loss-policy=stop, initial)" ],
[ "ticket-clone-2", "Ticket - Clone (loss-policy=stop, granted)" ],
[ "ticket-clone-3", "Ticket - Clone (loss-policy-stop, revoked)" ],
[ "ticket-clone-4", "Ticket - Clone (loss-policy=demote, initial)" ],
[ "ticket-clone-5", "Ticket - Clone (loss-policy=demote, granted)" ],
[ "ticket-clone-6", "Ticket - Clone (loss-policy=demote, revoked)" ],
[ "ticket-clone-7", "Ticket - Clone (loss-policy=fence, initial)" ],
[ "ticket-clone-8", "Ticket - Clone (loss-policy=fence, granted)" ],
[ "ticket-clone-9", "Ticket - Clone (loss-policy=fence, revoked)" ],
[ "ticket-clone-10", "Ticket - Clone (loss-policy=freeze, initial)" ],
[ "ticket-clone-11", "Ticket - Clone (loss-policy=freeze, granted)" ],
[ "ticket-clone-12", "Ticket - Clone (loss-policy=freeze, revoked)" ],
[ "ticket-clone-13", "Ticket - Clone (loss-policy=stop, standby, granted)" ],
[ "ticket-clone-14", "Ticket - Clone (loss-policy=stop, granted, standby)" ],
[ "ticket-clone-15", "Ticket - Clone (loss-policy=stop, standby, revoked)" ],
[ "ticket-clone-16", "Ticket - Clone (loss-policy=demote, standby, granted)" ],
[ "ticket-clone-17", "Ticket - Clone (loss-policy=demote, granted, standby)" ],
[ "ticket-clone-18", "Ticket - Clone (loss-policy=demote, standby, revoked)" ],
[ "ticket-clone-19", "Ticket - Clone (loss-policy=fence, standby, granted)" ],
[ "ticket-clone-20", "Ticket - Clone (loss-policy=fence, granted, standby)" ],
[ "ticket-clone-21", "Ticket - Clone (loss-policy=fence, standby, revoked)" ],
[ "ticket-clone-22", "Ticket - Clone (loss-policy=freeze, standby, granted)" ],
[ "ticket-clone-23", "Ticket - Clone (loss-policy=freeze, granted, standby)" ],
[ "ticket-clone-24", "Ticket - Clone (loss-policy=freeze, standby, revoked)" ],
],
[
[ "ticket-master-1", "Ticket - Master (loss-policy=stop, initial)" ],
[ "ticket-master-2", "Ticket - Master (loss-policy=stop, granted)" ],
[ "ticket-master-3", "Ticket - Master (loss-policy-stop, revoked)" ],
[ "ticket-master-4", "Ticket - Master (loss-policy=demote, initial)" ],
[ "ticket-master-5", "Ticket - Master (loss-policy=demote, granted)" ],
[ "ticket-master-6", "Ticket - Master (loss-policy=demote, revoked)" ],
[ "ticket-master-7", "Ticket - Master (loss-policy=fence, initial)" ],
[ "ticket-master-8", "Ticket - Master (loss-policy=fence, granted)" ],
[ "ticket-master-9", "Ticket - Master (loss-policy=fence, revoked)" ],
[ "ticket-master-10", "Ticket - Master (loss-policy=freeze, initial)" ],
[ "ticket-master-11", "Ticket - Master (loss-policy=freeze, granted)" ],
[ "ticket-master-12", "Ticket - Master (loss-policy=freeze, revoked)" ],
[ "ticket-master-13", "Ticket - Master (loss-policy=stop, standby, granted)" ],
[ "ticket-master-14", "Ticket - Master (loss-policy=stop, granted, standby)" ],
[ "ticket-master-15", "Ticket - Master (loss-policy=stop, standby, revoked)" ],
[ "ticket-master-16", "Ticket - Master (loss-policy=demote, standby, granted)" ],
[ "ticket-master-17", "Ticket - Master (loss-policy=demote, granted, standby)" ],
[ "ticket-master-18", "Ticket - Master (loss-policy=demote, standby, revoked)" ],
[ "ticket-master-19", "Ticket - Master (loss-policy=fence, standby, granted)" ],
[ "ticket-master-20", "Ticket - Master (loss-policy=fence, granted, standby)" ],
[ "ticket-master-21", "Ticket - Master (loss-policy=fence, standby, revoked)" ],
[ "ticket-master-22", "Ticket - Master (loss-policy=freeze, standby, granted)" ],
[ "ticket-master-23", "Ticket - Master (loss-policy=freeze, granted, standby)" ],
[ "ticket-master-24", "Ticket - Master (loss-policy=freeze, standby, revoked)" ],
],
[
[ "ticket-rsc-sets-1", "Ticket - Resource sets (1 ticket, initial)" ],
[ "ticket-rsc-sets-2", "Ticket - Resource sets (1 ticket, granted)" ],
[ "ticket-rsc-sets-3", "Ticket - Resource sets (1 ticket, revoked)" ],
[ "ticket-rsc-sets-4", "Ticket - Resource sets (2 tickets, initial)" ],
[ "ticket-rsc-sets-5", "Ticket - Resource sets (2 tickets, granted)" ],
[ "ticket-rsc-sets-6", "Ticket - Resource sets (2 tickets, granted)" ],
[ "ticket-rsc-sets-7", "Ticket - Resource sets (2 tickets, revoked)" ],
[ "ticket-rsc-sets-8", "Ticket - Resource sets (1 ticket, standby, granted)" ],
[ "ticket-rsc-sets-9", "Ticket - Resource sets (1 ticket, granted, standby)" ],
[ "ticket-rsc-sets-10", "Ticket - Resource sets (1 ticket, standby, revoked)" ],
[ "ticket-rsc-sets-11", "Ticket - Resource sets (2 tickets, standby, granted)" ],
[ "ticket-rsc-sets-12", "Ticket - Resource sets (2 tickets, standby, granted)" ],
[ "ticket-rsc-sets-13", "Ticket - Resource sets (2 tickets, granted, standby)" ],
[ "ticket-rsc-sets-14", "Ticket - Resource sets (2 tickets, standby, revoked)" ],
[ "cluster-specific-params", "Cluster-specific instance attributes based on rules" ],
[ "site-specific-params", "Site-specific instance attributes based on rules" ],
],
[
[ "template-1", "Template - 1" ],
[ "template-2", "Template - 2" ],
[ "template-3", "Template - 3 (merge operations)" ],
[ "template-coloc-1", "Template - Colocation 1" ],
[ "template-coloc-2", "Template - Colocation 2" ],
[ "template-coloc-3", "Template - Colocation 3" ],
[ "template-order-1", "Template - Order 1" ],
[ "template-order-2", "Template - Order 2" ],
[ "template-order-3", "Template - Order 3" ],
[ "template-ticket", "Template - Ticket" ],
[ "template-rsc-sets-1", "Template - Resource Sets 1" ],
[ "template-rsc-sets-2", "Template - Resource Sets 2" ],
[ "template-rsc-sets-3", "Template - Resource Sets 3" ],
[ "template-rsc-sets-4", "Template - Resource Sets 4" ],
[ "template-clone-primitive", "Cloned primitive from template" ],
[ "template-clone-group", "Cloned group from template" ],
[ "location-sets-templates", "Resource sets and templates - Location" ],
[ "tags-coloc-order-1", "Tags - Colocation and Order (Simple)" ],
[ "tags-coloc-order-2", "Tags - Colocation and Order (Resource Sets with Templates)" ],
[ "tags-location", "Tags - Location" ],
[ "tags-ticket", "Tags - Ticket" ],
],
[
[ "container-1", "Container - initial" ],
[ "container-2", "Container - monitor failed" ],
[ "container-3", "Container - stop failed" ],
[ "container-4", "Container - reached migration-threshold" ],
[ "container-group-1", "Container in group - initial" ],
[ "container-group-2", "Container in group - monitor failed" ],
[ "container-group-3", "Container in group - stop failed" ],
[ "container-group-4", "Container in group - reached migration-threshold" ],
[ "container-is-remote-node", "Place resource within container when container is remote-node" ],
[ "bug-rh-1097457", "Kill user defined container/contents ordering" ],
[ "bug-cl-5247", "Graph loop when recovering m/s resource in a container" ],
[ "bundle-order-startup", "Bundle startup ordering" ],
[ "bundle-order-partial-start",
"Bundle startup ordering when some dependencies are already running" ],
[ "bundle-order-partial-start-2",
"Bundle startup ordering when some dependencies and the container are already running" ],
[ "bundle-order-stop", "Bundle stop ordering" ],
[ "bundle-order-partial-stop", "Bundle startup ordering when some dependencies are already stopped" ],
[ "bundle-order-stop-on-remote", "Stop nested resource after bringing up the connection" ],
[ "bundle-order-startup-clone", "Prevent startup because bundle isn't promoted" ],
[ "bundle-order-startup-clone-2", "Bundle startup with clones" ],
[ "bundle-order-stop-clone", "Stop bundle because clone is stopping" ],
[ "bundle-nested-colocation", "Colocation of nested connection resources" ],
[ "bundle-order-fencing",
"Order pseudo bundle fencing after parent node fencing if both are happening" ],
[ "bundle-probe-order-1", "order 1" ],
[ "bundle-probe-order-2", "order 2" ],
[ "bundle-probe-order-3", "order 3" ],
[ "bundle-probe-remotes", "Ensure remotes get probed too" ],
[ "bundle-replicas-change", "Change bundle from 1 replica to multiple" ],
[ "nested-remote-recovery", "Recover bundle's container hosted on remote node" ],
],
[
[ "whitebox-fail1", "Fail whitebox container rsc" ],
[ "whitebox-fail2", "Fail cluster connection to guest node" ],
[ "whitebox-fail3", "Failed containers should not run nested on remote nodes" ],
[ "whitebox-start", "Start whitebox container with resources assigned to it" ],
[ "whitebox-stop", "Stop whitebox container with resources assigned to it" ],
[ "whitebox-move", "Move whitebox container with resources assigned to it" ],
[ "whitebox-asymmetric", "Verify connection rsc opts-in based on container resource" ],
[ "whitebox-ms-ordering", "Verify promote/demote can not occur before connection is established" ],
[ "whitebox-ms-ordering-move", "Stop/Start cycle within a moving container" ],
[ "whitebox-orphaned", "Properly shutdown orphaned whitebox container" ],
[ "whitebox-orphan-ms", "Properly tear down orphan ms resources on remote-nodes" ],
[ "whitebox-unexpectedly-running", "Recover container nodes the cluster did not start" ],
[ "whitebox-migrate1", "Migrate both container and connection resource" ],
[ "whitebox-imply-stop-on-fence",
"imply stop action on container node rsc when host node is fenced" ],
[ "whitebox-nested-group", "Verify guest remote-node works nested in a group" ],
[ "guest-node-host-dies", "Verify guest node is recovered if host goes away" ],
[ "guest-node-cleanup", "Order guest node connection recovery after container probe" ],
[ "guest-host-not-fenceable", "Actions on guest node are unrunnable if host is unclean and cannot be fenced" ],
],
[
[ "remote-startup-probes", "Baremetal remote-node startup probes" ],
[ "remote-startup", "Startup a newly discovered remote-nodes with no status" ],
[ "remote-fence-unclean", "Fence unclean baremetal remote-node" ],
[ "remote-fence-unclean2",
"Fence baremetal remote-node after cluster node fails and connection can not be recovered" ],
[ "remote-fence-unclean-3", "Probe failed remote nodes (triggers fencing)" ],
[ "remote-move", "Move remote-node connection resource" ],
[ "remote-disable", "Disable a baremetal remote-node" ],
[ "remote-probe-disable", "Probe then stop a baremetal remote-node" ],
[ "remote-orphaned", "Properly shutdown orphaned connection resource" ],
[ "remote-orphaned2",
"verify we can handle orphaned remote connections with active resources on the remote" ],
[ "remote-recover", "Recover connection resource after cluster-node fails" ],
[ "remote-stale-node-entry",
"Make sure we properly handle leftover remote-node entries in the node section" ],
[ "remote-partial-migrate",
"Make sure partial migrations are handled before ops on the remote node" ],
[ "remote-partial-migrate2",
"Make sure partial migration target is prefered for remote connection" ],
[ "remote-recover-fail", "Make sure start failure causes fencing if rsc are active on remote" ],
[ "remote-start-fail",
"Make sure a start failure does not result in fencing if no active resources are on remote" ],
[ "remote-unclean2",
"Make monitor failure always results in fencing, even if no rsc are active on remote" ],
[ "remote-fence-before-reconnect", "Fence before clearing recurring monitor failure" ],
[ "remote-recovery", "Recover remote connections before attempting demotion" ],
[ "remote-recover-connection", "Optimistically recovery of only the connection" ],
[ "remote-recover-all", "Fencing when the connection has no home" ],
[ "remote-recover-no-resources", "Fencing when the connection has no home and no active resources" ],
[ "remote-recover-unknown",
"Fencing when the connection has no home and the remote has no operation history" ],
[ "remote-reconnect-delay", "Waiting for remote reconnect interval to expire" ],
[ "remote-connection-unrecoverable",
"Remote connection host must be fenced, with connection unrecoverable" ],
],
[
[ "resource-discovery", "Exercises resource-discovery location constraint option" ],
[ "rsc-discovery-per-node", "Disable resource discovery per node" ],
[ "shutdown-lock", "Ensure shutdown lock works properly" ],
[ "shutdown-lock-expiration", "Ensure shutdown lock expiration works properly" ],
],
# @TODO: If pacemaker implements versioned attributes, uncomment these tests
#[
# [ "versioned-resources", "Start resources with #ra-version rules" ],
# [ "restart-versioned", "Restart resources on #ra-version change" ],
# [ "reload-versioned", "Reload resources on #ra-version change" ],
#],
#[
# [ "versioned-operations-1", "Use #ra-version to configure operations of native resources" ],
# [ "versioned-operations-2", "Use #ra-version to configure operations of stonith resources" ],
# [ "versioned-operations-3", "Use #ra-version to configure operations of master/slave resources" ],
# [ "versioned-operations-4", "Use #ra-version to configure operations of groups of the resources" ],
#],
]
TESTS_64BIT = [
[
[ "year-2038", "Check handling of timestamps beyond 2038-01-19 03:14:08 UTC" ],
],
]
# Constants substituted in the build process
class BuildVars(object):
SBINDIR = "@sbindir@"
BUILDDIR = "@abs_top_builddir@"
CRM_SCHEMA_DIRECTORY = "@CRM_SCHEMA_DIRECTORY@"
# These values must be kept in sync with include/crm/crm.h
class CrmExit(object):
OK = 0
ERROR = 1
NOT_INSTALLED = 5
NOINPUT = 66
def is_executable(path):
""" Check whether a file at a given path is executable. """
try:
return os.stat(path)[stat.ST_MODE] & stat.S_IXUSR
except OSError:
return False
def diff(file1, file2, **kwargs):
""" Call diff on two files """
return subprocess.call([ "diff", "-u", "-N", "--ignore-all-space",
"--ignore-blank-lines", file1, file2 ], **kwargs)
def sort_file(filename):
""" Sort a file alphabetically """
with io.open(filename, "rt") as f:
lines = sorted(f)
with io.open(filename, "wt") as f:
f.writelines(lines)
def remove_files(filenames):
""" Remove a list of files """
for filename in filenames:
try:
os.remove(filename)
except OSError:
pass
def normalize(filename):
""" Remove text from a file that isn't important for comparison """
if not hasattr(normalize, "patterns"):
normalize.patterns = [
re.compile(r'crm_feature_set="[^"]*"'),
re.compile(r'batch-limit="[0-9]*"')
]
if os.path.isfile(filename):
with io.open(filename, "rt") as f:
lines = f.readlines()
with io.open(filename, "wt") as f:
for line in lines:
for pattern in normalize.patterns:
line = pattern.sub("", line)
f.write(line)
def cat(filename, dest=sys.stdout):
""" Copy a file to a destination file descriptor """
with io.open(filename, "rt") as f:
shutil.copyfileobj(f, dest)
class CtsScheduler(object):
""" Regression tests for Pacemaker's scheduler """
def _parse_args(self, argv):
""" Parse command-line arguments """
parser = argparse.ArgumentParser(description=DESC)
parser.add_argument('-V', '--verbose', action='count',
help='Display any differences from expected output')
parser.add_argument('--run', metavar='TEST',
help=('Run only single specified test (any further '
'arguments will be passed to crm_simulate)'))
parser.add_argument('--update', action='store_true',
help='Update expected results with actual results')
parser.add_argument('-b', '--binary', metavar='PATH',
help='Specify path to crm_simulate')
parser.add_argument('-i', '--io-dir', metavar='PATH',
help='Specify path to regression test data directory')
parser.add_argument('-o', '--out-dir', metavar='PATH',
help='Specify where intermediate and output files should go')
parser.add_argument('-v', '--valgrind', action='store_true',
help='Run all commands under valgrind')
parser.add_argument('--valgrind-dhat', action='store_true',
help='Run all commands under valgrind with heap analyzer')
parser.add_argument('--valgrind-skip-output', action='store_true',
help='If running under valgrind, do not display output')
parser.add_argument('--testcmd-options', metavar='OPTIONS', default='',
help='Additional options for command under test')
# argparse can't handle "everything after --run TEST", so grab that
self.single_test_args = []
narg = 0
for arg in argv:
narg = narg + 1
if arg == '--run':
(argv, self.single_test_args) = (argv[:narg+1], argv[narg+1:])
break
self.args = parser.parse_args(argv[1:])
def _error(self, s):
print(" * ERROR: %s" % s)
def _failed(self, s):
print(" * FAILED: %s" % s)
def _get_valgrind_cmd(self):
""" Return command arguments needed (or not) to run valgrind """
if self.args.valgrind:
os.environ['G_SLICE'] = "always-malloc"
return [
"valgrind",
"-q",
"--gen-suppressions=all",
"--time-stamp=yes",
"--trace-children=no",
"--show-reachable=no",
"--leak-check=full",
"--num-callers=20",
"--suppressions=%s/valgrind-pcmk.suppressions" % (self.test_home)
]
if self.args.valgrind_dhat:
os.environ['G_SLICE'] = "always-malloc"
return [
"valgrind",
"--tool=exp-dhat",
"--time-stamp=yes",
"--trace-children=no",
"--show-top-n=100",
"--num-callers=4"
]
return []
def _get_simulator_cmd(self):
""" Locate the simulation binary """
if self.args.binary is None:
self.args.binary = BuildVars.BUILDDIR + "/tools/crm_simulate"
if not is_executable(self.args.binary):
self.args.binary = BuildVars.SBINDIR + "/crm_simulate"
if not is_executable(self.args.binary):
# @TODO it would be more pythonic to raise an exception
self._error("Test binary " + self.args.binary + " not found")
sys.exit(CrmExit.NOT_INSTALLED)
return [ self.args.binary ] + shlex.split(self.args.testcmd_options)
def set_schema_env(self):
""" Ensure schema directory environment variable is set, if possible """
try:
return os.environ['PCMK_schema_directory']
except KeyError:
for d in [ os.path.join(BuildVars.BUILDDIR, "xml"),
BuildVars.CRM_SCHEMA_DIRECTORY ]:
if os.path.isdir(d):
os.environ['PCMK_schema_directory'] = d
return d
return None
def __init__(self, argv=sys.argv):
self._parse_args(argv)
# Where this executable lives
self.test_home = os.path.dirname(os.path.realpath(argv[0]))
# Where test data resides
if self.args.io_dir is None:
self.args.io_dir = os.path.join(self.test_home, "scheduler")
# Where to store generated files
if self.args.out_dir is None:
self.args.out_dir = self.args.io_dir
self.failed_filename = os.path.join(self.test_home, ".regression.failed.diff")
else:
self.failed_filename = os.path.join(self.args.out_dir, ".regression.failed.diff")
os.environ['CIB_shadow_dir'] = self.args.out_dir
self.failed_file = None
# Single test mode (if requested)
try:
# User can give test base name or file name of a test input
self.args.run = os.path.splitext(os.path.basename(self.args.run))[0]
except (AttributeError, TypeError):
pass # --run was not specified
self.set_schema_env()
# Arguments needed (or not) to run commands
self.valgrind_args = self._get_valgrind_cmd()
self.simulate_args = self._get_simulator_cmd()
# Test counters
self.num_failed = 0
self.num_tests = 0
def _compare_files(self, filename1, filename2):
""" Add any file differences to failed results """
with io.open("/dev/null", "wt") as dev_null:
if diff(filename1, filename2, stdout=dev_null) != 0:
diff(filename1, filename2, stdout=self.failed_file, stderr=dev_null)
self.failed_file.write("\n");
return True
return False
def run_one(self, test_name, test_desc, test_args=[]):
""" Run one scheduler test """
print(" Test %-25s %s" % ((test_name + ":"), test_desc))
did_fail = False
self.num_tests = self.num_tests + 1
# Test inputs
input_filename = "%s/%s.xml" % (self.args.io_dir, test_name)
expected_filename = "%s/%s.exp" % (self.args.io_dir, test_name)
dot_expected_filename = "%s/%s.dot" % (self.args.io_dir, test_name)
scores_filename = "%s/%s.scores" % (self.args.io_dir, test_name)
summary_filename = "%s/%s.summary" % (self.args.io_dir, test_name)
stderr_expected_filename = "%s/%s.stderr" % (self.args.io_dir, test_name)
# (Intermediate) test outputs
output_filename = "%s/%s.out" % (self.args.out_dir, test_name)
dot_output_filename = "%s/%s.pe.dot" % (self.args.out_dir, test_name)
score_output_filename = "%s/%s.scores.pe" % (self.args.out_dir, test_name)
summary_output_filename = "%s/%s.summary.pe" % (self.args.out_dir, test_name)
stderr_output_filename = "%s/%s.stderr.pe" % (self.args.out_dir, test_name)
valgrind_output_filename = "%s/%s.valgrind" % (self.args.out_dir, test_name)
# Common arguments for running test
test_cmd = []
if self.valgrind_args:
test_cmd = self.valgrind_args + [ "--log-file=%s" % valgrind_output_filename ]
test_cmd = test_cmd + self.simulate_args
# @TODO It would be more pythonic to raise exceptions for errors,
# then perhaps it would be nice to make a single-test class
# Ensure necessary test inputs exist
if not os.path.isfile(input_filename):
self._error("No input")
self.num_failed = self.num_failed + 1
return CrmExit.NOINPUT
if not self.args.update and not os.path.isfile(expected_filename):
self._error("no stored output")
return CrmExit.NOINPUT
# Run simulation to generate summary output
if self.args.run: # Single test mode
test_cmd_full = test_cmd + [ '-x', input_filename, '-S' ] + test_args
print(" ".join(test_cmd_full))
else:
# @TODO Why isn't test_args added here?
test_cmd_full = test_cmd + [ '-x', input_filename, '-S' ]
with io.open(summary_output_filename, "wt") as f:
subprocess.call(test_cmd_full, stdout=f, stderr=subprocess.STDOUT, env=os.environ)
if self.args.run:
cat(summary_output_filename)
# Re-run simulation to generate dot, graph, and scores
test_cmd_full = test_cmd + [
'-x', input_filename,
'-D', dot_output_filename,
'-G', output_filename,
'-sSQ' ] + test_args
with io.open(stderr_output_filename, "wt") as f_stderr, \
io.open(score_output_filename, "wt") as f_score:
rc = subprocess.call(test_cmd_full, stdout=f_score, stderr=f_stderr, env=os.environ)
# Check for test command failure
if rc != CrmExit.OK:
self._failed("Test returned: %d" % rc)
did_fail = True
print(" ".join(test_cmd_full))
# Check for valgrind errors
if self.valgrind_args and not self.args.valgrind_skip_output:
if os.stat(valgrind_output_filename).st_size > 0:
self._failed("Valgrind reported errors")
did_fail = True
cat(valgrind_output_filename)
remove_files([ valgrind_output_filename ])
# Check for core dump
if os.path.isfile("core"):
self._failed("Core-file detected: core." + test_name)
did_fail = True
os.rename("core", "%s/core.%s" % (self.test_home, test_name))
# Check any stderr output
if os.path.isfile(stderr_expected_filename):
if self._compare_files(stderr_expected_filename, stderr_output_filename):
self._failed("stderr changed")
did_fail = True
elif os.stat(stderr_output_filename).st_size > 0:
self._failed("Output was written to stderr")
did_fail = True
cat(stderr_output_filename)
remove_files([ stderr_output_filename ])
# Check whether output graph exists, and normalize it
if (not os.path.isfile(output_filename)
or os.stat(output_filename).st_size == 0):
self._error("No graph produced")
did_fail = True
self.num_failed = self.num_failed + 1
remove_files([ output_filename ])
return CrmExit.ERROR
normalize(output_filename)
# Check whether dot output exists, and sort it
if (not os.path.isfile(dot_output_filename) or
os.stat(dot_output_filename).st_size == 0):
self._error("No dot-file summary produced")
did_fail = True
self.num_failed = self.num_failed + 1
remove_files([ dot_output_filename, output_filename ])
return CrmExit.ERROR
with io.open(dot_output_filename, "rt") as f:
first_line = f.readline() # "digraph" line with opening brace
lines = f.readlines()
last_line = lines[-1] # closing brace
del lines[-1]
lines = sorted(set(lines)) # unique sort
with io.open(dot_output_filename, "wt") as f:
f.write(first_line)
f.writelines(lines)
f.write(last_line)
# Check whether score output exists, and sort it
if (not os.path.isfile(score_output_filename)
or os.stat(score_output_filename).st_size == 0):
self._error("No allocation scores produced")
did_fail = True
self.num_failed = self.num_failed + 1
remove_files([ score_output_filename, output_filename ])
return CrmExit.ERROR
else:
sort_file(score_output_filename)
if self.args.update:
shutil.copyfile(output_filename, expected_filename)
shutil.copyfile(dot_output_filename, dot_expected_filename)
shutil.copyfile(score_output_filename, scores_filename)
shutil.copyfile(summary_output_filename, summary_filename)
print(" Updated expected outputs")
if self._compare_files(summary_filename, summary_output_filename):
self._failed("summary changed")
did_fail = True
if self._compare_files(dot_expected_filename, dot_output_filename):
self._failed("dot-file summary changed")
did_fail = True
else:
remove_files([ dot_output_filename ])
if self._compare_files(expected_filename, output_filename):
self._failed("xml-file changed")
did_fail = True
if self._compare_files(scores_filename, score_output_filename):
self._failed("scores-file changed")
did_fail = True
remove_files([ output_filename,
score_output_filename,
summary_output_filename])
if did_fail:
self.num_failed = self.num_failed + 1
return CrmExit.ERROR
return CrmExit.OK
def run_all(self):
""" Run all defined tests """
if platform.architecture()[0] == "64bit":
TESTS.extend(TESTS_64BIT)
for group in TESTS:
for test in group:
try:
args = test[2]
except IndexError:
args = []
self.run_one(test[0], test[1], args)
print()
def _print_summary(self):
""" Print a summary of parameters for this test run """
print("Test home is:\t" + self.test_home)
print("Test binary is:\t" + self.args.binary)
if 'PCMK_schema_directory' in os.environ:
print("Schema home is:\t" + os.environ['PCMK_schema_directory'])
if self.valgrind_args != []:
print("Activating memory testing with valgrind")
print()
def _test_results(self):
if self.num_failed == 0:
return CrmExit.OK
if os.path.isfile(self.failed_filename) and os.stat(self.failed_filename).st_size != 0:
if self.args.verbose:
self._error("Results of %d failed tests (out of %d):" %
(self.num_failed, self.num_tests))
cat(self.failed_filename)
else:
self._error("Results of %d failed tests (out of %d) are in %s" %
(self.num_failed, self.num_tests, self.failed_filename))
self._error("Use -V to display them after running the tests")
else:
self._error("%d (of %d) tests failed (no diff results)" %
(self.num_failed, self.num_tests))
if os.path.isfile(self.failed_filename):
os.remove(self.failed_filename)
return CrmExit.ERROR
def run(self):
""" Run test(s) as specified """
self._print_summary()
# Zero out the error log
self.failed_file = io.open(self.failed_filename, "wt")
if self.args.run is None:
print("Performing the following tests from " + self.args.io_dir)
print()
self.run_all()
print()
self.failed_file.close()
rc = self._test_results()
else:
rc = self.run_one(self.args.run, "Single shot", self.single_test_args)
self.failed_file.close()
cat(self.failed_filename)
return rc
if __name__ == "__main__":
sys.exit(CtsScheduler().run())
# vim: set filetype=python expandtab tabstop=4 softtabstop=4 shiftwidth=4 textwidth=120:
diff --git a/cts/scheduler/priority-fencing-delay.dot b/cts/scheduler/priority-fencing-delay.dot
new file mode 100644
index 0000000000..62ba699eb6
--- /dev/null
+++ b/cts/scheduler/priority-fencing-delay.dot
@@ -0,0 +1,109 @@
+ digraph "g" {
+"R-lxc-01_kiff-01_monitor_10000 kiff-02" [ style=bold color="green" fontcolor="black"]
+"R-lxc-01_kiff-01_start_0 kiff-02" -> "R-lxc-01_kiff-01_monitor_10000 kiff-02" [ style = bold]
+"R-lxc-01_kiff-01_start_0 kiff-02" -> "lxc-01_kiff-01_start_0 kiff-02" [ style = bold]
+"R-lxc-01_kiff-01_start_0 kiff-02" -> "vm-fs_start_0 lxc-01_kiff-01" [ style = bold]
+"R-lxc-01_kiff-01_start_0 kiff-02" [ style=bold color="green" fontcolor="black"]
+"R-lxc-01_kiff-01_stop_0 kiff-01" -> "R-lxc-01_kiff-01_start_0 kiff-02" [ style = bold]
+"R-lxc-01_kiff-01_stop_0 kiff-01" -> "shared0-clone_stop_0" [ style = bold]
+"R-lxc-01_kiff-01_stop_0 kiff-01" [ style=bold color="green" fontcolor="orange"]
+"R-lxc-02_kiff-01_monitor_10000 kiff-02" [ style=bold color="green" fontcolor="black"]
+"R-lxc-02_kiff-01_start_0 kiff-02" -> "R-lxc-02_kiff-01_monitor_10000 kiff-02" [ style = bold]
+"R-lxc-02_kiff-01_start_0 kiff-02" -> "lxc-02_kiff-01_start_0 kiff-02" [ style = bold]
+"R-lxc-02_kiff-01_start_0 kiff-02" [ style=bold color="green" fontcolor="black"]
+"R-lxc-02_kiff-01_stop_0 kiff-01" -> "R-lxc-02_kiff-01_start_0 kiff-02" [ style = bold]
+"R-lxc-02_kiff-01_stop_0 kiff-01" -> "shared0-clone_stop_0" [ style = bold]
+"R-lxc-02_kiff-01_stop_0 kiff-01" -> "vm-fs_start_0 lxc-01_kiff-01" [ style = bold]
+"R-lxc-02_kiff-01_stop_0 kiff-01" [ style=bold color="green" fontcolor="orange"]
+"clvmd-clone_stop_0" -> "clvmd-clone_stopped_0" [ style = bold]
+"clvmd-clone_stop_0" -> "clvmd_stop_0 kiff-01" [ style = bold]
+"clvmd-clone_stop_0" [ style=bold color="green" fontcolor="orange"]
+"clvmd-clone_stopped_0" -> "dlm-clone_stop_0" [ style = bold]
+"clvmd-clone_stopped_0" [ style=bold color="green" fontcolor="orange"]
+"clvmd_monitor_0 lxc-01_kiff-02" -> "clvmd-clone_stopped_0" [ style = bold]
+"clvmd_monitor_0 lxc-01_kiff-02" [ style=bold color="green" fontcolor="black"]
+"clvmd_monitor_0 lxc-02_kiff-02" -> "clvmd-clone_stopped_0" [ style = bold]
+"clvmd_monitor_0 lxc-02_kiff-02" [ style=bold color="green" fontcolor="black"]
+"clvmd_stop_0 kiff-01" -> "clvmd-clone_stopped_0" [ style = bold]
+"clvmd_stop_0 kiff-01" -> "dlm_stop_0 kiff-01" [ style = bold]
+"clvmd_stop_0 kiff-01" [ style=bold color="green" fontcolor="orange"]
+"dlm-clone_stop_0" -> "dlm-clone_stopped_0" [ style = bold]
+"dlm-clone_stop_0" -> "dlm_stop_0 kiff-01" [ style = bold]
+"dlm-clone_stop_0" [ style=bold color="green" fontcolor="orange"]
+"dlm-clone_stopped_0" [ style=bold color="green" fontcolor="orange"]
+"dlm_monitor_0 lxc-01_kiff-02" -> "dlm-clone_stopped_0" [ style = bold]
+"dlm_monitor_0 lxc-01_kiff-02" [ style=bold color="green" fontcolor="black"]
+"dlm_monitor_0 lxc-02_kiff-02" -> "dlm-clone_stopped_0" [ style = bold]
+"dlm_monitor_0 lxc-02_kiff-02" [ style=bold color="green" fontcolor="black"]
+"dlm_stop_0 kiff-01" -> "dlm-clone_stopped_0" [ style = bold]
+"dlm_stop_0 kiff-01" [ style=bold color="green" fontcolor="orange"]
+"fence-kiff-02_monitor_60000 kiff-02" [ style=bold color="green" fontcolor="black"]
+"fence-kiff-02_start_0 kiff-02" -> "fence-kiff-02_monitor_60000 kiff-02" [ style = bold]
+"fence-kiff-02_start_0 kiff-02" [ style=bold color="green" fontcolor="black"]
+"fence-kiff-02_stop_0 kiff-01" -> "fence-kiff-02_start_0 kiff-02" [ style = bold]
+"fence-kiff-02_stop_0 kiff-01" [ style=bold color="green" fontcolor="orange"]
+"lxc-01_kiff-01_monitor_30000 kiff-02" [ style=bold color="green" fontcolor="black"]
+"lxc-01_kiff-01_start_0 kiff-02" -> "lxc-01_kiff-01_monitor_30000 kiff-02" [ style = bold]
+"lxc-01_kiff-01_start_0 kiff-02" -> "vm-fs_monitor_20000 lxc-01_kiff-01" [ style = bold]
+"lxc-01_kiff-01_start_0 kiff-02" -> "vm-fs_start_0 lxc-01_kiff-01" [ style = bold]
+"lxc-01_kiff-01_start_0 kiff-02" [ style=bold color="green" fontcolor="black"]
+"lxc-01_kiff-01_stop_0 kiff-01" -> "R-lxc-01_kiff-01_stop_0 kiff-01" [ style = bold]
+"lxc-01_kiff-01_stop_0 kiff-01" -> "lxc-01_kiff-01_start_0 kiff-02" [ style = bold]
+"lxc-01_kiff-01_stop_0 kiff-01" [ style=bold color="green" fontcolor="orange"]
+"lxc-02_kiff-01_monitor_30000 kiff-02" [ style=bold color="green" fontcolor="black"]
+"lxc-02_kiff-01_start_0 kiff-02" -> "lxc-02_kiff-01_monitor_30000 kiff-02" [ style = bold]
+"lxc-02_kiff-01_start_0 kiff-02" [ style=bold color="green" fontcolor="black"]
+"lxc-02_kiff-01_stop_0 kiff-01" -> "R-lxc-02_kiff-01_stop_0 kiff-01" [ style = bold]
+"lxc-02_kiff-01_stop_0 kiff-01" -> "lxc-02_kiff-01_start_0 kiff-02" [ style = bold]
+"lxc-02_kiff-01_stop_0 kiff-01" [ style=bold color="green" fontcolor="orange"]
+"shared0-clone_stop_0" -> "shared0-clone_stopped_0" [ style = bold]
+"shared0-clone_stop_0" -> "shared0_stop_0 kiff-01" [ style = bold]
+"shared0-clone_stop_0" [ style=bold color="green" fontcolor="orange"]
+"shared0-clone_stopped_0" -> "clvmd-clone_stop_0" [ style = bold]
+"shared0-clone_stopped_0" [ style=bold color="green" fontcolor="orange"]
+"shared0_monitor_0 lxc-01_kiff-02" -> "shared0-clone_stopped_0" [ style = bold]
+"shared0_monitor_0 lxc-01_kiff-02" [ style=bold color="green" fontcolor="black"]
+"shared0_monitor_0 lxc-02_kiff-02" -> "shared0-clone_stopped_0" [ style = bold]
+"shared0_monitor_0 lxc-02_kiff-02" [ style=bold color="green" fontcolor="black"]
+"shared0_stop_0 kiff-01" -> "clvmd_stop_0 kiff-01" [ style = bold]
+"shared0_stop_0 kiff-01" -> "shared0-clone_stopped_0" [ style = bold]
+"shared0_stop_0 kiff-01" [ style=bold color="green" fontcolor="orange"]
+"stonith 'reboot' kiff-01" -> "R-lxc-01_kiff-01_start_0 kiff-02" [ style = bold]
+"stonith 'reboot' kiff-01" -> "R-lxc-01_kiff-01_stop_0 kiff-01" [ style = bold]
+"stonith 'reboot' kiff-01" -> "R-lxc-02_kiff-01_start_0 kiff-02" [ style = bold]
+"stonith 'reboot' kiff-01" -> "R-lxc-02_kiff-01_stop_0 kiff-01" [ style = bold]
+"stonith 'reboot' kiff-01" -> "clvmd-clone_stop_0" [ style = bold]
+"stonith 'reboot' kiff-01" -> "clvmd_stop_0 kiff-01" [ style = bold]
+"stonith 'reboot' kiff-01" -> "dlm-clone_stop_0" [ style = bold]
+"stonith 'reboot' kiff-01" -> "dlm_stop_0 kiff-01" [ style = bold]
+"stonith 'reboot' kiff-01" -> "shared0-clone_stop_0" [ style = bold]
+"stonith 'reboot' kiff-01" -> "shared0_stop_0 kiff-01" [ style = bold]
+"stonith 'reboot' kiff-01" -> "stonith 'reboot' lxc-01_kiff-01" [ style = bold]
+"stonith 'reboot' kiff-01" -> "stonith 'reboot' lxc-02_kiff-01" [ style = bold]
+"stonith 'reboot' kiff-01" -> "vm-fs_start_0 lxc-01_kiff-01" [ style = bold]
+"stonith 'reboot' kiff-01" [ style=bold color="green" fontcolor="black"]
+"stonith 'reboot' lxc-01_kiff-01" -> "R-lxc-01_kiff-01_start_0 kiff-02" [ style = bold]
+"stonith 'reboot' lxc-01_kiff-01" -> "R-lxc-02_kiff-01_start_0 kiff-02" [ style = bold]
+"stonith 'reboot' lxc-01_kiff-01" -> "fence-kiff-02_start_0 kiff-02" [ style = bold]
+"stonith 'reboot' lxc-01_kiff-01" -> "lxc-01_kiff-01_start_0 kiff-02" [ style = bold]
+"stonith 'reboot' lxc-01_kiff-01" -> "lxc-02_kiff-01_start_0 kiff-02" [ style = bold]
+"stonith 'reboot' lxc-01_kiff-01" -> "vm-fs_start_0 lxc-01_kiff-01" [ style = bold]
+"stonith 'reboot' lxc-01_kiff-01" -> "vm-fs_stop_0 lxc-01_kiff-01" [ style = bold]
+"stonith 'reboot' lxc-01_kiff-01" [ style=bold color="green" fontcolor="orange"]
+"stonith 'reboot' lxc-02_kiff-01" -> "R-lxc-01_kiff-01_start_0 kiff-02" [ style = bold]
+"stonith 'reboot' lxc-02_kiff-01" -> "R-lxc-02_kiff-01_start_0 kiff-02" [ style = bold]
+"stonith 'reboot' lxc-02_kiff-01" -> "fence-kiff-02_start_0 kiff-02" [ style = bold]
+"stonith 'reboot' lxc-02_kiff-01" -> "lxc-01_kiff-01_start_0 kiff-02" [ style = bold]
+"stonith 'reboot' lxc-02_kiff-01" -> "lxc-02_kiff-01_start_0 kiff-02" [ style = bold]
+"stonith 'reboot' lxc-02_kiff-01" -> "vm-fs_start_0 lxc-01_kiff-01" [ style = bold]
+"stonith 'reboot' lxc-02_kiff-01" [ style=bold color="green" fontcolor="orange"]
+"vm-fs_monitor_0 lxc-01_kiff-02" -> "vm-fs_start_0 lxc-01_kiff-01" [ style = bold]
+"vm-fs_monitor_0 lxc-01_kiff-02" [ style=bold color="green" fontcolor="black"]
+"vm-fs_monitor_0 lxc-02_kiff-02" -> "vm-fs_start_0 lxc-01_kiff-01" [ style = bold]
+"vm-fs_monitor_0 lxc-02_kiff-02" [ style=bold color="green" fontcolor="black"]
+"vm-fs_monitor_20000 lxc-01_kiff-01" [ style=bold color="green" fontcolor="black"]
+"vm-fs_start_0 lxc-01_kiff-01" -> "vm-fs_monitor_20000 lxc-01_kiff-01" [ style = bold]
+"vm-fs_start_0 lxc-01_kiff-01" [ style=bold color="green" fontcolor="black"]
+"vm-fs_stop_0 lxc-01_kiff-01" -> "vm-fs_start_0 lxc-01_kiff-01" [ style = bold]
+"vm-fs_stop_0 lxc-01_kiff-01" [ style=bold color="green" fontcolor="orange"]
+}
diff --git a/cts/scheduler/priority-fencing-delay.exp b/cts/scheduler/priority-fencing-delay.exp
new file mode 100644
index 0000000000..c6315a1280
--- /dev/null
+++ b/cts/scheduler/priority-fencing-delay.exp
@@ -0,0 +1,570 @@
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
diff --git a/cts/scheduler/priority-fencing-delay.scores b/cts/scheduler/priority-fencing-delay.scores
new file mode 100644
index 0000000000..b96175efbf
--- /dev/null
+++ b/cts/scheduler/priority-fencing-delay.scores
@@ -0,0 +1,301 @@
+Allocation scores:
+pcmk__clone_allocate: clvmd-clone allocation score on kiff-01: 0
+pcmk__clone_allocate: clvmd-clone allocation score on kiff-02: 0
+pcmk__clone_allocate: clvmd-clone allocation score on lxc-01_kiff-01: 0
+pcmk__clone_allocate: clvmd-clone allocation score on lxc-01_kiff-02: 0
+pcmk__clone_allocate: clvmd-clone allocation score on lxc-02_kiff-01: 0
+pcmk__clone_allocate: clvmd-clone allocation score on lxc-02_kiff-02: 0
+pcmk__clone_allocate: clvmd:0 allocation score on kiff-01: 1
+pcmk__clone_allocate: clvmd:0 allocation score on kiff-02: 0
+pcmk__clone_allocate: clvmd:0 allocation score on lxc-01_kiff-01: 0
+pcmk__clone_allocate: clvmd:0 allocation score on lxc-01_kiff-02: 0
+pcmk__clone_allocate: clvmd:0 allocation score on lxc-02_kiff-01: 0
+pcmk__clone_allocate: clvmd:0 allocation score on lxc-02_kiff-02: 0
+pcmk__clone_allocate: clvmd:1 allocation score on kiff-01: 0
+pcmk__clone_allocate: clvmd:1 allocation score on kiff-02: 1
+pcmk__clone_allocate: clvmd:1 allocation score on lxc-01_kiff-01: 0
+pcmk__clone_allocate: clvmd:1 allocation score on lxc-01_kiff-02: 0
+pcmk__clone_allocate: clvmd:1 allocation score on lxc-02_kiff-01: 0
+pcmk__clone_allocate: clvmd:1 allocation score on lxc-02_kiff-02: 0
+pcmk__clone_allocate: clvmd:2 allocation score on kiff-01: 0
+pcmk__clone_allocate: clvmd:2 allocation score on kiff-02: 0
+pcmk__clone_allocate: clvmd:2 allocation score on lxc-01_kiff-01: 0
+pcmk__clone_allocate: clvmd:2 allocation score on lxc-01_kiff-02: 0
+pcmk__clone_allocate: clvmd:2 allocation score on lxc-02_kiff-01: 0
+pcmk__clone_allocate: clvmd:2 allocation score on lxc-02_kiff-02: 0
+pcmk__clone_allocate: clvmd:3 allocation score on kiff-01: 0
+pcmk__clone_allocate: clvmd:3 allocation score on kiff-02: 0
+pcmk__clone_allocate: clvmd:3 allocation score on lxc-01_kiff-01: 0
+pcmk__clone_allocate: clvmd:3 allocation score on lxc-01_kiff-02: 0
+pcmk__clone_allocate: clvmd:3 allocation score on lxc-02_kiff-01: 0
+pcmk__clone_allocate: clvmd:3 allocation score on lxc-02_kiff-02: 0
+pcmk__clone_allocate: clvmd:4 allocation score on kiff-01: 0
+pcmk__clone_allocate: clvmd:4 allocation score on kiff-02: 0
+pcmk__clone_allocate: clvmd:4 allocation score on lxc-01_kiff-01: 0
+pcmk__clone_allocate: clvmd:4 allocation score on lxc-01_kiff-02: 0
+pcmk__clone_allocate: clvmd:4 allocation score on lxc-02_kiff-01: 0
+pcmk__clone_allocate: clvmd:4 allocation score on lxc-02_kiff-02: 0
+pcmk__clone_allocate: clvmd:5 allocation score on kiff-01: 0
+pcmk__clone_allocate: clvmd:5 allocation score on kiff-02: 0
+pcmk__clone_allocate: clvmd:5 allocation score on lxc-01_kiff-01: 0
+pcmk__clone_allocate: clvmd:5 allocation score on lxc-01_kiff-02: 0
+pcmk__clone_allocate: clvmd:5 allocation score on lxc-02_kiff-01: 0
+pcmk__clone_allocate: clvmd:5 allocation score on lxc-02_kiff-02: 0
+pcmk__clone_allocate: dlm-clone allocation score on kiff-01: 0
+pcmk__clone_allocate: dlm-clone allocation score on kiff-02: 0
+pcmk__clone_allocate: dlm-clone allocation score on lxc-01_kiff-01: -INFINITY
+pcmk__clone_allocate: dlm-clone allocation score on lxc-01_kiff-02: -INFINITY
+pcmk__clone_allocate: dlm-clone allocation score on lxc-02_kiff-01: -INFINITY
+pcmk__clone_allocate: dlm-clone allocation score on lxc-02_kiff-02: -INFINITY
+pcmk__clone_allocate: dlm:0 allocation score on kiff-01: 1
+pcmk__clone_allocate: dlm:0 allocation score on kiff-02: 0
+pcmk__clone_allocate: dlm:0 allocation score on lxc-01_kiff-01: -INFINITY
+pcmk__clone_allocate: dlm:0 allocation score on lxc-01_kiff-02: -INFINITY
+pcmk__clone_allocate: dlm:0 allocation score on lxc-02_kiff-01: -INFINITY
+pcmk__clone_allocate: dlm:0 allocation score on lxc-02_kiff-02: -INFINITY
+pcmk__clone_allocate: dlm:1 allocation score on kiff-01: 0
+pcmk__clone_allocate: dlm:1 allocation score on kiff-02: 1
+pcmk__clone_allocate: dlm:1 allocation score on lxc-01_kiff-01: -INFINITY
+pcmk__clone_allocate: dlm:1 allocation score on lxc-01_kiff-02: -INFINITY
+pcmk__clone_allocate: dlm:1 allocation score on lxc-02_kiff-01: -INFINITY
+pcmk__clone_allocate: dlm:1 allocation score on lxc-02_kiff-02: -INFINITY
+pcmk__clone_allocate: dlm:2 allocation score on kiff-01: 0
+pcmk__clone_allocate: dlm:2 allocation score on kiff-02: 0
+pcmk__clone_allocate: dlm:2 allocation score on lxc-01_kiff-01: -INFINITY
+pcmk__clone_allocate: dlm:2 allocation score on lxc-01_kiff-02: -INFINITY
+pcmk__clone_allocate: dlm:2 allocation score on lxc-02_kiff-01: -INFINITY
+pcmk__clone_allocate: dlm:2 allocation score on lxc-02_kiff-02: -INFINITY
+pcmk__clone_allocate: dlm:3 allocation score on kiff-01: 0
+pcmk__clone_allocate: dlm:3 allocation score on kiff-02: 0
+pcmk__clone_allocate: dlm:3 allocation score on lxc-01_kiff-01: -INFINITY
+pcmk__clone_allocate: dlm:3 allocation score on lxc-01_kiff-02: -INFINITY
+pcmk__clone_allocate: dlm:3 allocation score on lxc-02_kiff-01: -INFINITY
+pcmk__clone_allocate: dlm:3 allocation score on lxc-02_kiff-02: -INFINITY
+pcmk__clone_allocate: dlm:4 allocation score on kiff-01: 0
+pcmk__clone_allocate: dlm:4 allocation score on kiff-02: 0
+pcmk__clone_allocate: dlm:4 allocation score on lxc-01_kiff-01: -INFINITY
+pcmk__clone_allocate: dlm:4 allocation score on lxc-01_kiff-02: -INFINITY
+pcmk__clone_allocate: dlm:4 allocation score on lxc-02_kiff-01: -INFINITY
+pcmk__clone_allocate: dlm:4 allocation score on lxc-02_kiff-02: -INFINITY
+pcmk__clone_allocate: dlm:5 allocation score on kiff-01: 0
+pcmk__clone_allocate: dlm:5 allocation score on kiff-02: 0
+pcmk__clone_allocate: dlm:5 allocation score on lxc-01_kiff-01: -INFINITY
+pcmk__clone_allocate: dlm:5 allocation score on lxc-01_kiff-02: -INFINITY
+pcmk__clone_allocate: dlm:5 allocation score on lxc-02_kiff-01: -INFINITY
+pcmk__clone_allocate: dlm:5 allocation score on lxc-02_kiff-02: -INFINITY
+pcmk__clone_allocate: shared0-clone allocation score on kiff-01: 0
+pcmk__clone_allocate: shared0-clone allocation score on kiff-02: 0
+pcmk__clone_allocate: shared0-clone allocation score on lxc-01_kiff-01: 0
+pcmk__clone_allocate: shared0-clone allocation score on lxc-01_kiff-02: 0
+pcmk__clone_allocate: shared0-clone allocation score on lxc-02_kiff-01: 0
+pcmk__clone_allocate: shared0-clone allocation score on lxc-02_kiff-02: 0
+pcmk__clone_allocate: shared0:0 allocation score on kiff-01: 1
+pcmk__clone_allocate: shared0:0 allocation score on kiff-02: 0
+pcmk__clone_allocate: shared0:0 allocation score on lxc-01_kiff-01: 0
+pcmk__clone_allocate: shared0:0 allocation score on lxc-01_kiff-02: 0
+pcmk__clone_allocate: shared0:0 allocation score on lxc-02_kiff-01: 0
+pcmk__clone_allocate: shared0:0 allocation score on lxc-02_kiff-02: 0
+pcmk__clone_allocate: shared0:1 allocation score on kiff-01: 0
+pcmk__clone_allocate: shared0:1 allocation score on kiff-02: 1
+pcmk__clone_allocate: shared0:1 allocation score on lxc-01_kiff-01: 0
+pcmk__clone_allocate: shared0:1 allocation score on lxc-01_kiff-02: 0
+pcmk__clone_allocate: shared0:1 allocation score on lxc-02_kiff-01: 0
+pcmk__clone_allocate: shared0:1 allocation score on lxc-02_kiff-02: 0
+pcmk__clone_allocate: shared0:2 allocation score on kiff-01: 0
+pcmk__clone_allocate: shared0:2 allocation score on kiff-02: 0
+pcmk__clone_allocate: shared0:2 allocation score on lxc-01_kiff-01: 0
+pcmk__clone_allocate: shared0:2 allocation score on lxc-01_kiff-02: 0
+pcmk__clone_allocate: shared0:2 allocation score on lxc-02_kiff-01: 0
+pcmk__clone_allocate: shared0:2 allocation score on lxc-02_kiff-02: 0
+pcmk__clone_allocate: shared0:3 allocation score on kiff-01: 0
+pcmk__clone_allocate: shared0:3 allocation score on kiff-02: 0
+pcmk__clone_allocate: shared0:3 allocation score on lxc-01_kiff-01: 0
+pcmk__clone_allocate: shared0:3 allocation score on lxc-01_kiff-02: 0
+pcmk__clone_allocate: shared0:3 allocation score on lxc-02_kiff-01: 0
+pcmk__clone_allocate: shared0:3 allocation score on lxc-02_kiff-02: 0
+pcmk__clone_allocate: shared0:4 allocation score on kiff-01: 0
+pcmk__clone_allocate: shared0:4 allocation score on kiff-02: 0
+pcmk__clone_allocate: shared0:4 allocation score on lxc-01_kiff-01: 0
+pcmk__clone_allocate: shared0:4 allocation score on lxc-01_kiff-02: 0
+pcmk__clone_allocate: shared0:4 allocation score on lxc-02_kiff-01: 0
+pcmk__clone_allocate: shared0:4 allocation score on lxc-02_kiff-02: 0
+pcmk__clone_allocate: shared0:5 allocation score on kiff-01: 0
+pcmk__clone_allocate: shared0:5 allocation score on kiff-02: 0
+pcmk__clone_allocate: shared0:5 allocation score on lxc-01_kiff-01: 0
+pcmk__clone_allocate: shared0:5 allocation score on lxc-01_kiff-02: 0
+pcmk__clone_allocate: shared0:5 allocation score on lxc-02_kiff-01: 0
+pcmk__clone_allocate: shared0:5 allocation score on lxc-02_kiff-02: 0
+pcmk__native_allocate: R-lxc-01_kiff-01 allocation score on kiff-01: -INFINITY
+pcmk__native_allocate: R-lxc-01_kiff-01 allocation score on kiff-02: 0
+pcmk__native_allocate: R-lxc-01_kiff-01 allocation score on lxc-01_kiff-01: -INFINITY
+pcmk__native_allocate: R-lxc-01_kiff-01 allocation score on lxc-01_kiff-02: -INFINITY
+pcmk__native_allocate: R-lxc-01_kiff-01 allocation score on lxc-02_kiff-01: -INFINITY
+pcmk__native_allocate: R-lxc-01_kiff-01 allocation score on lxc-02_kiff-02: -INFINITY
+pcmk__native_allocate: R-lxc-01_kiff-02 allocation score on kiff-01: -INFINITY
+pcmk__native_allocate: R-lxc-01_kiff-02 allocation score on kiff-02: 100
+pcmk__native_allocate: R-lxc-01_kiff-02 allocation score on lxc-01_kiff-01: -INFINITY
+pcmk__native_allocate: R-lxc-01_kiff-02 allocation score on lxc-01_kiff-02: -INFINITY
+pcmk__native_allocate: R-lxc-01_kiff-02 allocation score on lxc-02_kiff-01: -INFINITY
+pcmk__native_allocate: R-lxc-01_kiff-02 allocation score on lxc-02_kiff-02: -INFINITY
+pcmk__native_allocate: R-lxc-02_kiff-01 allocation score on kiff-01: -INFINITY
+pcmk__native_allocate: R-lxc-02_kiff-01 allocation score on kiff-02: 0
+pcmk__native_allocate: R-lxc-02_kiff-01 allocation score on lxc-01_kiff-01: -INFINITY
+pcmk__native_allocate: R-lxc-02_kiff-01 allocation score on lxc-01_kiff-02: -INFINITY
+pcmk__native_allocate: R-lxc-02_kiff-01 allocation score on lxc-02_kiff-01: -INFINITY
+pcmk__native_allocate: R-lxc-02_kiff-01 allocation score on lxc-02_kiff-02: -INFINITY
+pcmk__native_allocate: R-lxc-02_kiff-02 allocation score on kiff-01: -INFINITY
+pcmk__native_allocate: R-lxc-02_kiff-02 allocation score on kiff-02: 100
+pcmk__native_allocate: R-lxc-02_kiff-02 allocation score on lxc-01_kiff-01: -INFINITY
+pcmk__native_allocate: R-lxc-02_kiff-02 allocation score on lxc-01_kiff-02: -INFINITY
+pcmk__native_allocate: R-lxc-02_kiff-02 allocation score on lxc-02_kiff-01: -INFINITY
+pcmk__native_allocate: R-lxc-02_kiff-02 allocation score on lxc-02_kiff-02: -INFINITY
+pcmk__native_allocate: clvmd:0 allocation score on kiff-01: -INFINITY
+pcmk__native_allocate: clvmd:0 allocation score on kiff-02: -INFINITY
+pcmk__native_allocate: clvmd:0 allocation score on lxc-01_kiff-01: -INFINITY
+pcmk__native_allocate: clvmd:0 allocation score on lxc-01_kiff-02: -INFINITY
+pcmk__native_allocate: clvmd:0 allocation score on lxc-02_kiff-01: -INFINITY
+pcmk__native_allocate: clvmd:0 allocation score on lxc-02_kiff-02: -INFINITY
+pcmk__native_allocate: clvmd:1 allocation score on kiff-01: -INFINITY
+pcmk__native_allocate: clvmd:1 allocation score on kiff-02: 1
+pcmk__native_allocate: clvmd:1 allocation score on lxc-01_kiff-01: -INFINITY
+pcmk__native_allocate: clvmd:1 allocation score on lxc-01_kiff-02: -INFINITY
+pcmk__native_allocate: clvmd:1 allocation score on lxc-02_kiff-01: -INFINITY
+pcmk__native_allocate: clvmd:1 allocation score on lxc-02_kiff-02: -INFINITY
+pcmk__native_allocate: clvmd:2 allocation score on kiff-01: -INFINITY
+pcmk__native_allocate: clvmd:2 allocation score on kiff-02: -INFINITY
+pcmk__native_allocate: clvmd:2 allocation score on lxc-01_kiff-01: -INFINITY
+pcmk__native_allocate: clvmd:2 allocation score on lxc-01_kiff-02: -INFINITY
+pcmk__native_allocate: clvmd:2 allocation score on lxc-02_kiff-01: -INFINITY
+pcmk__native_allocate: clvmd:2 allocation score on lxc-02_kiff-02: -INFINITY
+pcmk__native_allocate: clvmd:3 allocation score on kiff-01: -INFINITY
+pcmk__native_allocate: clvmd:3 allocation score on kiff-02: -INFINITY
+pcmk__native_allocate: clvmd:3 allocation score on lxc-01_kiff-01: -INFINITY
+pcmk__native_allocate: clvmd:3 allocation score on lxc-01_kiff-02: -INFINITY
+pcmk__native_allocate: clvmd:3 allocation score on lxc-02_kiff-01: -INFINITY
+pcmk__native_allocate: clvmd:3 allocation score on lxc-02_kiff-02: -INFINITY
+pcmk__native_allocate: clvmd:4 allocation score on kiff-01: -INFINITY
+pcmk__native_allocate: clvmd:4 allocation score on kiff-02: -INFINITY
+pcmk__native_allocate: clvmd:4 allocation score on lxc-01_kiff-01: -INFINITY
+pcmk__native_allocate: clvmd:4 allocation score on lxc-01_kiff-02: -INFINITY
+pcmk__native_allocate: clvmd:4 allocation score on lxc-02_kiff-01: -INFINITY
+pcmk__native_allocate: clvmd:4 allocation score on lxc-02_kiff-02: -INFINITY
+pcmk__native_allocate: clvmd:5 allocation score on kiff-01: -INFINITY
+pcmk__native_allocate: clvmd:5 allocation score on kiff-02: -INFINITY
+pcmk__native_allocate: clvmd:5 allocation score on lxc-01_kiff-01: -INFINITY
+pcmk__native_allocate: clvmd:5 allocation score on lxc-01_kiff-02: -INFINITY
+pcmk__native_allocate: clvmd:5 allocation score on lxc-02_kiff-01: -INFINITY
+pcmk__native_allocate: clvmd:5 allocation score on lxc-02_kiff-02: -INFINITY
+pcmk__native_allocate: dlm:0 allocation score on kiff-01: -INFINITY
+pcmk__native_allocate: dlm:0 allocation score on kiff-02: -INFINITY
+pcmk__native_allocate: dlm:0 allocation score on lxc-01_kiff-01: -INFINITY
+pcmk__native_allocate: dlm:0 allocation score on lxc-01_kiff-02: -INFINITY
+pcmk__native_allocate: dlm:0 allocation score on lxc-02_kiff-01: -INFINITY
+pcmk__native_allocate: dlm:0 allocation score on lxc-02_kiff-02: -INFINITY
+pcmk__native_allocate: dlm:1 allocation score on kiff-01: -INFINITY
+pcmk__native_allocate: dlm:1 allocation score on kiff-02: 1
+pcmk__native_allocate: dlm:1 allocation score on lxc-01_kiff-01: -INFINITY
+pcmk__native_allocate: dlm:1 allocation score on lxc-01_kiff-02: -INFINITY
+pcmk__native_allocate: dlm:1 allocation score on lxc-02_kiff-01: -INFINITY
+pcmk__native_allocate: dlm:1 allocation score on lxc-02_kiff-02: -INFINITY
+pcmk__native_allocate: dlm:2 allocation score on kiff-01: -INFINITY
+pcmk__native_allocate: dlm:2 allocation score on kiff-02: -INFINITY
+pcmk__native_allocate: dlm:2 allocation score on lxc-01_kiff-01: -INFINITY
+pcmk__native_allocate: dlm:2 allocation score on lxc-01_kiff-02: -INFINITY
+pcmk__native_allocate: dlm:2 allocation score on lxc-02_kiff-01: -INFINITY
+pcmk__native_allocate: dlm:2 allocation score on lxc-02_kiff-02: -INFINITY
+pcmk__native_allocate: dlm:3 allocation score on kiff-01: -INFINITY
+pcmk__native_allocate: dlm:3 allocation score on kiff-02: -INFINITY
+pcmk__native_allocate: dlm:3 allocation score on lxc-01_kiff-01: -INFINITY
+pcmk__native_allocate: dlm:3 allocation score on lxc-01_kiff-02: -INFINITY
+pcmk__native_allocate: dlm:3 allocation score on lxc-02_kiff-01: -INFINITY
+pcmk__native_allocate: dlm:3 allocation score on lxc-02_kiff-02: -INFINITY
+pcmk__native_allocate: dlm:4 allocation score on kiff-01: -INFINITY
+pcmk__native_allocate: dlm:4 allocation score on kiff-02: -INFINITY
+pcmk__native_allocate: dlm:4 allocation score on lxc-01_kiff-01: -INFINITY
+pcmk__native_allocate: dlm:4 allocation score on lxc-01_kiff-02: -INFINITY
+pcmk__native_allocate: dlm:4 allocation score on lxc-02_kiff-01: -INFINITY
+pcmk__native_allocate: dlm:4 allocation score on lxc-02_kiff-02: -INFINITY
+pcmk__native_allocate: dlm:5 allocation score on kiff-01: -INFINITY
+pcmk__native_allocate: dlm:5 allocation score on kiff-02: -INFINITY
+pcmk__native_allocate: dlm:5 allocation score on lxc-01_kiff-01: -INFINITY
+pcmk__native_allocate: dlm:5 allocation score on lxc-01_kiff-02: -INFINITY
+pcmk__native_allocate: dlm:5 allocation score on lxc-02_kiff-01: -INFINITY
+pcmk__native_allocate: dlm:5 allocation score on lxc-02_kiff-02: -INFINITY
+pcmk__native_allocate: fence-kiff-01 allocation score on kiff-01: 0
+pcmk__native_allocate: fence-kiff-01 allocation score on kiff-02: 0
+pcmk__native_allocate: fence-kiff-01 allocation score on lxc-01_kiff-01: -INFINITY
+pcmk__native_allocate: fence-kiff-01 allocation score on lxc-01_kiff-02: -INFINITY
+pcmk__native_allocate: fence-kiff-01 allocation score on lxc-02_kiff-01: -INFINITY
+pcmk__native_allocate: fence-kiff-01 allocation score on lxc-02_kiff-02: -INFINITY
+pcmk__native_allocate: fence-kiff-02 allocation score on kiff-01: 0
+pcmk__native_allocate: fence-kiff-02 allocation score on kiff-02: 0
+pcmk__native_allocate: fence-kiff-02 allocation score on lxc-01_kiff-01: -INFINITY
+pcmk__native_allocate: fence-kiff-02 allocation score on lxc-01_kiff-02: -INFINITY
+pcmk__native_allocate: fence-kiff-02 allocation score on lxc-02_kiff-01: -INFINITY
+pcmk__native_allocate: fence-kiff-02 allocation score on lxc-02_kiff-02: -INFINITY
+pcmk__native_allocate: lxc-01_kiff-01 allocation score on kiff-01: -INFINITY
+pcmk__native_allocate: lxc-01_kiff-01 allocation score on kiff-02: 0
+pcmk__native_allocate: lxc-01_kiff-01 allocation score on lxc-01_kiff-01: -INFINITY
+pcmk__native_allocate: lxc-01_kiff-01 allocation score on lxc-01_kiff-02: -INFINITY
+pcmk__native_allocate: lxc-01_kiff-01 allocation score on lxc-02_kiff-01: -INFINITY
+pcmk__native_allocate: lxc-01_kiff-01 allocation score on lxc-02_kiff-02: -INFINITY
+pcmk__native_allocate: lxc-01_kiff-02 allocation score on kiff-01: -INFINITY
+pcmk__native_allocate: lxc-01_kiff-02 allocation score on kiff-02: 0
+pcmk__native_allocate: lxc-01_kiff-02 allocation score on lxc-01_kiff-01: -INFINITY
+pcmk__native_allocate: lxc-01_kiff-02 allocation score on lxc-01_kiff-02: -INFINITY
+pcmk__native_allocate: lxc-01_kiff-02 allocation score on lxc-02_kiff-01: -INFINITY
+pcmk__native_allocate: lxc-01_kiff-02 allocation score on lxc-02_kiff-02: -INFINITY
+pcmk__native_allocate: lxc-02_kiff-01 allocation score on kiff-01: -INFINITY
+pcmk__native_allocate: lxc-02_kiff-01 allocation score on kiff-02: 0
+pcmk__native_allocate: lxc-02_kiff-01 allocation score on lxc-01_kiff-01: -INFINITY
+pcmk__native_allocate: lxc-02_kiff-01 allocation score on lxc-01_kiff-02: -INFINITY
+pcmk__native_allocate: lxc-02_kiff-01 allocation score on lxc-02_kiff-01: -INFINITY
+pcmk__native_allocate: lxc-02_kiff-01 allocation score on lxc-02_kiff-02: -INFINITY
+pcmk__native_allocate: lxc-02_kiff-02 allocation score on kiff-01: -INFINITY
+pcmk__native_allocate: lxc-02_kiff-02 allocation score on kiff-02: 0
+pcmk__native_allocate: lxc-02_kiff-02 allocation score on lxc-01_kiff-01: -INFINITY
+pcmk__native_allocate: lxc-02_kiff-02 allocation score on lxc-01_kiff-02: -INFINITY
+pcmk__native_allocate: lxc-02_kiff-02 allocation score on lxc-02_kiff-01: -INFINITY
+pcmk__native_allocate: lxc-02_kiff-02 allocation score on lxc-02_kiff-02: -INFINITY
+pcmk__native_allocate: shared0:0 allocation score on kiff-01: -INFINITY
+pcmk__native_allocate: shared0:0 allocation score on kiff-02: -INFINITY
+pcmk__native_allocate: shared0:0 allocation score on lxc-01_kiff-01: -INFINITY
+pcmk__native_allocate: shared0:0 allocation score on lxc-01_kiff-02: -INFINITY
+pcmk__native_allocate: shared0:0 allocation score on lxc-02_kiff-01: -INFINITY
+pcmk__native_allocate: shared0:0 allocation score on lxc-02_kiff-02: -INFINITY
+pcmk__native_allocate: shared0:1 allocation score on kiff-01: -INFINITY
+pcmk__native_allocate: shared0:1 allocation score on kiff-02: 1
+pcmk__native_allocate: shared0:1 allocation score on lxc-01_kiff-01: -INFINITY
+pcmk__native_allocate: shared0:1 allocation score on lxc-01_kiff-02: -INFINITY
+pcmk__native_allocate: shared0:1 allocation score on lxc-02_kiff-01: -INFINITY
+pcmk__native_allocate: shared0:1 allocation score on lxc-02_kiff-02: -INFINITY
+pcmk__native_allocate: shared0:2 allocation score on kiff-01: -INFINITY
+pcmk__native_allocate: shared0:2 allocation score on kiff-02: -INFINITY
+pcmk__native_allocate: shared0:2 allocation score on lxc-01_kiff-01: -INFINITY
+pcmk__native_allocate: shared0:2 allocation score on lxc-01_kiff-02: -INFINITY
+pcmk__native_allocate: shared0:2 allocation score on lxc-02_kiff-01: -INFINITY
+pcmk__native_allocate: shared0:2 allocation score on lxc-02_kiff-02: -INFINITY
+pcmk__native_allocate: shared0:3 allocation score on kiff-01: -INFINITY
+pcmk__native_allocate: shared0:3 allocation score on kiff-02: -INFINITY
+pcmk__native_allocate: shared0:3 allocation score on lxc-01_kiff-01: -INFINITY
+pcmk__native_allocate: shared0:3 allocation score on lxc-01_kiff-02: -INFINITY
+pcmk__native_allocate: shared0:3 allocation score on lxc-02_kiff-01: -INFINITY
+pcmk__native_allocate: shared0:3 allocation score on lxc-02_kiff-02: -INFINITY
+pcmk__native_allocate: shared0:4 allocation score on kiff-01: -INFINITY
+pcmk__native_allocate: shared0:4 allocation score on kiff-02: -INFINITY
+pcmk__native_allocate: shared0:4 allocation score on lxc-01_kiff-01: -INFINITY
+pcmk__native_allocate: shared0:4 allocation score on lxc-01_kiff-02: -INFINITY
+pcmk__native_allocate: shared0:4 allocation score on lxc-02_kiff-01: -INFINITY
+pcmk__native_allocate: shared0:4 allocation score on lxc-02_kiff-02: -INFINITY
+pcmk__native_allocate: shared0:5 allocation score on kiff-01: -INFINITY
+pcmk__native_allocate: shared0:5 allocation score on kiff-02: -INFINITY
+pcmk__native_allocate: shared0:5 allocation score on lxc-01_kiff-01: -INFINITY
+pcmk__native_allocate: shared0:5 allocation score on lxc-01_kiff-02: -INFINITY
+pcmk__native_allocate: shared0:5 allocation score on lxc-02_kiff-01: -INFINITY
+pcmk__native_allocate: shared0:5 allocation score on lxc-02_kiff-02: -INFINITY
+pcmk__native_allocate: vm-fs allocation score on kiff-01: 0
+pcmk__native_allocate: vm-fs allocation score on kiff-02: 0
+pcmk__native_allocate: vm-fs allocation score on lxc-01_kiff-01: 0
+pcmk__native_allocate: vm-fs allocation score on lxc-01_kiff-02: 0
+pcmk__native_allocate: vm-fs allocation score on lxc-02_kiff-01: 0
+pcmk__native_allocate: vm-fs allocation score on lxc-02_kiff-02: 0
diff --git a/cts/scheduler/priority-fencing-delay.summary b/cts/scheduler/priority-fencing-delay.summary
new file mode 100644
index 0000000000..20a9a4a91b
--- /dev/null
+++ b/cts/scheduler/priority-fencing-delay.summary
@@ -0,0 +1,102 @@
+
+Current cluster status:
+Node kiff-01 (1): UNCLEAN (offline)
+Online: [ kiff-02 ]
+GuestOnline: [ lxc-01_kiff-02:R-lxc-01_kiff-02 lxc-02_kiff-02:R-lxc-02_kiff-02 ]
+
+ vm-fs (ocf::heartbeat:Filesystem): FAILED lxc-01_kiff-01
+ R-lxc-01_kiff-02 (ocf::heartbeat:VirtualDomain): Started kiff-02
+ fence-kiff-01 (stonith:fence_ipmilan): Started kiff-02
+ fence-kiff-02 (stonith:fence_ipmilan): Started kiff-01 (UNCLEAN)
+ Clone Set: dlm-clone [dlm]
+ dlm (ocf::pacemaker:controld): Started kiff-01 (UNCLEAN)
+ Started: [ kiff-02 ]
+ Stopped: [ lxc-01_kiff-01 lxc-01_kiff-02 lxc-02_kiff-01 lxc-02_kiff-02 ]
+ Clone Set: clvmd-clone [clvmd]
+ clvmd (ocf::heartbeat:clvm): Started kiff-01 (UNCLEAN)
+ Started: [ kiff-02 ]
+ Stopped: [ lxc-01_kiff-01 lxc-01_kiff-02 lxc-02_kiff-01 lxc-02_kiff-02 ]
+ Clone Set: shared0-clone [shared0]
+ shared0 (ocf::heartbeat:Filesystem): Started kiff-01 (UNCLEAN)
+ Started: [ kiff-02 ]
+ Stopped: [ lxc-01_kiff-01 lxc-01_kiff-02 lxc-02_kiff-01 lxc-02_kiff-02 ]
+ R-lxc-01_kiff-01 (ocf::heartbeat:VirtualDomain): FAILED kiff-01 (UNCLEAN)
+ R-lxc-02_kiff-01 (ocf::heartbeat:VirtualDomain): Started kiff-01 (UNCLEAN)
+ R-lxc-02_kiff-02 (ocf::heartbeat:VirtualDomain): Started kiff-02
+
+Transition Summary:
+ * Fence (reboot) lxc-02_kiff-01 (resource: R-lxc-02_kiff-01) 'guest is unclean'
+ * Fence (reboot) lxc-01_kiff-01 (resource: R-lxc-01_kiff-01) 'guest is unclean'
+ * Fence (reboot) kiff-01 'peer is no longer part of the cluster'
+ * Recover vm-fs ( lxc-01_kiff-01 )
+ * Move fence-kiff-02 ( kiff-01 -> kiff-02 )
+ * Stop dlm:0 ( kiff-01 ) due to node availability
+ * Stop clvmd:0 ( kiff-01 ) due to node availability
+ * Stop shared0:0 ( kiff-01 ) due to node availability
+ * Recover R-lxc-01_kiff-01 ( kiff-01 -> kiff-02 )
+ * Move R-lxc-02_kiff-01 ( kiff-01 -> kiff-02 )
+ * Move lxc-01_kiff-01 ( kiff-01 -> kiff-02 )
+ * Move lxc-02_kiff-01 ( kiff-01 -> kiff-02 )
+
+Executing cluster transition:
+ * Resource action: vm-fs monitor on lxc-02_kiff-02
+ * Resource action: vm-fs monitor on lxc-01_kiff-02
+ * Pseudo action: fence-kiff-02_stop_0
+ * Resource action: dlm monitor on lxc-02_kiff-02
+ * Resource action: dlm monitor on lxc-01_kiff-02
+ * Resource action: clvmd monitor on lxc-02_kiff-02
+ * Resource action: clvmd monitor on lxc-01_kiff-02
+ * Resource action: shared0 monitor on lxc-02_kiff-02
+ * Resource action: shared0 monitor on lxc-01_kiff-02
+ * Pseudo action: lxc-01_kiff-01_stop_0
+ * Pseudo action: lxc-02_kiff-01_stop_0
+ * Fencing kiff-01 (reboot)
+ * Pseudo action: R-lxc-01_kiff-01_stop_0
+ * Pseudo action: R-lxc-02_kiff-01_stop_0
+ * Pseudo action: stonith-lxc-02_kiff-01-reboot on lxc-02_kiff-01
+ * Pseudo action: stonith-lxc-01_kiff-01-reboot on lxc-01_kiff-01
+ * Pseudo action: vm-fs_stop_0
+ * Resource action: fence-kiff-02 start on kiff-02
+ * Pseudo action: shared0-clone_stop_0
+ * Resource action: R-lxc-01_kiff-01 start on kiff-02
+ * Resource action: R-lxc-02_kiff-01 start on kiff-02
+ * Resource action: lxc-01_kiff-01 start on kiff-02
+ * Resource action: lxc-02_kiff-01 start on kiff-02
+ * Resource action: vm-fs start on lxc-01_kiff-01
+ * Resource action: fence-kiff-02 monitor=60000 on kiff-02
+ * Pseudo action: shared0_stop_0
+ * Pseudo action: shared0-clone_stopped_0
+ * Resource action: R-lxc-01_kiff-01 monitor=10000 on kiff-02
+ * Resource action: R-lxc-02_kiff-01 monitor=10000 on kiff-02
+ * Resource action: lxc-01_kiff-01 monitor=30000 on kiff-02
+ * Resource action: lxc-02_kiff-01 monitor=30000 on kiff-02
+ * Resource action: vm-fs monitor=20000 on lxc-01_kiff-01
+ * Pseudo action: clvmd-clone_stop_0
+ * Pseudo action: clvmd_stop_0
+ * Pseudo action: clvmd-clone_stopped_0
+ * Pseudo action: dlm-clone_stop_0
+ * Pseudo action: dlm_stop_0
+ * Pseudo action: dlm-clone_stopped_0
+
+Revised cluster status:
+Online: [ kiff-02 ]
+OFFLINE: [ kiff-01 ]
+GuestOnline: [ lxc-01_kiff-01:R-lxc-01_kiff-01 lxc-01_kiff-02:R-lxc-01_kiff-02 lxc-02_kiff-01:R-lxc-02_kiff-01 lxc-02_kiff-02:R-lxc-02_kiff-02 ]
+
+ vm-fs (ocf::heartbeat:Filesystem): Started lxc-01_kiff-01
+ R-lxc-01_kiff-02 (ocf::heartbeat:VirtualDomain): Started kiff-02
+ fence-kiff-01 (stonith:fence_ipmilan): Started kiff-02
+ fence-kiff-02 (stonith:fence_ipmilan): Started kiff-02
+ Clone Set: dlm-clone [dlm]
+ Started: [ kiff-02 ]
+ Stopped: [ kiff-01 lxc-01_kiff-01 lxc-01_kiff-02 lxc-02_kiff-01 lxc-02_kiff-02 ]
+ Clone Set: clvmd-clone [clvmd]
+ Started: [ kiff-02 ]
+ Stopped: [ kiff-01 lxc-01_kiff-01 lxc-01_kiff-02 lxc-02_kiff-01 lxc-02_kiff-02 ]
+ Clone Set: shared0-clone [shared0]
+ Started: [ kiff-02 ]
+ Stopped: [ kiff-01 lxc-01_kiff-01 lxc-01_kiff-02 lxc-02_kiff-01 lxc-02_kiff-02 ]
+ R-lxc-01_kiff-01 (ocf::heartbeat:VirtualDomain): Started kiff-02
+ R-lxc-02_kiff-01 (ocf::heartbeat:VirtualDomain): Started kiff-02
+ R-lxc-02_kiff-02 (ocf::heartbeat:VirtualDomain): Started kiff-02
+
diff --git a/cts/scheduler/priority-fencing-delay.xml b/cts/scheduler/priority-fencing-delay.xml
new file mode 100644
index 0000000000..04d1d3e290
--- /dev/null
+++ b/cts/scheduler/priority-fencing-delay.xml
@@ -0,0 +1,351 @@
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
diff --git a/daemons/controld/controld_fencing.c b/daemons/controld/controld_fencing.c
index c93efb3dcd..97fd4ad928 100644
--- a/daemons/controld/controld_fencing.c
+++ b/daemons/controld/controld_fencing.c
@@ -1,955 +1,958 @@
/*
* Copyright 2004-2020 the Pacemaker project contributors
*
* The version control history for this file may have further details.
*
* This source code is licensed under the GNU General Public License version 2
* or later (GPLv2+) WITHOUT ANY WARRANTY.
*/
#include
#include
#include
#include
#include
static void
tengine_stonith_history_synced(stonith_t *st, stonith_event_t *st_event);
/*
* stonith failure counting
*
* We don't want to get stuck in a permanent fencing loop. Keep track of the
* number of fencing failures for each target node, and the most we'll restart a
* transition for.
*/
struct st_fail_rec {
int count;
};
static bool fence_reaction_panic = FALSE;
static unsigned long int stonith_max_attempts = 10;
static GHashTable *stonith_failures = NULL;
// crmd_opts defines default for stonith-max-attempts, so value is never NULL
void
update_stonith_max_attempts(const char *value)
{
if (safe_str_eq(value, CRM_INFINITY_S)) {
stonith_max_attempts = CRM_SCORE_INFINITY;
} else {
stonith_max_attempts = (unsigned long int) crm_parse_ll(value, NULL);
}
}
void
set_fence_reaction(const char *reaction_s)
{
if (safe_str_eq(reaction_s, "panic")) {
fence_reaction_panic = TRUE;
} else {
if (safe_str_neq(reaction_s, "stop")) {
crm_warn("Invalid value '%s' for %s, using 'stop'",
reaction_s, XML_CONFIG_ATTR_FENCE_REACTION);
}
fence_reaction_panic = FALSE;
}
}
static gboolean
too_many_st_failures(const char *target)
{
GHashTableIter iter;
const char *key = NULL;
struct st_fail_rec *value = NULL;
if (stonith_failures == NULL) {
return FALSE;
}
if (target == NULL) {
g_hash_table_iter_init(&iter, stonith_failures);
while (g_hash_table_iter_next(&iter, (gpointer *) &key,
(gpointer *) &value)) {
if (value->count >= stonith_max_attempts) {
target = (const char*)key;
goto too_many;
}
}
} else {
value = g_hash_table_lookup(stonith_failures, target);
if ((value != NULL) && (value->count >= stonith_max_attempts)) {
goto too_many;
}
}
return FALSE;
too_many:
crm_warn("Too many failures (%d) to fence %s, giving up",
value->count, target);
return TRUE;
}
/*!
* \internal
* \brief Reset a stonith fail count
*
* \param[in] target Name of node to reset, or NULL for all
*/
void
st_fail_count_reset(const char *target)
{
if (stonith_failures == NULL) {
return;
}
if (target) {
struct st_fail_rec *rec = NULL;
rec = g_hash_table_lookup(stonith_failures, target);
if (rec) {
rec->count = 0;
}
} else {
GHashTableIter iter;
const char *key = NULL;
struct st_fail_rec *rec = NULL;
g_hash_table_iter_init(&iter, stonith_failures);
while (g_hash_table_iter_next(&iter, (gpointer *) &key,
(gpointer *) &rec)) {
rec->count = 0;
}
}
}
static void
st_fail_count_increment(const char *target)
{
struct st_fail_rec *rec = NULL;
if (stonith_failures == NULL) {
stonith_failures = crm_str_table_new();
}
rec = g_hash_table_lookup(stonith_failures, target);
if (rec) {
rec->count++;
} else {
rec = malloc(sizeof(struct st_fail_rec));
if(rec == NULL) {
return;
}
rec->count = 1;
g_hash_table_insert(stonith_failures, strdup(target), rec);
}
}
/* end stonith fail count functions */
static void
cib_fencing_updated(xmlNode *msg, int call_id, int rc, xmlNode *output,
void *user_data)
{
if (rc < pcmk_ok) {
crm_err("Fencing update %d for %s: failed - %s (%d)",
call_id, (char *)user_data, pcmk_strerror(rc), rc);
crm_log_xml_warn(msg, "Failed update");
abort_transition(INFINITY, tg_shutdown, "CIB update failed", NULL);
} else {
crm_info("Fencing update %d for %s: complete", call_id, (char *)user_data);
}
}
static void
send_stonith_update(crm_action_t *action, const char *target, const char *uuid)
{
int rc = pcmk_ok;
crm_node_t *peer = NULL;
/* We (usually) rely on the membership layer to do node_update_cluster,
* and the peer status callback to do node_update_peer, because the node
* might have already rejoined before we get the stonith result here.
*/
int flags = node_update_join | node_update_expected;
/* zero out the node-status & remove all LRM status info */
xmlNode *node_state = NULL;
CRM_CHECK(target != NULL, return);
CRM_CHECK(uuid != NULL, return);
/* Make sure the membership and join caches are accurate */
peer = crm_get_peer_full(0, target, CRM_GET_PEER_ANY);
CRM_CHECK(peer != NULL, return);
if (peer->state == NULL) {
/* Usually, we rely on the membership layer to update the cluster state
* in the CIB. However, if the node has never been seen, do it here, so
* the node is not considered unclean.
*/
flags |= node_update_cluster;
}
if (peer->uuid == NULL) {
crm_info("Recording uuid '%s' for node '%s'", uuid, target);
peer->uuid = strdup(uuid);
}
crmd_peer_down(peer, TRUE);
/* Generate a node state update for the CIB */
node_state = create_node_state_update(peer, flags, NULL, __FUNCTION__);
/* we have to mark whether or not remote nodes have already been fenced */
if (peer->flags & crm_remote_node) {
time_t now = time(NULL);
char *now_s = crm_itoa(now);
crm_xml_add(node_state, XML_NODE_IS_FENCED, now_s);
free(now_s);
}
/* Force our known ID */
crm_xml_add(node_state, XML_ATTR_UUID, uuid);
rc = fsa_cib_conn->cmds->update(fsa_cib_conn, XML_CIB_TAG_STATUS, node_state,
cib_quorum_override | cib_scope_local | cib_can_create);
/* Delay processing the trigger until the update completes */
crm_debug("Sending fencing update %d for %s", rc, target);
fsa_register_cib_callback(rc, FALSE, strdup(target), cib_fencing_updated);
/* Make sure it sticks */
/* fsa_cib_conn->cmds->bump_epoch(fsa_cib_conn, cib_quorum_override|cib_scope_local); */
controld_delete_node_state(peer->uname, controld_section_all,
cib_scope_local);
free_xml(node_state);
return;
}
/*!
* \internal
* \brief Abort transition due to stonith failure
*
* \param[in] abort_action Whether to restart or stop transition
* \param[in] target Don't restart if this (NULL for any) has too many failures
* \param[in] reason Log this stonith action XML as abort reason (or NULL)
*/
static void
abort_for_stonith_failure(enum transition_action abort_action,
const char *target, xmlNode *reason)
{
/* If stonith repeatedly fails, we eventually give up on starting a new
* transition for that reason.
*/
if ((abort_action != tg_stop) && too_many_st_failures(target)) {
abort_action = tg_stop;
}
abort_transition(INFINITY, abort_action, "Stonith failed", reason);
}
/*
* stonith cleanup list
*
* If the DC is shot, proper notifications might not go out.
* The stonith cleanup list allows the cluster to (re-)send
* notifications once a new DC is elected.
*/
static GListPtr stonith_cleanup_list = NULL;
/*!
* \internal
* \brief Add a node to the stonith cleanup list
*
* \param[in] target Name of node to add
*/
void
add_stonith_cleanup(const char *target) {
stonith_cleanup_list = g_list_append(stonith_cleanup_list, strdup(target));
}
/*!
* \internal
* \brief Remove a node from the stonith cleanup list
*
* \param[in] Name of node to remove
*/
void
remove_stonith_cleanup(const char *target)
{
GListPtr iter = stonith_cleanup_list;
while (iter != NULL) {
GListPtr tmp = iter;
char *iter_name = tmp->data;
iter = iter->next;
if (safe_str_eq(target, iter_name)) {
crm_trace("Removing %s from the cleanup list", iter_name);
stonith_cleanup_list = g_list_delete_link(stonith_cleanup_list, tmp);
free(iter_name);
}
}
}
/*!
* \internal
* \brief Purge all entries from the stonith cleanup list
*/
void
purge_stonith_cleanup()
{
if (stonith_cleanup_list) {
GListPtr iter = NULL;
for (iter = stonith_cleanup_list; iter != NULL; iter = iter->next) {
char *target = iter->data;
crm_info("Purging %s from stonith cleanup list", target);
free(target);
}
g_list_free(stonith_cleanup_list);
stonith_cleanup_list = NULL;
}
}
/*!
* \internal
* \brief Send stonith updates for all entries in cleanup list, then purge it
*/
void
execute_stonith_cleanup()
{
GListPtr iter;
for (iter = stonith_cleanup_list; iter != NULL; iter = iter->next) {
char *target = iter->data;
crm_node_t *target_node = crm_get_peer(0, target);
const char *uuid = crm_peer_uuid(target_node);
crm_notice("Marking %s, target of a previous stonith action, as clean", target);
send_stonith_update(NULL, target, uuid);
free(target);
}
g_list_free(stonith_cleanup_list);
stonith_cleanup_list = NULL;
}
/* end stonith cleanup list functions */
/* stonith API client
*
* Functions that need to interact directly with the fencer via its API
*/
static stonith_t *stonith_api = NULL;
static crm_trigger_t *stonith_reconnect = NULL;
static char *te_client_id = NULL;
static gboolean
fail_incompletable_stonith(crm_graph_t *graph)
{
GListPtr lpc = NULL;
const char *task = NULL;
xmlNode *last_action = NULL;
if (graph == NULL) {
return FALSE;
}
for (lpc = graph->synapses; lpc != NULL; lpc = lpc->next) {
GListPtr lpc2 = NULL;
synapse_t *synapse = (synapse_t *) lpc->data;
if (synapse->confirmed) {
continue;
}
for (lpc2 = synapse->actions; lpc2 != NULL; lpc2 = lpc2->next) {
crm_action_t *action = (crm_action_t *) lpc2->data;
if (action->type != action_type_crm || action->confirmed) {
continue;
}
task = crm_element_value(action->xml, XML_LRM_ATTR_TASK);
if (task && safe_str_eq(task, CRM_OP_FENCE)) {
action->failed = TRUE;
last_action = action->xml;
update_graph(graph, action);
crm_notice("Failing action %d (%s): fencer terminated",
action->id, ID(action->xml));
}
}
}
if (last_action != NULL) {
crm_warn("Fencer failure resulted in unrunnable actions");
abort_for_stonith_failure(tg_restart, NULL, last_action);
return TRUE;
}
return FALSE;
}
static void
tengine_stonith_connection_destroy(stonith_t *st, stonith_event_t *e)
{
te_cleanup_stonith_history_sync(st, FALSE);
if (is_set(fsa_input_register, R_ST_REQUIRED)) {
crm_crit("Fencing daemon connection failed");
mainloop_set_trigger(stonith_reconnect);
} else {
crm_info("Fencing daemon disconnected");
}
if (stonith_api) {
/* the client API won't properly reconnect notifications
* if they are still in the table - so remove them
*/
if (stonith_api->state != stonith_disconnected) {
stonith_api->cmds->disconnect(st);
}
stonith_api->cmds->remove_notification(stonith_api, T_STONITH_NOTIFY_DISCONNECT);
stonith_api->cmds->remove_notification(stonith_api, T_STONITH_NOTIFY_FENCE);
stonith_api->cmds->remove_notification(stonith_api, T_STONITH_NOTIFY_HISTORY_SYNCED);
}
if (AM_I_DC) {
fail_incompletable_stonith(transition_graph);
trigger_graph();
}
}
static void
tengine_stonith_notify(stonith_t *st, stonith_event_t *st_event)
{
if (te_client_id == NULL) {
te_client_id = crm_strdup_printf("%s.%lu", crm_system_name,
(unsigned long) getpid());
}
if (st_event == NULL) {
crm_err("Notify data not found");
return;
}
crmd_alert_fencing_op(st_event);
if ((st_event->result == pcmk_ok) && safe_str_eq("on", st_event->action)) {
crm_notice("%s was successfully unfenced by %s (at the request of %s)",
st_event->target,
st_event->executioner? st_event->executioner : "",
st_event->origin);
/* TODO: Hook up st_event->device */
return;
} else if (safe_str_eq("on", st_event->action)) {
crm_err("Unfencing of %s by %s failed: %s (%d)",
st_event->target,
st_event->executioner? st_event->executioner : "",
pcmk_strerror(st_event->result), st_event->result);
return;
} else if ((st_event->result == pcmk_ok)
&& crm_str_eq(st_event->target, fsa_our_uname, TRUE)) {
/* We were notified of our own fencing. Most likely, either fencing was
* misconfigured, or fabric fencing that doesn't cut cluster
* communication is in use.
*
* Either way, shutting down the local host is a good idea, to require
* administrator intervention. Also, other nodes would otherwise likely
* set our status to lost because of the fencing callback and discard
* our subsequent election votes as "not part of our cluster".
*/
crm_crit("We were allegedly just fenced by %s for %s!",
st_event->executioner? st_event->executioner : "the cluster",
st_event->origin); /* Dumps blackbox if enabled */
if (fence_reaction_panic) {
pcmk_panic(__FUNCTION__);
} else {
crm_exit(CRM_EX_FATAL);
}
return;
}
/* Update the count of stonith failures for this target, in case we become
* DC later. The current DC has already updated its fail count in
* tengine_stonith_callback().
*/
if (!AM_I_DC && safe_str_eq(st_event->operation, T_STONITH_NOTIFY_FENCE)) {
if (st_event->result == pcmk_ok) {
st_fail_count_reset(st_event->target);
} else {
st_fail_count_increment(st_event->target);
}
}
crm_notice("Peer %s was%s terminated (%s) by %s on behalf of %s: %s "
CRM_XS " initiator=%s ref=%s",
st_event->target, st_event->result == pcmk_ok ? "" : " not",
st_event->action,
st_event->executioner ? st_event->executioner : "",
(st_event->client_origin? st_event->client_origin : ""),
pcmk_strerror(st_event->result),
st_event->origin, st_event->id);
if (st_event->result == pcmk_ok) {
crm_node_t *peer = crm_find_known_peer_full(0, st_event->target, CRM_GET_PEER_ANY);
const char *uuid = NULL;
gboolean we_are_executioner = safe_str_eq(st_event->executioner, fsa_our_uname);
if (peer == NULL) {
return;
}
uuid = crm_peer_uuid(peer);
crm_trace("target=%s dc=%s", st_event->target, fsa_our_dc);
if(AM_I_DC) {
/* The DC always sends updates */
send_stonith_update(NULL, st_event->target, uuid);
/* @TODO Ideally, at this point, we'd check whether the fenced node
* hosted any guest nodes, and call remote_node_down() for them.
* Unfortunately, the controller doesn't have a simple, reliable way
* to map hosts to guests. It might be possible to track this in the
* peer cache via crm_remote_peer_cache_refresh(). For now, we rely
* on the scheduler creating fence pseudo-events for the guests.
*/
if (st_event->client_origin
&& safe_str_neq(st_event->client_origin, te_client_id)) {
/* Abort the current transition graph if it wasn't us
* that invoked stonith to fence someone
*/
crm_info("External fencing operation from %s fenced %s", st_event->client_origin, st_event->target);
abort_transition(INFINITY, tg_restart, "External Fencing Operation", NULL);
}
/* Assume it was our leader if we don't currently have one */
} else if (((fsa_our_dc == NULL) || safe_str_eq(fsa_our_dc, st_event->target))
&& is_not_set(peer->flags, crm_remote_node)) {
crm_notice("Target %s our leader %s (recorded: %s)",
fsa_our_dc ? "was" : "may have been", st_event->target,
fsa_our_dc ? fsa_our_dc : "");
/* Given the CIB resyncing that occurs around elections,
* have one node update the CIB now and, if the new DC is different,
* have them do so too after the election
*/
if (we_are_executioner) {
send_stonith_update(NULL, st_event->target, uuid);
}
add_stonith_cleanup(st_event->target);
}
/* If the target is a remote node, and we host its connection,
* immediately fail all monitors so it can be recovered quickly.
* The connection won't necessarily drop when a remote node is fenced,
* so the failure might not otherwise be detected until the next poke.
*/
if (is_set(peer->flags, crm_remote_node)) {
remote_ra_fail(st_event->target);
}
crmd_peer_down(peer, TRUE);
}
}
/*!
* \brief Connect to fencer
*
* \param[in] user_data If NULL, retry failures now, otherwise retry in main loop
*
* \return TRUE
* \note If user_data is NULL, this will wait 2s between attempts, for up to
* 30 attempts, meaning the controller could be blocked as long as 58s.
*/
static gboolean
te_connect_stonith(gpointer user_data)
{
int rc = pcmk_ok;
if (stonith_api == NULL) {
stonith_api = stonith_api_new();
if (stonith_api == NULL) {
crm_err("Could not connect to fencer: API memory allocation failed");
return TRUE;
}
}
if (stonith_api->state != stonith_disconnected) {
crm_trace("Already connected to fencer, no need to retry");
return TRUE;
}
if (user_data == NULL) {
// Blocking (retry failures now until successful)
rc = stonith_api_connect_retry(stonith_api, crm_system_name, 30);
if (rc != pcmk_ok) {
crm_err("Could not connect to fencer in 30 attempts: %s "
CRM_XS " rc=%d", pcmk_strerror(rc), rc);
}
} else {
// Non-blocking (retry failures later in main loop)
rc = stonith_api->cmds->connect(stonith_api, crm_system_name, NULL);
if (rc != pcmk_ok) {
if (is_set(fsa_input_register, R_ST_REQUIRED)) {
crm_err("Fencer connection failed (will retry): %s "
CRM_XS " rc=%d", pcmk_strerror(rc), rc);
mainloop_set_trigger(stonith_reconnect);
} else {
crm_info("Fencer connection failed (ignoring because no longer required): %s "
CRM_XS " rc=%d", pcmk_strerror(rc), rc);
}
return TRUE;
}
}
if (rc == pcmk_ok) {
stonith_api->cmds->register_notification(stonith_api,
T_STONITH_NOTIFY_DISCONNECT,
tengine_stonith_connection_destroy);
stonith_api->cmds->register_notification(stonith_api,
T_STONITH_NOTIFY_FENCE,
tengine_stonith_notify);
stonith_api->cmds->register_notification(stonith_api,
T_STONITH_NOTIFY_HISTORY_SYNCED,
tengine_stonith_history_synced);
te_trigger_stonith_history_sync(TRUE);
crm_notice("Fencer successfully connected");
}
return TRUE;
}
/*!
\internal
\brief Schedule fencer connection attempt in main loop
*/
void
controld_trigger_fencer_connect()
{
if (stonith_reconnect == NULL) {
stonith_reconnect = mainloop_add_trigger(G_PRIORITY_LOW,
te_connect_stonith,
GINT_TO_POINTER(TRUE));
}
set_bit(fsa_input_register, R_ST_REQUIRED);
mainloop_set_trigger(stonith_reconnect);
}
void
controld_disconnect_fencer(bool destroy)
{
if (stonith_api) {
// Prevent fencer connection from coming up again
clear_bit(fsa_input_register, R_ST_REQUIRED);
if (stonith_api->state != stonith_disconnected) {
stonith_api->cmds->disconnect(stonith_api);
}
stonith_api->cmds->remove_notification(stonith_api, T_STONITH_NOTIFY_DISCONNECT);
stonith_api->cmds->remove_notification(stonith_api, T_STONITH_NOTIFY_FENCE);
stonith_api->cmds->remove_notification(stonith_api, T_STONITH_NOTIFY_HISTORY_SYNCED);
}
if (destroy) {
if (stonith_api) {
stonith_api->cmds->free(stonith_api);
stonith_api = NULL;
}
if (stonith_reconnect) {
mainloop_destroy_trigger(stonith_reconnect);
stonith_reconnect = NULL;
}
if (te_client_id) {
free(te_client_id);
te_client_id = NULL;
}
}
}
static gboolean
do_stonith_history_sync(gpointer user_data)
{
if (stonith_api && (stonith_api->state != stonith_disconnected)) {
stonith_history_t *history = NULL;
te_cleanup_stonith_history_sync(stonith_api, FALSE);
stonith_api->cmds->history(stonith_api,
st_opt_sync_call | st_opt_broadcast,
NULL, &history, 5);
stonith_history_free(history);
return TRUE;
} else {
crm_info("Skip triggering stonith history-sync as stonith is disconnected");
return FALSE;
}
}
static void
tengine_stonith_callback(stonith_t *stonith, stonith_callback_data_t *data)
{
char *uuid = NULL;
int stonith_id = -1;
int transition_id = -1;
crm_action_t *action = NULL;
int call_id = data->call_id;
int rc = data->rc;
char *userdata = data->userdata;
CRM_CHECK(userdata != NULL, return);
crm_notice("Stonith operation %d/%s: %s (%d)", call_id, (char *)userdata,
pcmk_strerror(rc), rc);
if (AM_I_DC == FALSE) {
return;
}
/* crm_info("call=%d, optype=%d, node_name=%s, result=%d, node_list=%s, action=%s", */
/* op->call_id, op->optype, op->node_name, op->op_result, */
/* (char *)op->node_list, op->private_data); */
/* filter out old STONITH actions */
CRM_CHECK(decode_transition_key(userdata, &uuid, &transition_id, &stonith_id, NULL),
goto bail);
if (transition_graph->complete || stonith_id < 0 || safe_str_neq(uuid, te_uuid)
|| transition_graph->id != transition_id) {
crm_info("Ignoring STONITH action initiated outside of the current transition");
goto bail;
}
action = controld_get_action(stonith_id);
if (action == NULL) {
crm_err("Stonith action not matched");
goto bail;
}
stop_te_timer(action->timer);
if (rc == pcmk_ok) {
const char *target = crm_element_value(action->xml, XML_LRM_ATTR_TARGET);
const char *uuid = crm_element_value(action->xml, XML_LRM_ATTR_TARGET_UUID);
const char *op = crm_meta_value(action->params, "stonith_action");
crm_info("Stonith operation %d for %s passed", call_id, target);
if (action->confirmed == FALSE) {
te_action_confirmed(action, NULL);
if (safe_str_eq("on", op)) {
const char *value = NULL;
char *now = crm_ttoa(time(NULL));
update_attrd(target, CRM_ATTR_UNFENCED, now, NULL, FALSE);
free(now);
value = crm_meta_value(action->params, XML_OP_ATTR_DIGESTS_ALL);
update_attrd(target, CRM_ATTR_DIGESTS_ALL, value, NULL, FALSE);
value = crm_meta_value(action->params, XML_OP_ATTR_DIGESTS_SECURE);
update_attrd(target, CRM_ATTR_DIGESTS_SECURE, value, NULL, FALSE);
} else if (action->sent_update == FALSE) {
send_stonith_update(action, target, uuid);
action->sent_update = TRUE;
}
}
st_fail_count_reset(target);
} else {
const char *target = crm_element_value(action->xml, XML_LRM_ATTR_TARGET);
enum transition_action abort_action = tg_restart;
action->failed = TRUE;
crm_notice("Stonith operation %d for %s failed (%s): aborting transition.",
call_id, target, pcmk_strerror(rc));
/* If no fence devices were available, there's no use in immediately
* checking again, so don't start a new transition in that case.
*/
if (rc == -ENODEV) {
crm_warn("No devices found in cluster to fence %s, giving up",
target);
abort_action = tg_stop;
}
/* Increment the fail count now, so abort_for_stonith_failure() can
* check it. Non-DC nodes will increment it in tengine_stonith_notify().
*/
st_fail_count_increment(target);
abort_for_stonith_failure(abort_action, target, NULL);
}
update_graph(transition_graph, action);
trigger_graph();
bail:
free(userdata);
free(uuid);
return;
}
gboolean
te_fence_node(crm_graph_t *graph, crm_action_t *action)
{
int rc = 0;
const char *id = NULL;
const char *uuid = NULL;
const char *target = NULL;
const char *type = NULL;
char *transition_key = NULL;
+ const char *priority_delay = NULL;
gboolean invalid_action = FALSE;
enum stonith_call_options options = st_opt_none;
id = ID(action->xml);
target = crm_element_value(action->xml, XML_LRM_ATTR_TARGET);
uuid = crm_element_value(action->xml, XML_LRM_ATTR_TARGET_UUID);
type = crm_meta_value(action->params, "stonith_action");
CRM_CHECK(id != NULL, invalid_action = TRUE);
CRM_CHECK(uuid != NULL, invalid_action = TRUE);
CRM_CHECK(type != NULL, invalid_action = TRUE);
CRM_CHECK(target != NULL, invalid_action = TRUE);
if (invalid_action) {
crm_log_xml_warn(action->xml, "BadAction");
return FALSE;
}
+ priority_delay = crm_meta_value(action->params, XML_CONFIG_ATTR_PRIORITY_FENCING_DELAY);
+
crm_notice("Requesting fencing (%s) of node %s "
- CRM_XS " action=%s timeout=%u",
- type, target, id, transition_graph->stonith_timeout);
+ CRM_XS " action=%s timeout=%u priority_delay=%s",
+ type, target, id, transition_graph->stonith_timeout, priority_delay);
/* Passing NULL means block until we can connect... */
te_connect_stonith(NULL);
if (crmd_join_phase_count(crm_join_confirmed) == 1) {
options |= st_opt_allow_suicide;
}
- rc = stonith_api->cmds->fence(stonith_api, options, target, type,
- (int) (transition_graph->stonith_timeout / 1000),
- 0);
+ rc = stonith_api->cmds->fence_with_delay(stonith_api, options, target, type,
+ (int) (transition_graph->stonith_timeout / 1000),
+ 0, crm_atoi(priority_delay, "-1"));
transition_key = pcmk__transition_key(transition_graph->id, action->id, 0,
te_uuid),
stonith_api->cmds->register_callback(stonith_api, rc,
(int) (transition_graph->stonith_timeout / 1000),
st_opt_timeout_updates, transition_key,
"tengine_stonith_callback", tengine_stonith_callback);
return TRUE;
}
/* end stonith API client functions */
/*
* stonith history synchronization
*
* Each node's fencer keeps track of a cluster-wide fencing history. When a node
* joins or leaves, we need to synchronize the history across all nodes.
*/
static crm_trigger_t *stonith_history_sync_trigger = NULL;
static mainloop_timer_t *stonith_history_sync_timer_short = NULL;
static mainloop_timer_t *stonith_history_sync_timer_long = NULL;
void
te_cleanup_stonith_history_sync(stonith_t *st, bool free_timers)
{
if (free_timers) {
mainloop_timer_del(stonith_history_sync_timer_short);
stonith_history_sync_timer_short = NULL;
mainloop_timer_del(stonith_history_sync_timer_long);
stonith_history_sync_timer_long = NULL;
} else {
mainloop_timer_stop(stonith_history_sync_timer_short);
mainloop_timer_stop(stonith_history_sync_timer_long);
}
if (st) {
st->cmds->remove_notification(st, T_STONITH_NOTIFY_HISTORY_SYNCED);
}
}
static void
tengine_stonith_history_synced(stonith_t *st, stonith_event_t *st_event)
{
te_cleanup_stonith_history_sync(st, FALSE);
crm_debug("Fence-history synced - cancel all timers");
}
static gboolean
stonith_history_sync_set_trigger(gpointer user_data)
{
mainloop_set_trigger(stonith_history_sync_trigger);
return FALSE;
}
void
te_trigger_stonith_history_sync(bool long_timeout)
{
/* trigger a sync in 5s to give more nodes the
* chance to show up so that we don't create
* unnecessary stonith-history-sync traffic
*
* the long timeout of 30s is there as a fallback
* so that after a successful connection to fenced
* we will wait for 30s for the DC to trigger a
* history-sync
* if this doesn't happen we trigger a sync locally
* (e.g. fenced segfaults and is restarted by pacemakerd)
*/
/* as we are finally checking the stonith-connection
* in do_stonith_history_sync we should be fine
* leaving stonith_history_sync_time & stonith_history_sync_trigger
* around
*/
if (stonith_history_sync_trigger == NULL) {
stonith_history_sync_trigger =
mainloop_add_trigger(G_PRIORITY_LOW,
do_stonith_history_sync, NULL);
}
if (long_timeout) {
if(stonith_history_sync_timer_long == NULL) {
stonith_history_sync_timer_long =
mainloop_timer_add("history_sync_long", 30000,
FALSE, stonith_history_sync_set_trigger,
NULL);
}
crm_info("Fence history will be synchronized cluster-wide within 30 seconds");
mainloop_timer_start(stonith_history_sync_timer_long);
} else {
if(stonith_history_sync_timer_short == NULL) {
stonith_history_sync_timer_short =
mainloop_timer_add("history_sync_short", 5000,
FALSE, stonith_history_sync_set_trigger,
NULL);
}
crm_info("Fence history will be synchronized cluster-wide within 5 seconds");
mainloop_timer_start(stonith_history_sync_timer_short);
}
}
/* end stonith history synchronization functions */
diff --git a/daemons/fenced/fenced_commands.c b/daemons/fenced/fenced_commands.c
index d92dd9a1e0..aa224c5967 100644
--- a/daemons/fenced/fenced_commands.c
+++ b/daemons/fenced/fenced_commands.c
@@ -1,2732 +1,2743 @@
/*
* Copyright 2009-2020 the Pacemaker project contributors
*
* The version control history for this file may have further details.
*
* This source code is licensed under the GNU General Public License version 2
* or later (GPLv2+) WITHOUT ANY WARRANTY.
*/
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
GHashTable *device_list = NULL;
GHashTable *topology = NULL;
GList *cmd_list = NULL;
struct device_search_s {
/* target of fence action */
char *host;
/* requested fence action */
char *action;
/* timeout to use if a device is queried dynamically for possible targets */
int per_device_timeout;
/* number of registered fencing devices at time of request */
int replies_needed;
/* number of device replies received so far */
int replies_received;
/* whether the target is eligible to perform requested action (or off) */
bool allow_suicide;
/* private data to pass to search callback function */
void *user_data;
/* function to call when all replies have been received */
void (*callback) (GList * devices, void *user_data);
/* devices capable of performing requested action (or off if remapping) */
GListPtr capable;
};
static gboolean stonith_device_dispatch(gpointer user_data);
static void st_child_done(GPid pid, int rc, const char *output, gpointer user_data);
static void stonith_send_reply(xmlNode * reply, int call_options, const char *remote_peer,
const char *client_id);
static void search_devices_record_result(struct device_search_s *search, const char *device,
gboolean can_fence);
typedef struct async_command_s {
int id;
int pid;
int fd_stdout;
int options;
int default_timeout; /* seconds */
int timeout; /* seconds */
- int start_delay; /* milliseconds */
+ int start_delay; /* seconds */
int delay_id;
char *op;
char *origin;
char *client;
char *client_name;
char *remote_op_id;
char *victim;
uint32_t victim_nodeid;
char *action;
char *device;
char *mode;
GListPtr device_list;
GListPtr device_next;
void *internal_user_data;
void (*done_cb) (GPid pid, int rc, const char *output, gpointer user_data);
guint timer_sigterm;
guint timer_sigkill;
/*! If the operation timed out, this is the last signal
* we sent to the process to get it to terminate */
int last_timeout_signo;
stonith_device_t *active_on;
stonith_device_t *activating_on;
} async_command_t;
static xmlNode *stonith_construct_async_reply(async_command_t * cmd, const char *output,
xmlNode * data, int rc);
static gboolean
is_action_required(const char *action, stonith_device_t *device)
{
return device && device->automatic_unfencing && safe_str_eq(action, "on");
}
static int
get_action_delay_max(stonith_device_t * device, const char * action)
{
const char *value = NULL;
- int delay_max_ms = 0;
+ int delay_max = 0;
if (safe_str_neq(action, "off") && safe_str_neq(action, "reboot")) {
return 0;
}
value = g_hash_table_lookup(device->params, STONITH_ATTR_DELAY_MAX);
if (value) {
- delay_max_ms = crm_get_msec(value);
+ delay_max = crm_parse_interval_spec(value) / 1000;
}
- return delay_max_ms;
+ return delay_max;
}
static int
get_action_delay_base(stonith_device_t * device, const char * action)
{
const char *value = NULL;
- int delay_base_ms = 0;
+ int delay_base = 0;
if (safe_str_neq(action, "off") && safe_str_neq(action, "reboot")) {
return 0;
}
value = g_hash_table_lookup(device->params, STONITH_ATTR_DELAY_BASE);
if (value) {
- delay_base_ms = crm_get_msec(value);
+ delay_base = crm_parse_interval_spec(value) / 1000;
}
- return delay_base_ms;
+ return delay_base;
}
/*!
* \internal
* \brief Override STONITH timeout with pcmk_*_timeout if available
*
* \param[in] device STONITH device to use
* \param[in] action STONITH action name
* \param[in] default_timeout Timeout to use if device does not have
* a pcmk_*_timeout parameter for action
*
* \return Value of pcmk_(action)_timeout if available, otherwise default_timeout
* \note For consistency, it would be nice if reboot/off/on timeouts could be
* set the same way as start/stop/monitor timeouts, i.e. with an
* entry in the fencing resource configuration. However that
* is insufficient because fencing devices may be registered directly via
* the fencer's register_device() API instead of going through the CIB
* (e.g. stonith_admin uses it for its -R option, and the executor uses it
* to ensure a device is registered when a command is issued). As device
* properties, pcmk_*_timeout parameters can be grabbed by the fencer when
* the device is registered, whether by CIB change or API call.
*/
static int
get_action_timeout(stonith_device_t * device, const char *action, int default_timeout)
{
if (action && device && device->params) {
char buffer[64] = { 0, };
const char *value = NULL;
/* If "reboot" was requested but the device does not support it,
* we will remap to "off", so check timeout for "off" instead
*/
if (safe_str_eq(action, "reboot")
&& is_not_set(device->flags, st_device_supports_reboot)) {
crm_trace("%s doesn't support reboot, using timeout for off instead",
device->id);
action = "off";
}
/* If the device config specified an action-specific timeout, use it */
snprintf(buffer, sizeof(buffer), "pcmk_%s_timeout", action);
value = g_hash_table_lookup(device->params, buffer);
if (value) {
return atoi(value);
}
}
return default_timeout;
}
static void
free_async_command(async_command_t * cmd)
{
if (!cmd) {
return;
}
if (cmd->delay_id) {
g_source_remove(cmd->delay_id);
}
cmd_list = g_list_remove(cmd_list, cmd);
g_list_free_full(cmd->device_list, free);
free(cmd->device);
free(cmd->action);
free(cmd->victim);
free(cmd->remote_op_id);
free(cmd->client);
free(cmd->client_name);
free(cmd->origin);
free(cmd->mode);
free(cmd->op);
free(cmd);
}
static async_command_t *
create_async_command(xmlNode * msg)
{
async_command_t *cmd = NULL;
xmlNode *op = get_xpath_object("//@" F_STONITH_ACTION, msg, LOG_ERR);
const char *action = crm_element_value(op, F_STONITH_ACTION);
CRM_CHECK(action != NULL, crm_log_xml_warn(msg, "NoAction"); return NULL);
crm_log_xml_trace(msg, "Command");
cmd = calloc(1, sizeof(async_command_t));
crm_element_value_int(msg, F_STONITH_CALLID, &(cmd->id));
crm_element_value_int(msg, F_STONITH_CALLOPTS, &(cmd->options));
crm_element_value_int(msg, F_STONITH_TIMEOUT, &(cmd->default_timeout));
cmd->timeout = cmd->default_timeout;
+ // Default value -1 means no enforced fencing delay
+ cmd->start_delay = -1;
+ crm_element_value_int(msg, F_STONITH_DELAY, &(cmd->start_delay));
cmd->origin = crm_element_value_copy(msg, F_ORIG);
cmd->remote_op_id = crm_element_value_copy(msg, F_STONITH_REMOTE_OP_ID);
cmd->client = crm_element_value_copy(msg, F_STONITH_CLIENTID);
cmd->client_name = crm_element_value_copy(msg, F_STONITH_CLIENTNAME);
cmd->op = crm_element_value_copy(msg, F_STONITH_OPERATION);
cmd->action = strdup(action);
cmd->victim = crm_element_value_copy(op, F_STONITH_TARGET);
cmd->mode = crm_element_value_copy(op, F_STONITH_MODE);
cmd->device = crm_element_value_copy(op, F_STONITH_DEVICE);
CRM_CHECK(cmd->op != NULL, crm_log_xml_warn(msg, "NoOp"); free_async_command(cmd); return NULL);
CRM_CHECK(cmd->client != NULL, crm_log_xml_warn(msg, "NoClient"));
cmd->done_cb = st_child_done;
cmd_list = g_list_append(cmd_list, cmd);
return cmd;
}
static int
get_action_limit(stonith_device_t * device)
{
const char *value = NULL;
int action_limit = 1;
value = g_hash_table_lookup(device->params, STONITH_ATTR_ACTION_LIMIT);
if (value) {
action_limit = crm_parse_int(value, "1");
if (action_limit == 0) {
/* pcmk_action_limit should not be 0. Enforce it to be 1. */
action_limit = 1;
}
}
return action_limit;
}
static int
get_active_cmds(stonith_device_t * device)
{
int counter = 0;
GListPtr gIter = NULL;
GListPtr gIterNext = NULL;
CRM_CHECK(device != NULL, return 0);
for (gIter = cmd_list; gIter != NULL; gIter = gIterNext) {
async_command_t *cmd = gIter->data;
gIterNext = gIter->next;
if (cmd->active_on == device) {
counter++;
}
}
return counter;
}
static void
fork_cb(GPid pid, gpointer user_data)
{
async_command_t *cmd = (async_command_t *) user_data;
stonith_device_t * device =
/* in case of a retry we've done the move from
activating_on to active_on already
*/
cmd->activating_on?cmd->activating_on:cmd->active_on;
CRM_ASSERT(device);
crm_debug("Operation '%s'%s%s on %s now running with pid=%d, timeout=%ds",
cmd->action, cmd->victim ? " targeting " : "", cmd->victim ? cmd->victim : "",
device->id, pid, cmd->timeout);
cmd->active_on = device;
cmd->activating_on = NULL;
}
static gboolean
stonith_device_execute(stonith_device_t * device)
{
int exec_rc = 0;
const char *action_str = NULL;
async_command_t *cmd = NULL;
stonith_action_t *action = NULL;
int active_cmds = 0;
int action_limit = 0;
GListPtr gIter = NULL;
GListPtr gIterNext = NULL;
CRM_CHECK(device != NULL, return FALSE);
active_cmds = get_active_cmds(device);
action_limit = get_action_limit(device);
if (action_limit > -1 && active_cmds >= action_limit) {
crm_trace("%s is over its action limit of %d (%u active action%s)",
device->id, action_limit, active_cmds, active_cmds > 1 ? "s" : "");
return TRUE;
}
for (gIter = device->pending_ops; gIter != NULL; gIter = gIterNext) {
async_command_t *pending_op = gIter->data;
gIterNext = gIter->next;
if (pending_op && pending_op->delay_id) {
crm_trace
- ("Operation '%s'%s%s on %s was asked to run too early, waiting for start_delay timeout of %dms",
+ ("Operation '%s'%s%s on %s was asked to run too early, waiting for start_delay timeout of %ds",
pending_op->action, pending_op->victim ? " targeting " : "",
pending_op->victim ? pending_op->victim : "",
device->id, pending_op->start_delay);
continue;
}
device->pending_ops = g_list_remove_link(device->pending_ops, gIter);
g_list_free_1(gIter);
cmd = pending_op;
break;
}
if (cmd == NULL) {
crm_trace("Nothing further to do for %s for now", device->id);
return TRUE;
}
if(safe_str_eq(device->agent, STONITH_WATCHDOG_AGENT)) {
if(safe_str_eq(cmd->action, "reboot")) {
pcmk_panic(__FUNCTION__);
goto done;
} else if(safe_str_eq(cmd->action, "off")) {
pcmk_panic(__FUNCTION__);
goto done;
} else {
crm_info("Faking success for %s watchdog operation", cmd->action);
cmd->done_cb(0, 0, NULL, cmd);
goto done;
}
}
#if SUPPORT_CIBSECRETS
if (pcmk__substitute_secrets(device->id, device->params) != pcmk_rc_ok) {
/* replacing secrets failed! */
if (safe_str_eq(cmd->action,"stop")) {
/* don't fail on stop! */
crm_info("proceeding with the stop operation for %s", device->id);
} else {
crm_err("failed to get secrets for %s, "
"considering resource not configured", device->id);
exec_rc = PCMK_OCF_NOT_CONFIGURED;
cmd->done_cb(0, exec_rc, NULL, cmd);
goto done;
}
}
#endif
action_str = cmd->action;
if (safe_str_eq(cmd->action, "reboot") && is_not_set(device->flags, st_device_supports_reboot)) {
crm_warn("Agent '%s' does not advertise support for 'reboot', performing 'off' action instead", device->agent);
action_str = "off";
}
action = stonith_action_create(device->agent,
action_str,
cmd->victim,
cmd->victim_nodeid,
cmd->timeout, device->params, device->aliases);
/* for async exec, exec_rc is negative for early error exit
otherwise handling of success/errors is done via callbacks */
cmd->activating_on = device;
exec_rc = stonith_action_execute_async(action, (void *)cmd,
cmd->done_cb, fork_cb);
if (exec_rc < 0) {
crm_warn("Operation '%s'%s%s on %s failed: %s (%d)",
cmd->action, cmd->victim ? " targeting " : "", cmd->victim ? cmd->victim : "",
device->id, pcmk_strerror(exec_rc), exec_rc);
cmd->activating_on = NULL;
cmd->done_cb(0, exec_rc, NULL, cmd);
}
done:
/* Device might get triggered to work by multiple fencing commands
* simultaneously. Trigger the device again to make sure any
* remaining concurrent commands get executed. */
if (device->pending_ops) {
mainloop_set_trigger(device->work);
}
return TRUE;
}
static gboolean
stonith_device_dispatch(gpointer user_data)
{
return stonith_device_execute(user_data);
}
static gboolean
start_delay_helper(gpointer data)
{
async_command_t *cmd = data;
stonith_device_t *device = NULL;
cmd->delay_id = 0;
device = cmd->device ? g_hash_table_lookup(device_list, cmd->device) : NULL;
if (device) {
mainloop_set_trigger(device->work);
}
return FALSE;
}
static void
schedule_stonith_command(async_command_t * cmd, stonith_device_t * device)
{
int delay_max = 0;
int delay_base = 0;
+ bool delay_enforced = (cmd->start_delay >= 0);
CRM_CHECK(cmd != NULL, return);
CRM_CHECK(device != NULL, return);
if (cmd->device) {
free(cmd->device);
}
if (device->include_nodeid && cmd->victim) {
crm_node_t *node = crm_get_peer(0, cmd->victim);
cmd->victim_nodeid = node->id;
}
cmd->device = strdup(device->id);
cmd->timeout = get_action_timeout(device, cmd->action, cmd->default_timeout);
if (cmd->remote_op_id) {
crm_debug("Scheduling '%s' action%s%s on %s for remote peer %s with op id (%s) (timeout=%ds)",
cmd->action,
cmd->victim ? " targeting " : "", cmd->victim ? cmd->victim : "",
device->id, cmd->origin, cmd->remote_op_id, cmd->timeout);
} else {
crm_debug("Scheduling '%s' action%s%s on %s for %s (timeout=%ds)",
cmd->action,
cmd->victim ? " targeting " : "", cmd->victim ? cmd->victim : "",
device->id, cmd->client, cmd->timeout);
}
device->pending_ops = g_list_append(device->pending_ops, cmd);
mainloop_set_trigger(device->work);
- delay_max = get_action_delay_max(device, cmd->action);
- delay_base = get_action_delay_base(device, cmd->action);
- if (delay_max == 0) {
- delay_max = delay_base;
- }
- if (delay_max < delay_base) {
- crm_warn("Base-delay (%dms) is larger than max-delay (%dms) "
- "for %s on %s - limiting to max-delay",
- delay_base, delay_max, cmd->action, device->id);
- delay_base = delay_max;
+ // No enforced fencing delay
+ if (delay_enforced == FALSE) {
+ delay_max = get_action_delay_max(device, cmd->action);
+ delay_base = get_action_delay_base(device, cmd->action);
+ if (delay_max == 0) {
+ delay_max = delay_base;
+ }
+ if (delay_max < delay_base) {
+ crm_warn("Base-delay (%ds) is larger than max-delay (%ds) "
+ "for %s on %s - limiting to max-delay",
+ delay_base, delay_max, cmd->action, device->id);
+ delay_base = delay_max;
+ }
+ if (delay_max > 0) {
+ // coverity[dont_call] We're not using rand() for security
+ cmd->start_delay =
+ ((delay_max != delay_base)?(rand() % (delay_max - delay_base)):0)
+ + delay_base;
+ }
}
- if (delay_max > 0) {
- // coverity[dont_call] We're not using rand() for security
- cmd->start_delay =
- ((delay_max != delay_base)?(rand() % (delay_max - delay_base)):0)
- + delay_base;
- crm_notice("Delaying '%s' action%s%s on %s for %dms (timeout=%ds, base=%dms, "
- "max=%dms)",
- cmd->action,
- cmd->victim ? " targeting " : "", cmd->victim ? cmd->victim : "",
- device->id, cmd->start_delay, cmd->timeout,
- delay_base, delay_max);
+
+ if (cmd->start_delay > 0) {
+ crm_notice("Delaying '%s' action%s%s on %s for %s%ds (timeout=%ds, base=%ds, "
+ "max=%ds)",
+ cmd->action,
+ cmd->victim ? " targeting " : "", cmd->victim ? cmd->victim : "",
+ device->id, delay_enforced ? "enforced " : "",
+ cmd->start_delay, cmd->timeout,
+ delay_base, delay_max);
cmd->delay_id =
- g_timeout_add(cmd->start_delay, start_delay_helper, cmd);
+ g_timeout_add_seconds(cmd->start_delay, start_delay_helper, cmd);
}
}
static void
free_device(gpointer data)
{
GListPtr gIter = NULL;
stonith_device_t *device = data;
g_hash_table_destroy(device->params);
g_hash_table_destroy(device->aliases);
for (gIter = device->pending_ops; gIter != NULL; gIter = gIter->next) {
async_command_t *cmd = gIter->data;
crm_warn("Removal of device '%s' purged operation '%s'", device->id, cmd->action);
cmd->done_cb(0, -ENODEV, NULL, cmd);
}
g_list_free(device->pending_ops);
g_list_free_full(device->targets, free);
mainloop_destroy_trigger(device->work);
free_xml(device->agent_metadata);
free(device->namespace);
free(device->on_target_actions);
free(device->agent);
free(device->id);
free(device);
}
void free_device_list()
{
if (device_list != NULL) {
g_hash_table_destroy(device_list);
device_list = NULL;
}
}
void
init_device_list()
{
if (device_list == NULL) {
device_list = g_hash_table_new_full(crm_str_hash, g_str_equal, NULL,
free_device);
}
}
static GHashTable *
build_port_aliases(const char *hostmap, GListPtr * targets)
{
char *name = NULL;
int last = 0, lpc = 0, max = 0, added = 0;
GHashTable *aliases = crm_strcase_table_new();
if (hostmap == NULL) {
return aliases;
}
max = strlen(hostmap);
for (; lpc <= max; lpc++) {
switch (hostmap[lpc]) {
/* Assignment chars */
case '=':
case ':':
if (lpc > last) {
free(name);
name = calloc(1, 1 + lpc - last);
memcpy(name, hostmap + last, lpc - last);
}
last = lpc + 1;
break;
/* Delimeter chars */
/* case ',': Potentially used to specify multiple ports */
case 0:
case ';':
case ' ':
case '\t':
if (name) {
char *value = NULL;
value = calloc(1, 1 + lpc - last);
memcpy(value, hostmap + last, lpc - last);
crm_debug("Adding alias '%s'='%s'", name, value);
g_hash_table_replace(aliases, name, value);
if (targets) {
*targets = g_list_append(*targets, strdup(value));
}
value = NULL;
name = NULL;
added++;
} else if (lpc > last) {
crm_debug("Parse error at offset %d near '%s'", lpc - last, hostmap + last);
}
last = lpc + 1;
break;
}
if (hostmap[lpc] == 0) {
break;
}
}
if (added == 0) {
crm_info("No host mappings detected in '%s'", hostmap);
}
free(name);
return aliases;
}
GHashTable *metadata_cache = NULL;
void
free_metadata_cache() {
if (metadata_cache != NULL) {
g_hash_table_destroy(metadata_cache);
metadata_cache = NULL;
}
}
static void
init_metadata_cache() {
if (metadata_cache == NULL) {
metadata_cache = crm_str_table_new();
}
}
static xmlNode *
get_agent_metadata(const char *agent)
{
xmlNode *xml = NULL;
char *buffer = NULL;
init_metadata_cache();
buffer = g_hash_table_lookup(metadata_cache, agent);
if(safe_str_eq(agent, STONITH_WATCHDOG_AGENT)) {
return NULL;
} else if(buffer == NULL) {
stonith_t *st = stonith_api_new();
int rc;
if (st == NULL) {
crm_warn("Could not get agent meta-data: "
"API memory allocation failed");
return NULL;
}
rc = st->cmds->metadata(st, st_opt_sync_call, agent, NULL, &buffer, 10);
stonith_api_delete(st);
if (rc || !buffer) {
crm_err("Could not retrieve metadata for fencing agent %s", agent);
return NULL;
}
g_hash_table_replace(metadata_cache, strdup(agent), buffer);
}
xml = string2xml(buffer);
return xml;
}
static gboolean
is_nodeid_required(xmlNode * xml)
{
xmlXPathObjectPtr xpath = NULL;
if (stand_alone) {
return FALSE;
}
if (!xml) {
return FALSE;
}
xpath = xpath_search(xml, "//parameter[@name='nodeid']");
if (numXpathResults(xpath) <= 0) {
freeXpathObject(xpath);
return FALSE;
}
freeXpathObject(xpath);
return TRUE;
}
#define MAX_ACTION_LEN 256
static char *
add_action(char *actions, const char *action)
{
int offset = 0;
if (actions == NULL) {
actions = calloc(1, MAX_ACTION_LEN);
} else {
offset = strlen(actions);
}
if (offset > 0) {
offset += snprintf(actions+offset, MAX_ACTION_LEN - offset, " ");
}
offset += snprintf(actions+offset, MAX_ACTION_LEN - offset, "%s", action);
return actions;
}
static void
read_action_metadata(stonith_device_t *device)
{
xmlXPathObjectPtr xpath = NULL;
int max = 0;
int lpc = 0;
if (device->agent_metadata == NULL) {
return;
}
xpath = xpath_search(device->agent_metadata, "//action");
max = numXpathResults(xpath);
if (max <= 0) {
freeXpathObject(xpath);
return;
}
for (lpc = 0; lpc < max; lpc++) {
const char *on_target = NULL;
const char *action = NULL;
xmlNode *match = getXpathResult(xpath, lpc);
CRM_LOG_ASSERT(match != NULL);
if(match == NULL) { continue; };
on_target = crm_element_value(match, "on_target");
action = crm_element_value(match, "name");
if(safe_str_eq(action, "list")) {
set_bit(device->flags, st_device_supports_list);
} else if(safe_str_eq(action, "status")) {
set_bit(device->flags, st_device_supports_status);
} else if(safe_str_eq(action, "reboot")) {
set_bit(device->flags, st_device_supports_reboot);
} else if (safe_str_eq(action, "on")) {
/* "automatic" means the cluster will unfence node when it joins */
const char *automatic = crm_element_value(match, "automatic");
/* "required" is a deprecated synonym for "automatic" */
const char *required = crm_element_value(match, "required");
if (crm_is_true(automatic) || crm_is_true(required)) {
device->automatic_unfencing = TRUE;
}
}
if (action && crm_is_true(on_target)) {
device->on_target_actions = add_action(device->on_target_actions, action);
}
}
freeXpathObject(xpath);
}
/*!
* \internal
* \brief Set a pcmk_*_action parameter if not already set
*
* \param[in,out] params Device parameters
* \param[in] action Name of action
* \param[in] value Value to use if action is not already set
*/
static void
map_action(GHashTable *params, const char *action, const char *value)
{
char *key = crm_strdup_printf("pcmk_%s_action", action);
if (g_hash_table_lookup(params, key)) {
crm_warn("Ignoring %s='%s', see %s instead",
STONITH_ATTR_ACTION_OP, value, key);
free(key);
} else {
crm_warn("Mapping %s='%s' to %s='%s'",
STONITH_ATTR_ACTION_OP, value, key, value);
g_hash_table_insert(params, key, strdup(value));
}
}
/*!
* \internal
* \brief Create device parameter table from XML
*
* \param[in] name Device name (used for logging only)
* \param[in,out] params Device parameters
*/
static GHashTable *
xml2device_params(const char *name, xmlNode *dev)
{
GHashTable *params = xml2list(dev);
const char *value;
/* Action should never be specified in the device configuration,
* but we support it for users who are familiar with other software
* that worked that way.
*/
value = g_hash_table_lookup(params, STONITH_ATTR_ACTION_OP);
if (value != NULL) {
crm_warn("%s has '%s' parameter, which should never be specified in configuration",
name, STONITH_ATTR_ACTION_OP);
if (*value == '\0') {
crm_warn("Ignoring empty '%s' parameter", STONITH_ATTR_ACTION_OP);
} else if (strcmp(value, "reboot") == 0) {
crm_warn("Ignoring %s='reboot' (see stonith-action cluster property instead)",
STONITH_ATTR_ACTION_OP);
} else if (strcmp(value, "off") == 0) {
map_action(params, "reboot", value);
} else {
map_action(params, "off", value);
map_action(params, "reboot", value);
}
g_hash_table_remove(params, STONITH_ATTR_ACTION_OP);
}
return params;
}
static stonith_device_t *
build_device_from_xml(xmlNode * msg)
{
const char *value;
xmlNode *dev = get_xpath_object("//" F_STONITH_DEVICE, msg, LOG_ERR);
stonith_device_t *device = NULL;
char *agent = crm_element_value_copy(dev, "agent");
CRM_CHECK(agent != NULL, return device);
device = calloc(1, sizeof(stonith_device_t));
CRM_CHECK(device != NULL, {free(agent); return device;});
device->id = crm_element_value_copy(dev, XML_ATTR_ID);
device->agent = agent;
device->namespace = crm_element_value_copy(dev, "namespace");
device->params = xml2device_params(device->id, dev);
value = g_hash_table_lookup(device->params, STONITH_ATTR_HOSTLIST);
if (value) {
device->targets = stonith__parse_targets(value);
}
value = g_hash_table_lookup(device->params, STONITH_ATTR_HOSTMAP);
device->aliases = build_port_aliases(value, &(device->targets));
device->agent_metadata = get_agent_metadata(device->agent);
read_action_metadata(device);
value = g_hash_table_lookup(device->params, "nodeid");
if (!value) {
device->include_nodeid = is_nodeid_required(device->agent_metadata);
}
value = crm_element_value(dev, "rsc_provides");
if (safe_str_eq(value, "unfencing")) {
device->automatic_unfencing = TRUE;
}
if (is_action_required("on", device)) {
crm_info("The fencing device '%s' requires unfencing", device->id);
}
if (device->on_target_actions) {
crm_info("The fencing device '%s' requires actions (%s) to be executed on the target node",
device->id, device->on_target_actions);
}
device->work = mainloop_add_trigger(G_PRIORITY_HIGH, stonith_device_dispatch, device);
/* TODO: Hook up priority */
return device;
}
static const char *
target_list_type(stonith_device_t * dev)
{
const char *check_type = NULL;
check_type = g_hash_table_lookup(dev->params, STONITH_ATTR_HOSTCHECK);
if (check_type == NULL) {
if (g_hash_table_lookup(dev->params, STONITH_ATTR_HOSTLIST)) {
check_type = "static-list";
} else if (g_hash_table_lookup(dev->params, STONITH_ATTR_HOSTMAP)) {
check_type = "static-list";
} else if(is_set(dev->flags, st_device_supports_list)){
check_type = "dynamic-list";
} else if(is_set(dev->flags, st_device_supports_status)){
check_type = "status";
} else {
check_type = "none";
}
}
return check_type;
}
static void
schedule_internal_command(const char *origin,
stonith_device_t * device,
const char *action,
const char *victim,
int timeout,
void *internal_user_data,
void (*done_cb) (GPid pid, int rc, const char *output,
gpointer user_data))
{
async_command_t *cmd = NULL;
cmd = calloc(1, sizeof(async_command_t));
cmd->id = -1;
cmd->default_timeout = timeout ? timeout : 60;
cmd->timeout = cmd->default_timeout;
cmd->action = strdup(action);
cmd->victim = victim ? strdup(victim) : NULL;
cmd->device = strdup(device->id);
cmd->origin = strdup(origin);
cmd->client = strdup(crm_system_name);
cmd->client_name = strdup(crm_system_name);
cmd->internal_user_data = internal_user_data;
cmd->done_cb = done_cb; /* cmd, not internal_user_data, is passed to 'done_cb' as the userdata */
schedule_stonith_command(cmd, device);
}
gboolean
string_in_list(GListPtr list, const char *item)
{
int lpc = 0;
int max = g_list_length(list);
for (lpc = 0; lpc < max; lpc++) {
const char *value = g_list_nth_data(list, lpc);
if (safe_str_eq(item, value)) {
return TRUE;
} else {
crm_trace("%d: '%s' != '%s'", lpc, item, value);
}
}
return FALSE;
}
static void
status_search_cb(GPid pid, int rc, const char *output, gpointer user_data)
{
async_command_t *cmd = user_data;
struct device_search_s *search = cmd->internal_user_data;
stonith_device_t *dev = cmd->device ? g_hash_table_lookup(device_list, cmd->device) : NULL;
gboolean can = FALSE;
free_async_command(cmd);
if (!dev) {
search_devices_record_result(search, NULL, FALSE);
return;
}
mainloop_set_trigger(dev->work);
if (rc == 1 /* unknown */ ) {
crm_trace("Host %s is not known by %s", search->host, dev->id);
} else if (rc == 0 /* active */ || rc == 2 /* inactive */ ) {
crm_trace("Host %s is known by %s", search->host, dev->id);
can = TRUE;
} else {
crm_notice("Unknown result when testing if %s can fence %s: rc=%d", dev->id, search->host,
rc);
}
search_devices_record_result(search, dev->id, can);
}
static void
dynamic_list_search_cb(GPid pid, int rc, const char *output, gpointer user_data)
{
async_command_t *cmd = user_data;
struct device_search_s *search = cmd->internal_user_data;
stonith_device_t *dev = cmd->device ? g_hash_table_lookup(device_list, cmd->device) : NULL;
gboolean can_fence = FALSE;
free_async_command(cmd);
/* Host/alias must be in the list output to be eligible to be fenced
*
* Will cause problems if down'd nodes aren't listed or (for virtual nodes)
* if the guest is still listed despite being moved to another machine
*/
if (!dev) {
search_devices_record_result(search, NULL, FALSE);
return;
}
mainloop_set_trigger(dev->work);
/* If we successfully got the targets earlier, don't disable. */
if (rc != 0 && !dev->targets) {
crm_notice("Disabling port list queries for %s (%d): %s", dev->id, rc, output);
/* Fall back to status */
g_hash_table_replace(dev->params, strdup(STONITH_ATTR_HOSTCHECK), strdup("status"));
g_list_free_full(dev->targets, free);
dev->targets = NULL;
} else if (!rc) {
crm_info("Refreshing port list for %s", dev->id);
g_list_free_full(dev->targets, free);
dev->targets = stonith__parse_targets(output);
dev->targets_age = time(NULL);
}
if (dev->targets) {
const char *alias = g_hash_table_lookup(dev->aliases, search->host);
if (!alias) {
alias = search->host;
}
if (string_in_list(dev->targets, alias)) {
can_fence = TRUE;
}
}
search_devices_record_result(search, dev->id, can_fence);
}
/*!
* \internal
* \brief Returns true if any key in first is not in second or second has a different value for key
*/
static int
device_params_diff(GHashTable *first, GHashTable *second) {
char *key = NULL;
char *value = NULL;
GHashTableIter gIter;
g_hash_table_iter_init(&gIter, first);
while (g_hash_table_iter_next(&gIter, (void **)&key, (void **)&value)) {
if(strstr(key, "CRM_meta") == key) {
continue;
} else if(strcmp(key, "crm_feature_set") == 0) {
continue;
} else {
char *other_value = g_hash_table_lookup(second, key);
if (!other_value || safe_str_neq(other_value, value)) {
crm_trace("Different value for %s: %s != %s", key, other_value, value);
return 1;
}
}
}
return 0;
}
/*!
* \internal
* \brief Checks to see if an identical device already exists in the device_list
*/
static stonith_device_t *
device_has_duplicate(stonith_device_t * device)
{
stonith_device_t *dup = g_hash_table_lookup(device_list, device->id);
if (!dup) {
crm_trace("No match for %s", device->id);
return NULL;
} else if (safe_str_neq(dup->agent, device->agent)) {
crm_trace("Different agent: %s != %s", dup->agent, device->agent);
return NULL;
}
/* Use calculate_operation_digest() here? */
if (device_params_diff(device->params, dup->params) ||
device_params_diff(dup->params, device->params)) {
return NULL;
}
crm_trace("Match");
return dup;
}
int
stonith_device_register(xmlNode * msg, const char **desc, gboolean from_cib)
{
stonith_device_t *dup = NULL;
stonith_device_t *device = build_device_from_xml(msg);
CRM_CHECK(device != NULL, return -ENOMEM);
dup = device_has_duplicate(device);
if (dup) {
crm_debug("Device '%s' already existed in device list (%d active devices)", device->id,
g_hash_table_size(device_list));
free_device(device);
device = dup;
} else {
stonith_device_t *old = g_hash_table_lookup(device_list, device->id);
if (from_cib && old && old->api_registered) {
/* If the cib is writing over an entry that is shared with a stonith client,
* copy any pending ops that currently exist on the old entry to the new one.
* Otherwise the pending ops will be reported as failures
*/
crm_info("Overwriting an existing entry for %s from the cib", device->id);
device->pending_ops = old->pending_ops;
device->api_registered = TRUE;
old->pending_ops = NULL;
if (device->pending_ops) {
mainloop_set_trigger(device->work);
}
}
g_hash_table_replace(device_list, device->id, device);
crm_notice("Added '%s' to the device list (%d active devices)", device->id,
g_hash_table_size(device_list));
}
if (desc) {
*desc = device->id;
}
if (from_cib) {
device->cib_registered = TRUE;
} else {
device->api_registered = TRUE;
}
return pcmk_ok;
}
int
stonith_device_remove(const char *id, gboolean from_cib)
{
stonith_device_t *device = g_hash_table_lookup(device_list, id);
if (!device) {
crm_info("Device '%s' not found (%d active devices)", id, g_hash_table_size(device_list));
return pcmk_ok;
}
if (from_cib) {
device->cib_registered = FALSE;
} else {
device->verified = FALSE;
device->api_registered = FALSE;
}
if (!device->cib_registered && !device->api_registered) {
g_hash_table_remove(device_list, id);
crm_info("Removed '%s' from the device list (%d active devices)",
id, g_hash_table_size(device_list));
} else {
crm_trace("Not removing '%s' from the device list (%d active devices) "
"- still %s%s_registered", id, g_hash_table_size(device_list),
device->cib_registered?"cib":"", device->api_registered?"api":"");
}
return pcmk_ok;
}
/*!
* \internal
* \brief Return the number of stonith levels registered for a node
*
* \param[in] tp Node's topology table entry
*
* \return Number of non-NULL levels in topology entry
* \note This function is used only for log messages.
*/
static int
count_active_levels(stonith_topology_t * tp)
{
int lpc = 0;
int count = 0;
for (lpc = 0; lpc < ST_LEVEL_MAX; lpc++) {
if (tp->levels[lpc] != NULL) {
count++;
}
}
return count;
}
static void
free_topology_entry(gpointer data)
{
stonith_topology_t *tp = data;
int lpc = 0;
for (lpc = 0; lpc < ST_LEVEL_MAX; lpc++) {
if (tp->levels[lpc] != NULL) {
g_list_free_full(tp->levels[lpc], free);
}
}
free(tp->target);
free(tp->target_value);
free(tp->target_pattern);
free(tp->target_attribute);
free(tp);
}
void
free_topology_list()
{
if (topology != NULL) {
g_hash_table_destroy(topology);
topology = NULL;
}
}
void
init_topology_list()
{
if (topology == NULL) {
topology = g_hash_table_new_full(crm_str_hash, g_str_equal, NULL,
free_topology_entry);
}
}
char *stonith_level_key(xmlNode *level, int mode)
{
if(mode == -1) {
mode = stonith_level_kind(level);
}
switch(mode) {
case 0:
return crm_element_value_copy(level, XML_ATTR_STONITH_TARGET);
case 1:
return crm_element_value_copy(level, XML_ATTR_STONITH_TARGET_PATTERN);
case 2:
{
const char *name = crm_element_value(level, XML_ATTR_STONITH_TARGET_ATTRIBUTE);
const char *value = crm_element_value(level, XML_ATTR_STONITH_TARGET_VALUE);
if(name && value) {
return crm_strdup_printf("%s=%s", name, value);
}
}
default:
return crm_strdup_printf("Unknown-%d-%s", mode, ID(level));
}
}
int stonith_level_kind(xmlNode * level)
{
int mode = 0;
const char *target = crm_element_value(level, XML_ATTR_STONITH_TARGET);
if(target == NULL) {
mode++;
target = crm_element_value(level, XML_ATTR_STONITH_TARGET_PATTERN);
}
if(stand_alone == FALSE && target == NULL) {
mode++;
if(crm_element_value(level, XML_ATTR_STONITH_TARGET_ATTRIBUTE) == NULL) {
mode++;
} else if(crm_element_value(level, XML_ATTR_STONITH_TARGET_VALUE) == NULL) {
mode++;
}
}
return mode;
}
static stonith_key_value_t *
parse_device_list(const char *devices)
{
int lpc = 0;
int max = 0;
int last = 0;
stonith_key_value_t *output = NULL;
if (devices == NULL) {
return output;
}
max = strlen(devices);
for (lpc = 0; lpc <= max; lpc++) {
if (devices[lpc] == ',' || devices[lpc] == 0) {
char *line = strndup(devices + last, lpc - last);
output = stonith_key_value_add(output, NULL, line);
free(line);
last = lpc + 1;
}
}
return output;
}
/*!
* \internal
* \brief Register a STONITH level for a target
*
* Given an XML request specifying the target name, level index, and device IDs
* for the level, this will create an entry for the target in the global topology
* table if one does not already exist, then append the specified device IDs to
* the entry's device list for the specified level.
*
* \param[in] msg XML request for STONITH level registration
* \param[out] desc If not NULL, will be set to string representation ("TARGET[LEVEL]")
*
* \return pcmk_ok on success, -EINVAL if XML does not specify valid level index
*/
int
stonith_level_register(xmlNode *msg, char **desc)
{
int id = 0;
xmlNode *level;
int mode;
char *target;
stonith_topology_t *tp;
stonith_key_value_t *dIter = NULL;
stonith_key_value_t *devices = NULL;
/* Allow the XML here to point to the level tag directly, or wrapped in
* another tag. If directly, don't search by xpath, because it might give
* multiple hits (e.g. if the XML is the CIB).
*/
if (safe_str_eq(TYPE(msg), XML_TAG_FENCING_LEVEL)) {
level = msg;
} else {
level = get_xpath_object("//" XML_TAG_FENCING_LEVEL, msg, LOG_ERR);
}
CRM_CHECK(level != NULL, return -EINVAL);
mode = stonith_level_kind(level);
target = stonith_level_key(level, mode);
crm_element_value_int(level, XML_ATTR_STONITH_INDEX, &id);
if (desc) {
*desc = crm_strdup_printf("%s[%d]", target, id);
}
/* Sanity-check arguments */
if (mode >= 3 || (id <= 0) || (id >= ST_LEVEL_MAX)) {
crm_trace("Could not add %s[%d] (%d) to the topology (%d active entries)", target, id, mode, g_hash_table_size(topology));
free(target);
crm_log_xml_err(level, "Bad topology");
return -EINVAL;
}
/* Find or create topology table entry */
tp = g_hash_table_lookup(topology, target);
if (tp == NULL) {
tp = calloc(1, sizeof(stonith_topology_t));
tp->kind = mode;
tp->target = target;
tp->target_value = crm_element_value_copy(level, XML_ATTR_STONITH_TARGET_VALUE);
tp->target_pattern = crm_element_value_copy(level, XML_ATTR_STONITH_TARGET_PATTERN);
tp->target_attribute = crm_element_value_copy(level, XML_ATTR_STONITH_TARGET_ATTRIBUTE);
g_hash_table_replace(topology, tp->target, tp);
crm_trace("Added %s (%d) to the topology (%d active entries)",
target, mode, g_hash_table_size(topology));
} else {
free(target);
}
if (tp->levels[id] != NULL) {
crm_info("Adding to the existing %s[%d] topology entry",
tp->target, id);
}
devices = parse_device_list(crm_element_value(level, XML_ATTR_STONITH_DEVICES));
for (dIter = devices; dIter; dIter = dIter->next) {
const char *device = dIter->value;
crm_trace("Adding device '%s' for %s[%d]", device, tp->target, id);
tp->levels[id] = g_list_append(tp->levels[id], strdup(device));
}
stonith_key_value_freeall(devices, 1, 1);
crm_info("Target %s has %d active fencing levels",
tp->target, count_active_levels(tp));
return pcmk_ok;
}
int
stonith_level_remove(xmlNode *msg, char **desc)
{
int id = 0;
stonith_topology_t *tp;
char *target;
/* Unlike additions, removal requests should always have one level tag */
xmlNode *level = get_xpath_object("//" XML_TAG_FENCING_LEVEL, msg, LOG_ERR);
CRM_CHECK(level != NULL, return -EINVAL);
target = stonith_level_key(level, -1);
crm_element_value_int(level, XML_ATTR_STONITH_INDEX, &id);
if (desc) {
*desc = crm_strdup_printf("%s[%d]", target, id);
}
/* Sanity-check arguments */
if (id >= ST_LEVEL_MAX) {
free(target);
return -EINVAL;
}
tp = g_hash_table_lookup(topology, target);
if (tp == NULL) {
crm_info("Topology for %s not found (%d active entries)",
target, g_hash_table_size(topology));
} else if (id == 0 && g_hash_table_remove(topology, target)) {
crm_info("Removed all %s related entries from the topology (%d active entries)",
target, g_hash_table_size(topology));
} else if (id > 0 && tp->levels[id] != NULL) {
g_list_free_full(tp->levels[id], free);
tp->levels[id] = NULL;
crm_info("Removed level '%d' from topology for %s (%d active levels remaining)",
id, target, count_active_levels(tp));
}
free(target);
return pcmk_ok;
}
/*!
* \internal
* \brief Schedule an (asynchronous) action directly on a stonith device
*
* Handle a STONITH_OP_EXEC API message by scheduling a requested agent action
* directly on a specified device. Only list, monitor, and status actions are
* expected to use this call, though it should work with any agent command.
*
* \param[in] msg API message XML with desired action
* \param[out] output Unused
*
* \return -EINPROGRESS on success, -errno otherwise
* \note If the action is monitor, the device must be registered via the API
* (CIB registration is not sufficient), because monitor should not be
* possible unless the device is "started" (API registered).
*/
static int
stonith_device_action(xmlNode * msg, char **output)
{
xmlNode *dev = get_xpath_object("//" F_STONITH_DEVICE, msg, LOG_ERR);
xmlNode *op = get_xpath_object("//@" F_STONITH_ACTION, msg, LOG_ERR);
const char *id = crm_element_value(dev, F_STONITH_DEVICE);
const char *action = crm_element_value(op, F_STONITH_ACTION);
async_command_t *cmd = NULL;
stonith_device_t *device = NULL;
if ((id == NULL) || (action == NULL)) {
crm_info("Malformed API action request: device %s, action %s",
(id? id : "not specified"),
(action? action : "not specified"));
return -EPROTO;
}
device = g_hash_table_lookup(device_list, id);
if ((device == NULL)
|| (!device->api_registered && !strcmp(action, "monitor"))) {
// Monitors may run only on "started" (API-registered) devices
crm_info("Ignoring API '%s' action request because device %s not found",
action, id);
return -ENODEV;
}
cmd = create_async_command(msg);
if (cmd == NULL) {
return -EPROTO;
}
schedule_stonith_command(cmd, device);
return -EINPROGRESS;
}
static void
search_devices_record_result(struct device_search_s *search, const char *device, gboolean can_fence)
{
search->replies_received++;
if (can_fence && device) {
search->capable = g_list_append(search->capable, strdup(device));
}
if (search->replies_needed == search->replies_received) {
crm_debug("Finished Search. %d devices can perform action (%s) on node %s",
g_list_length(search->capable),
search->action ? search->action : "",
search->host ? search->host : "");
search->callback(search->capable, search->user_data);
free(search->host);
free(search->action);
free(search);
}
}
/*!
* \internal
* \brief Check whether the local host is allowed to execute a fencing action
*
* \param[in] device Fence device to check
* \param[in] action Fence action to check
* \param[in] target Hostname of fence target
* \param[in] allow_suicide Whether self-fencing is allowed for this operation
*
* \return TRUE if local host is allowed to execute action, FALSE otherwise
*/
static gboolean
localhost_is_eligible(const stonith_device_t *device, const char *action,
const char *target, gboolean allow_suicide)
{
gboolean localhost_is_target = safe_str_eq(target, stonith_our_uname);
if (device && action && device->on_target_actions
&& strstr(device->on_target_actions, action)) {
if (!localhost_is_target) {
crm_trace("'%s' operation with %s can only be executed for localhost not %s",
action, device->id, target);
return FALSE;
}
} else if (localhost_is_target && !allow_suicide) {
crm_trace("'%s' operation does not support self-fencing", action);
return FALSE;
}
return TRUE;
}
static void
can_fence_host_with_device(stonith_device_t * dev, struct device_search_s *search)
{
gboolean can = FALSE;
const char *check_type = NULL;
const char *host = search->host;
const char *alias = NULL;
CRM_LOG_ASSERT(dev != NULL);
if (dev == NULL) {
goto search_report_results;
} else if (host == NULL) {
can = TRUE;
goto search_report_results;
}
/* Short-circuit query if this host is not allowed to perform the action */
if (safe_str_eq(search->action, "reboot")) {
/* A "reboot" *might* get remapped to "off" then "on", so short-circuit
* only if all three are disallowed. If only one or two are disallowed,
* we'll report that with the results. We never allow suicide for
* remapped "on" operations because the host is off at that point.
*/
if (!localhost_is_eligible(dev, "reboot", host, search->allow_suicide)
&& !localhost_is_eligible(dev, "off", host, search->allow_suicide)
&& !localhost_is_eligible(dev, "on", host, FALSE)) {
goto search_report_results;
}
} else if (!localhost_is_eligible(dev, search->action, host,
search->allow_suicide)) {
goto search_report_results;
}
alias = g_hash_table_lookup(dev->aliases, host);
if (alias == NULL) {
alias = host;
}
check_type = target_list_type(dev);
if (safe_str_eq(check_type, "none")) {
can = TRUE;
} else if (safe_str_eq(check_type, "static-list")) {
/* Presence in the hostmap is sufficient
* Only use if all hosts on which the device can be active can always fence all listed hosts
*/
if (string_in_list(dev->targets, host)) {
can = TRUE;
} else if (g_hash_table_lookup(dev->params, STONITH_ATTR_HOSTMAP)
&& g_hash_table_lookup(dev->aliases, host)) {
can = TRUE;
}
} else if (safe_str_eq(check_type, "dynamic-list")) {
time_t now = time(NULL);
if (dev->targets == NULL || dev->targets_age + 60 < now) {
crm_trace("Running '%s' to check whether %s is eligible to fence %s (%s)",
check_type, dev->id, search->host, search->action);
schedule_internal_command(__FUNCTION__, dev, "list", NULL,
search->per_device_timeout, search, dynamic_list_search_cb);
/* we'll respond to this search request async in the cb */
return;
}
if (string_in_list(dev->targets, alias)) {
can = TRUE;
}
} else if (safe_str_eq(check_type, "status")) {
crm_trace("Running '%s' to check whether %s is eligible to fence %s (%s)",
check_type, dev->id, search->host, search->action);
schedule_internal_command(__FUNCTION__, dev, "status", search->host,
search->per_device_timeout, search, status_search_cb);
/* we'll respond to this search request async in the cb */
return;
} else {
crm_err("Invalid value for " STONITH_ATTR_HOSTCHECK ": %s", check_type);
check_type = "Invalid " STONITH_ATTR_HOSTCHECK;
}
if (safe_str_eq(host, alias)) {
crm_notice("%s is%s eligible to fence (%s) %s: %s",
dev->id, (can? "" : " not"), search->action, host,
check_type);
} else {
crm_notice("%s is%s eligible to fence (%s) %s (aka. '%s'): %s",
dev->id, (can? "" : " not"), search->action, host, alias,
check_type);
}
search_report_results:
search_devices_record_result(search, dev ? dev->id : NULL, can);
}
static void
search_devices(gpointer key, gpointer value, gpointer user_data)
{
stonith_device_t *dev = value;
struct device_search_s *search = user_data;
can_fence_host_with_device(dev, search);
}
#define DEFAULT_QUERY_TIMEOUT 20
static void
get_capable_devices(const char *host, const char *action, int timeout, bool suicide, void *user_data,
void (*callback) (GList * devices, void *user_data))
{
struct device_search_s *search;
int per_device_timeout = DEFAULT_QUERY_TIMEOUT;
int devices_needing_async_query = 0;
char *key = NULL;
const char *check_type = NULL;
GHashTableIter gIter;
stonith_device_t *device = NULL;
if (!g_hash_table_size(device_list)) {
callback(NULL, user_data);
return;
}
search = calloc(1, sizeof(struct device_search_s));
if (!search) {
callback(NULL, user_data);
return;
}
g_hash_table_iter_init(&gIter, device_list);
while (g_hash_table_iter_next(&gIter, (void **)&key, (void **)&device)) {
check_type = target_list_type(device);
if (safe_str_eq(check_type, "status") || safe_str_eq(check_type, "dynamic-list")) {
devices_needing_async_query++;
}
}
/* If we have devices that require an async event in order to know what
* nodes they can fence, we have to give the events a timeout. The total
* query timeout is divided among those events. */
if (devices_needing_async_query) {
per_device_timeout = timeout / devices_needing_async_query;
if (!per_device_timeout) {
crm_err("STONITH timeout %ds is too low; using %ds, but consider raising to at least %ds",
timeout, DEFAULT_QUERY_TIMEOUT,
DEFAULT_QUERY_TIMEOUT * devices_needing_async_query);
per_device_timeout = DEFAULT_QUERY_TIMEOUT;
} else if (per_device_timeout < DEFAULT_QUERY_TIMEOUT) {
crm_notice("STONITH timeout %ds is low for the current configuration;"
" consider raising to at least %ds",
timeout, DEFAULT_QUERY_TIMEOUT * devices_needing_async_query);
}
}
search->host = host ? strdup(host) : NULL;
search->action = action ? strdup(action) : NULL;
search->per_device_timeout = per_device_timeout;
/* We are guaranteed this many replies. Even if a device gets
* unregistered some how during the async search, we will get
* the correct number of replies. */
search->replies_needed = g_hash_table_size(device_list);
search->allow_suicide = suicide;
search->callback = callback;
search->user_data = user_data;
/* kick off the search */
crm_debug("Searching through %d devices to see what is capable of action (%s) for target %s",
search->replies_needed,
search->action ? search->action : "",
search->host ? search->host : "");
g_hash_table_foreach(device_list, search_devices, search);
}
struct st_query_data {
xmlNode *reply;
char *remote_peer;
char *client_id;
char *target;
char *action;
int call_options;
};
/*!
* \internal
* \brief Add action-specific attributes to query reply XML
*
* \param[in,out] xml XML to add attributes to
* \param[in] action Fence action
* \param[in] device Fence device
*/
static void
add_action_specific_attributes(xmlNode *xml, const char *action,
stonith_device_t *device)
{
int action_specific_timeout;
int delay_max;
int delay_base;
CRM_CHECK(xml && action && device, return);
if (is_action_required(action, device)) {
crm_trace("Action '%s' is required on %s", action, device->id);
crm_xml_add_int(xml, F_STONITH_DEVICE_REQUIRED, 1);
}
action_specific_timeout = get_action_timeout(device, action, 0);
if (action_specific_timeout) {
crm_trace("Action '%s' has timeout %dms on %s",
action, action_specific_timeout, device->id);
crm_xml_add_int(xml, F_STONITH_ACTION_TIMEOUT, action_specific_timeout);
}
delay_max = get_action_delay_max(device, action);
if (delay_max > 0) {
crm_trace("Action '%s' has maximum random delay %dms on %s",
action, delay_max, device->id);
crm_xml_add_int(xml, F_STONITH_DELAY_MAX, delay_max / 1000);
}
delay_base = get_action_delay_base(device, action);
if (delay_base > 0) {
crm_xml_add_int(xml, F_STONITH_DELAY_BASE, delay_base / 1000);
}
if ((delay_max > 0) && (delay_base == 0)) {
crm_trace("Action '%s' has maximum random delay %dms on %s",
action, delay_max, device->id);
} else if ((delay_max == 0) && (delay_base > 0)) {
crm_trace("Action '%s' has a static delay of %dms on %s",
action, delay_base, device->id);
} else if ((delay_max > 0) && (delay_base > 0)) {
crm_trace("Action '%s' has a minimum delay of %dms and a randomly chosen "
"maximum delay of %dms on %s",
action, delay_base, delay_max, device->id);
}
}
/*!
* \internal
* \brief Add "disallowed" attribute to query reply XML if appropriate
*
* \param[in,out] xml XML to add attribute to
* \param[in] action Fence action
* \param[in] device Fence device
* \param[in] target Fence target
* \param[in] allow_suicide Whether self-fencing is allowed
*/
static void
add_disallowed(xmlNode *xml, const char *action, stonith_device_t *device,
const char *target, gboolean allow_suicide)
{
if (!localhost_is_eligible(device, action, target, allow_suicide)) {
crm_trace("Action '%s' on %s is disallowed for local host",
action, device->id);
crm_xml_add(xml, F_STONITH_ACTION_DISALLOWED, XML_BOOLEAN_TRUE);
}
}
/*!
* \internal
* \brief Add child element with action-specific values to query reply XML
*
* \param[in,out] xml XML to add attribute to
* \param[in] action Fence action
* \param[in] device Fence device
* \param[in] target Fence target
* \param[in] allow_suicide Whether self-fencing is allowed
*/
static void
add_action_reply(xmlNode *xml, const char *action, stonith_device_t *device,
const char *target, gboolean allow_suicide)
{
xmlNode *child = create_xml_node(xml, F_STONITH_ACTION);
crm_xml_add(child, XML_ATTR_ID, action);
add_action_specific_attributes(child, action, device);
add_disallowed(child, action, device, target, allow_suicide);
}
static void
stonith_query_capable_device_cb(GList * devices, void *user_data)
{
struct st_query_data *query = user_data;
int available_devices = 0;
xmlNode *dev = NULL;
xmlNode *list = NULL;
GListPtr lpc = NULL;
/* Pack the results into XML */
list = create_xml_node(NULL, __FUNCTION__);
crm_xml_add(list, F_STONITH_TARGET, query->target);
for (lpc = devices; lpc != NULL; lpc = lpc->next) {
stonith_device_t *device = g_hash_table_lookup(device_list, lpc->data);
const char *action = query->action;
if (!device) {
/* It is possible the device got unregistered while
* determining who can fence the target */
continue;
}
available_devices++;
dev = create_xml_node(list, F_STONITH_DEVICE);
crm_xml_add(dev, XML_ATTR_ID, device->id);
crm_xml_add(dev, "namespace", device->namespace);
crm_xml_add(dev, "agent", device->agent);
crm_xml_add_int(dev, F_STONITH_DEVICE_VERIFIED, device->verified);
/* If the originating fencer wants to reboot the node, and we have a
* capable device that doesn't support "reboot", remap to "off" instead.
*/
if (is_not_set(device->flags, st_device_supports_reboot)
&& safe_str_eq(query->action, "reboot")) {
crm_trace("%s doesn't support reboot, using values for off instead",
device->id);
action = "off";
}
/* Add action-specific values if available */
add_action_specific_attributes(dev, action, device);
if (safe_str_eq(query->action, "reboot")) {
/* A "reboot" *might* get remapped to "off" then "on", so after
* sending the "reboot"-specific values in the main element, we add
* sub-elements for "off" and "on" values.
*
* We short-circuited earlier if "reboot", "off" and "on" are all
* disallowed for the local host. However if only one or two are
* disallowed, we send back the results and mark which ones are
* disallowed. If "reboot" is disallowed, this might cause problems
* with older fencer versions, which won't check for it. Older
* versions will ignore "off" and "on", so they are not a problem.
*/
add_disallowed(dev, action, device, query->target,
is_set(query->call_options, st_opt_allow_suicide));
add_action_reply(dev, "off", device, query->target,
is_set(query->call_options, st_opt_allow_suicide));
add_action_reply(dev, "on", device, query->target, FALSE);
}
/* A query without a target wants device parameters */
if (query->target == NULL) {
xmlNode *attrs = create_xml_node(dev, XML_TAG_ATTRS);
g_hash_table_foreach(device->params, hash2field, attrs);
}
}
crm_xml_add_int(list, F_STONITH_AVAILABLE_DEVICES, available_devices);
if (query->target) {
crm_debug("Found %d matching devices for '%s'", available_devices, query->target);
} else {
crm_debug("%d devices installed", available_devices);
}
if (list != NULL) {
crm_log_xml_trace(list, "Add query results");
add_message_xml(query->reply, F_STONITH_CALLDATA, list);
}
stonith_send_reply(query->reply, query->call_options, query->remote_peer, query->client_id);
free_xml(query->reply);
free(query->remote_peer);
free(query->client_id);
free(query->target);
free(query->action);
free(query);
free_xml(list);
g_list_free_full(devices, free);
}
static void
stonith_query(xmlNode * msg, const char *remote_peer, const char *client_id, int call_options)
{
struct st_query_data *query = NULL;
const char *action = NULL;
const char *target = NULL;
int timeout = 0;
xmlNode *dev = get_xpath_object("//@" F_STONITH_ACTION, msg, LOG_NEVER);
crm_element_value_int(msg, F_STONITH_TIMEOUT, &timeout);
if (dev) {
const char *device = crm_element_value(dev, F_STONITH_DEVICE);
target = crm_element_value(dev, F_STONITH_TARGET);
action = crm_element_value(dev, F_STONITH_ACTION);
if (device && safe_str_eq(device, "manual_ack")) {
/* No query or reply necessary */
return;
}
}
crm_log_xml_debug(msg, "Query");
query = calloc(1, sizeof(struct st_query_data));
query->reply = stonith_construct_reply(msg, NULL, NULL, pcmk_ok);
query->remote_peer = remote_peer ? strdup(remote_peer) : NULL;
query->client_id = client_id ? strdup(client_id) : NULL;
query->target = target ? strdup(target) : NULL;
query->action = action ? strdup(action) : NULL;
query->call_options = call_options;
get_capable_devices(target, action, timeout,
is_set(call_options, st_opt_allow_suicide),
query, stonith_query_capable_device_cb);
}
#define ST_LOG_OUTPUT_MAX 512
static void
log_operation(async_command_t * cmd, int rc, int pid, const char *next, const char *output, gboolean op_merged)
{
if (rc == 0) {
next = NULL;
}
if (cmd->victim != NULL) {
do_crm_log(rc == 0 ? LOG_NOTICE : LOG_ERR,
"Operation '%s' [%d] (call %d from %s) for host '%s' with device '%s' returned%s: %d (%s)%s%s",
cmd->action, pid, cmd->id, cmd->client_name, cmd->victim,
cmd->device, (op_merged? " (merged)" : ""),
rc, pcmk_strerror(rc),
(next? ", retrying with " : ""), (next ? next : ""));
} else {
do_crm_log_unlikely(rc == 0 ? LOG_DEBUG : LOG_NOTICE,
"Operation '%s' [%d] for device '%s' returned%s: %d (%s)%s%s",
cmd->action, pid, cmd->device,
(op_merged? " (merged)" : ""),
rc, pcmk_strerror(rc),
(next? ", retrying with " : ""), (next ? next : ""));
}
if (output) {
/* Logging the whole string confuses syslog when the string is xml */
char *prefix = crm_strdup_printf("%s:%d", cmd->device, pid);
crm_log_output(rc == 0 ? LOG_DEBUG : LOG_WARNING, prefix, output);
free(prefix);
}
}
static void
stonith_send_async_reply(async_command_t * cmd, const char *output, int rc, GPid pid, int options)
{
xmlNode *reply = NULL;
gboolean bcast = FALSE;
reply = stonith_construct_async_reply(cmd, output, NULL, rc);
if (safe_str_eq(cmd->action, "metadata")) {
/* Too verbose to log */
crm_trace("Metadata query for %s", cmd->device);
output = NULL;
} else if (crm_str_eq(cmd->action, "monitor", TRUE) ||
crm_str_eq(cmd->action, "list", TRUE) || crm_str_eq(cmd->action, "status", TRUE)) {
crm_trace("Never broadcast '%s' replies", cmd->action);
} else if (!stand_alone && safe_str_eq(cmd->origin, cmd->victim) && safe_str_neq(cmd->action, "on")) {
crm_trace("Broadcast '%s' reply for %s", cmd->action, cmd->victim);
crm_xml_add(reply, F_SUBTYPE, "broadcast");
bcast = TRUE;
}
log_operation(cmd, rc, pid, NULL, output, (options & st_reply_opt_merged ? TRUE : FALSE));
crm_log_xml_trace(reply, "Reply");
if (options & st_reply_opt_merged) {
crm_xml_add(reply, F_STONITH_MERGED, "true");
}
if (bcast) {
crm_xml_add(reply, F_STONITH_OPERATION, T_STONITH_NOTIFY);
send_cluster_message(NULL, crm_msg_stonith_ng, reply, FALSE);
} else if (cmd->origin) {
crm_trace("Directed reply to %s", cmd->origin);
send_cluster_message(crm_get_peer(0, cmd->origin), crm_msg_stonith_ng, reply, FALSE);
} else {
crm_trace("Directed local %ssync reply to %s",
(cmd->options & st_opt_sync_call) ? "" : "a-", cmd->client_name);
do_local_reply(reply, cmd->client, cmd->options & st_opt_sync_call, FALSE);
}
if (stand_alone) {
/* Do notification with a clean data object */
xmlNode *notify_data = create_xml_node(NULL, T_STONITH_NOTIFY_FENCE);
crm_xml_add_int(notify_data, F_STONITH_RC, rc);
crm_xml_add(notify_data, F_STONITH_TARGET, cmd->victim);
crm_xml_add(notify_data, F_STONITH_OPERATION, cmd->op);
crm_xml_add(notify_data, F_STONITH_DELEGATE, "localhost");
crm_xml_add(notify_data, F_STONITH_DEVICE, cmd->device);
crm_xml_add(notify_data, F_STONITH_REMOTE_OP_ID, cmd->remote_op_id);
crm_xml_add(notify_data, F_STONITH_ORIGIN, cmd->client);
do_stonith_notify(0, T_STONITH_NOTIFY_FENCE, rc, notify_data);
do_stonith_notify(0, T_STONITH_NOTIFY_HISTORY, 0, NULL);
}
free_xml(reply);
}
static void
cancel_stonith_command(async_command_t * cmd)
{
stonith_device_t *device;
CRM_CHECK(cmd != NULL, return);
if (!cmd->device) {
return;
}
device = g_hash_table_lookup(device_list, cmd->device);
if (device) {
crm_trace("Cancel scheduled '%s' action on %s", cmd->action, device->id);
device->pending_ops = g_list_remove(device->pending_ops, cmd);
}
}
static void
st_child_done(GPid pid, int rc, const char *output, gpointer user_data)
{
stonith_device_t *device = NULL;
stonith_device_t *next_device = NULL;
async_command_t *cmd = user_data;
GListPtr gIter = NULL;
GListPtr gIterNext = NULL;
CRM_CHECK(cmd != NULL, return);
cmd->active_on = NULL;
/* The device is ready to do something else now */
device = g_hash_table_lookup(device_list, cmd->device);
if (device) {
if (!device->verified && (rc == pcmk_ok) &&
(safe_str_eq(cmd->action, "list") ||
safe_str_eq(cmd->action, "monitor") || safe_str_eq(cmd->action, "status"))) {
device->verified = TRUE;
}
mainloop_set_trigger(device->work);
}
crm_debug("Operation '%s' on '%s' completed with rc=%d (%d remaining)",
cmd->action, cmd->device, rc, g_list_length(cmd->device_next));
if (rc == 0) {
GListPtr iter;
/* see if there are any required devices left to execute for this op */
for (iter = cmd->device_next; iter != NULL; iter = iter->next) {
next_device = g_hash_table_lookup(device_list, iter->data);
if (next_device != NULL && is_action_required(cmd->action, next_device)) {
cmd->device_next = iter->next;
break;
}
next_device = NULL;
}
} else if (rc != 0 && cmd->device_next && (is_action_required(cmd->action, device) == FALSE)) {
/* if this device didn't work out, see if there are any others we can try.
* if the failed device was 'required', we can't pick another device. */
next_device = g_hash_table_lookup(device_list, cmd->device_next->data);
cmd->device_next = cmd->device_next->next;
}
/* this operation requires more fencing, hooray! */
if (next_device) {
log_operation(cmd, rc, pid, next_device->id, output, FALSE);
schedule_stonith_command(cmd, next_device);
/* Prevent cmd from being freed */
cmd = NULL;
goto done;
}
stonith_send_async_reply(cmd, output, rc, pid, st_reply_opt_none);
if (rc != 0) {
goto done;
}
/* Check to see if any operations are scheduled to do the exact
* same thing that just completed. If so, rather than
* performing the same fencing operation twice, return the result
* of this operation for all pending commands it matches. */
for (gIter = cmd_list; gIter != NULL; gIter = gIterNext) {
async_command_t *cmd_other = gIter->data;
gIterNext = gIter->next;
if (cmd == cmd_other) {
continue;
}
/* A pending scheduled command matches the command that just finished if.
* 1. The client connections are different.
* 2. The node victim is the same.
* 3. The fencing action is the same.
* 4. The device scheduled to execute the action is the same.
*/
if (safe_str_eq(cmd->client, cmd_other->client) ||
safe_str_neq(cmd->victim, cmd_other->victim) ||
safe_str_neq(cmd->action, cmd_other->action) ||
safe_str_neq(cmd->device, cmd_other->device)) {
continue;
}
/* Duplicate merging will do the right thing for either type of remapped
* reboot. If the executing fencer remapped an unsupported reboot to
* off, then cmd->action will be reboot and will be merged with any
* other reboot requests. If the originating fencer remapped a
* topology reboot to off then on, we will get here once with
* cmd->action "off" and once with "on", and they will be merged
* separately with similar requests.
*/
crm_notice
("Merging stonith action '%s' targeting %s originating from client %s with identical stonith request from client %s",
cmd_other->action, cmd_other->victim, cmd_other->client_name, cmd->client_name);
cmd_list = g_list_remove_link(cmd_list, gIter);
stonith_send_async_reply(cmd_other, output, rc, pid, st_reply_opt_merged);
cancel_stonith_command(cmd_other);
free_async_command(cmd_other);
g_list_free_1(gIter);
}
done:
free_async_command(cmd);
}
static gint
sort_device_priority(gconstpointer a, gconstpointer b)
{
const stonith_device_t *dev_a = a;
const stonith_device_t *dev_b = b;
if (dev_a->priority > dev_b->priority) {
return -1;
} else if (dev_a->priority < dev_b->priority) {
return 1;
}
return 0;
}
static void
stonith_fence_get_devices_cb(GList * devices, void *user_data)
{
async_command_t *cmd = user_data;
stonith_device_t *device = NULL;
crm_info("Found %d matching devices for '%s'", g_list_length(devices), cmd->victim);
if (devices != NULL) {
/* Order based on priority */
devices = g_list_sort(devices, sort_device_priority);
device = g_hash_table_lookup(device_list, devices->data);
if (device) {
cmd->device_list = devices;
cmd->device_next = devices->next;
devices = NULL; /* list owned by cmd now */
}
}
/* we have a device, schedule it for fencing. */
if (device) {
schedule_stonith_command(cmd, device);
/* in progress */
return;
}
/* no device found! */
stonith_send_async_reply(cmd, NULL, -ENODEV, 0, st_reply_opt_none);
free_async_command(cmd);
g_list_free_full(devices, free);
}
static int
stonith_fence(xmlNode * msg)
{
const char *device_id = NULL;
stonith_device_t *device = NULL;
async_command_t *cmd = create_async_command(msg);
xmlNode *dev = get_xpath_object("//@" F_STONITH_TARGET, msg, LOG_ERR);
if (cmd == NULL) {
return -EPROTO;
}
device_id = crm_element_value(dev, F_STONITH_DEVICE);
if (device_id) {
device = g_hash_table_lookup(device_list, device_id);
if (device == NULL) {
crm_err("Requested device '%s' is not available", device_id);
return -ENODEV;
}
schedule_stonith_command(cmd, device);
} else {
const char *host = crm_element_value(dev, F_STONITH_TARGET);
if (cmd->options & st_opt_cs_nodeid) {
int nodeid = crm_atoi(host, NULL);
crm_node_t *node = crm_find_known_peer_full(nodeid, NULL, CRM_GET_PEER_ANY);
if (node) {
host = node->uname;
}
}
/* If we get to here, then self-fencing is implicitly allowed */
get_capable_devices(host, cmd->action, cmd->default_timeout,
TRUE, cmd, stonith_fence_get_devices_cb);
}
return -EINPROGRESS;
}
xmlNode *
stonith_construct_reply(xmlNode * request, const char *output, xmlNode * data, int rc)
{
int lpc = 0;
xmlNode *reply = NULL;
const char *name = NULL;
const char *value = NULL;
const char *names[] = {
F_STONITH_OPERATION,
F_STONITH_CALLID,
F_STONITH_CLIENTID,
F_STONITH_CLIENTNAME,
F_STONITH_REMOTE_OP_ID,
F_STONITH_CALLOPTS
};
crm_trace("Creating a basic reply");
reply = create_xml_node(NULL, T_STONITH_REPLY);
crm_xml_add(reply, "st_origin", __FUNCTION__);
crm_xml_add(reply, F_TYPE, T_STONITH_NG);
crm_xml_add(reply, "st_output", output);
crm_xml_add_int(reply, F_STONITH_RC, rc);
CRM_CHECK(request != NULL, crm_warn("Can't create a sane reply"); return reply);
for (lpc = 0; lpc < DIMOF(names); lpc++) {
name = names[lpc];
value = crm_element_value(request, name);
crm_xml_add(reply, name, value);
}
if (data != NULL) {
crm_trace("Attaching reply output");
add_message_xml(reply, F_STONITH_CALLDATA, data);
}
return reply;
}
static xmlNode *
stonith_construct_async_reply(async_command_t * cmd, const char *output, xmlNode * data, int rc)
{
xmlNode *reply = NULL;
crm_trace("Creating a basic reply");
reply = create_xml_node(NULL, T_STONITH_REPLY);
crm_xml_add(reply, "st_origin", __FUNCTION__);
crm_xml_add(reply, F_TYPE, T_STONITH_NG);
crm_xml_add(reply, F_STONITH_OPERATION, cmd->op);
crm_xml_add(reply, F_STONITH_DEVICE, cmd->device);
crm_xml_add(reply, F_STONITH_REMOTE_OP_ID, cmd->remote_op_id);
crm_xml_add(reply, F_STONITH_CLIENTID, cmd->client);
crm_xml_add(reply, F_STONITH_CLIENTNAME, cmd->client_name);
crm_xml_add(reply, F_STONITH_TARGET, cmd->victim);
crm_xml_add(reply, F_STONITH_ACTION, cmd->op);
crm_xml_add(reply, F_STONITH_ORIGIN, cmd->origin);
crm_xml_add_int(reply, F_STONITH_CALLID, cmd->id);
crm_xml_add_int(reply, F_STONITH_CALLOPTS, cmd->options);
crm_xml_add_int(reply, F_STONITH_RC, rc);
crm_xml_add(reply, "st_output", output);
if (data != NULL) {
crm_info("Attaching reply output");
add_message_xml(reply, F_STONITH_CALLDATA, data);
}
return reply;
}
bool fencing_peer_active(crm_node_t *peer)
{
if (peer == NULL) {
return FALSE;
} else if (peer->uname == NULL) {
return FALSE;
} else if (is_set(peer->processes, crm_get_cluster_proc())) {
return TRUE;
}
return FALSE;
}
/*!
* \internal
* \brief Determine if we need to use an alternate node to
* fence the target. If so return that node's uname
*
* \retval NULL, no alternate host
* \retval uname, uname of alternate host to use
*/
static const char *
check_alternate_host(const char *target)
{
const char *alternate_host = NULL;
crm_trace("Checking if we (%s) can fence %s", stonith_our_uname, target);
if (find_topology_for_host(target) && safe_str_eq(target, stonith_our_uname)) {
GHashTableIter gIter;
crm_node_t *entry = NULL;
g_hash_table_iter_init(&gIter, crm_peer_cache);
while (g_hash_table_iter_next(&gIter, NULL, (void **)&entry)) {
crm_trace("Checking for %s.%d != %s", entry->uname, entry->id, target);
if (fencing_peer_active(entry)
&& safe_str_neq(entry->uname, target)) {
alternate_host = entry->uname;
break;
}
}
if (alternate_host == NULL) {
crm_err("No alternate host available to handle complex self fencing request");
g_hash_table_iter_init(&gIter, crm_peer_cache);
while (g_hash_table_iter_next(&gIter, NULL, (void **)&entry)) {
crm_notice("Peer[%d] %s", entry->id, entry->uname);
}
}
}
return alternate_host;
}
static void
stonith_send_reply(xmlNode * reply, int call_options, const char *remote_peer,
const char *client_id)
{
if (remote_peer) {
send_cluster_message(crm_get_peer(0, remote_peer), crm_msg_stonith_ng, reply, FALSE);
} else {
do_local_reply(reply, client_id, is_set(call_options, st_opt_sync_call), remote_peer != NULL);
}
}
static int
handle_request(pcmk__client_t *client, uint32_t id, uint32_t flags,
xmlNode *request, const char *remote_peer)
{
int call_options = 0;
int rc = -EOPNOTSUPP;
xmlNode *data = NULL;
xmlNode *reply = NULL;
char *output = NULL;
const char *op = crm_element_value(request, F_STONITH_OPERATION);
const char *client_id = crm_element_value(request, F_STONITH_CLIENTID);
crm_element_value_int(request, F_STONITH_CALLOPTS, &call_options);
if (is_set(call_options, st_opt_sync_call)) {
CRM_ASSERT(client == NULL || client->request_id == id);
}
if (crm_str_eq(op, CRM_OP_REGISTER, TRUE)) {
xmlNode *reply = create_xml_node(NULL, "reply");
CRM_ASSERT(client);
crm_xml_add(reply, F_STONITH_OPERATION, CRM_OP_REGISTER);
crm_xml_add(reply, F_STONITH_CLIENTID, client->id);
pcmk__ipc_send_xml(client, id, reply, flags);
client->request_id = 0;
free_xml(reply);
return 0;
} else if (crm_str_eq(op, STONITH_OP_EXEC, TRUE)) {
rc = stonith_device_action(request, &output);
} else if (crm_str_eq(op, STONITH_OP_TIMEOUT_UPDATE, TRUE)) {
const char *call_id = crm_element_value(request, F_STONITH_CALLID);
const char *client_id = crm_element_value(request, F_STONITH_CLIENTID);
int op_timeout = 0;
crm_element_value_int(request, F_STONITH_TIMEOUT, &op_timeout);
do_stonith_async_timeout_update(client_id, call_id, op_timeout);
return 0;
} else if (crm_str_eq(op, STONITH_OP_QUERY, TRUE)) {
if (remote_peer) {
create_remote_stonith_op(client_id, request, TRUE); /* Record it for the future notification */
}
stonith_query(request, remote_peer, client_id, call_options);
return 0;
} else if (crm_str_eq(op, T_STONITH_NOTIFY, TRUE)) {
const char *flag_name = NULL;
CRM_ASSERT(client);
flag_name = crm_element_value(request, F_STONITH_NOTIFY_ACTIVATE);
if (flag_name) {
crm_debug("Setting %s callbacks for %s (%s): ON", flag_name, client->name, client->id);
client->options |= get_stonith_flag(flag_name);
}
flag_name = crm_element_value(request, F_STONITH_NOTIFY_DEACTIVATE);
if (flag_name) {
crm_debug("Setting %s callbacks for %s (%s): off", flag_name, client->name, client->id);
client->options |= get_stonith_flag(flag_name);
}
if (flags & crm_ipc_client_response) {
pcmk__ipc_send_ack(client, id, flags, "ack");
}
return 0;
} else if (crm_str_eq(op, STONITH_OP_RELAY, TRUE)) {
xmlNode *dev = get_xpath_object("//@" F_STONITH_TARGET, request, LOG_TRACE);
crm_notice("Peer %s has received a forwarded fencing request from %s to fence (%s) peer %s",
stonith_our_uname,
client ? client->name : remote_peer,
crm_element_value(dev, F_STONITH_ACTION),
crm_element_value(dev, F_STONITH_TARGET));
if (initiate_remote_stonith_op(NULL, request, FALSE) != NULL) {
rc = -EINPROGRESS;
}
} else if (crm_str_eq(op, STONITH_OP_FENCE, TRUE)) {
if (remote_peer || stand_alone) {
rc = stonith_fence(request);
} else if (call_options & st_opt_manual_ack) {
remote_fencing_op_t *rop = NULL;
xmlNode *dev = get_xpath_object("//@" F_STONITH_TARGET, request, LOG_TRACE);
const char *target = crm_element_value(dev, F_STONITH_TARGET);
crm_notice("Received manual confirmation that %s is fenced", target);
rop = initiate_remote_stonith_op(client, request, TRUE);
rc = stonith_manual_ack(request, rop);
} else {
const char *alternate_host = NULL;
xmlNode *dev = get_xpath_object("//@" F_STONITH_TARGET, request, LOG_TRACE);
const char *target = crm_element_value(dev, F_STONITH_TARGET);
const char *action = crm_element_value(dev, F_STONITH_ACTION);
const char *device = crm_element_value(dev, F_STONITH_DEVICE);
if (client) {
int tolerance = 0;
crm_notice("Client %s.%.8s wants to fence (%s) '%s' with device '%s'",
client->name, client->id, action, target, device ? device : "(any)");
crm_element_value_int(dev, F_STONITH_TOLERANCE, &tolerance);
if (stonith_check_fence_tolerance(tolerance, target, action)) {
rc = 0;
goto done;
}
} else {
crm_notice("Peer %s wants to fence (%s) '%s' with device '%s'",
remote_peer, action, target, device ? device : "(any)");
}
alternate_host = check_alternate_host(target);
if (alternate_host && client) {
const char *client_id = NULL;
crm_notice("Forwarding complex self fencing request to peer %s", alternate_host);
if (client->id) {
client_id = client->id;
} else {
client_id = crm_element_value(request, F_STONITH_CLIENTID);
}
/* Create a record of it, otherwise call_id will be 0 if we need to notify of failures */
create_remote_stonith_op(client_id, request, FALSE);
crm_xml_add(request, F_STONITH_OPERATION, STONITH_OP_RELAY);
crm_xml_add(request, F_STONITH_CLIENTID, client->id);
send_cluster_message(crm_get_peer(0, alternate_host), crm_msg_stonith_ng, request,
FALSE);
rc = -EINPROGRESS;
} else if (initiate_remote_stonith_op(client, request, FALSE) != NULL) {
rc = -EINPROGRESS;
}
}
} else if (crm_str_eq(op, STONITH_OP_FENCE_HISTORY, TRUE)) {
rc = stonith_fence_history(request, &data, remote_peer, call_options);
if (call_options & st_opt_discard_reply) {
/* we don't expect answers to the broadcast
* we might have sent out
*/
free_xml(data);
return pcmk_ok;
}
} else if (crm_str_eq(op, STONITH_OP_DEVICE_ADD, TRUE)) {
const char *device_id = NULL;
rc = stonith_device_register(request, &device_id, FALSE);
do_stonith_notify_device(call_options, op, rc, device_id);
} else if (crm_str_eq(op, STONITH_OP_DEVICE_DEL, TRUE)) {
xmlNode *dev = get_xpath_object("//" F_STONITH_DEVICE, request, LOG_ERR);
const char *device_id = crm_element_value(dev, XML_ATTR_ID);
rc = stonith_device_remove(device_id, FALSE);
do_stonith_notify_device(call_options, op, rc, device_id);
} else if (crm_str_eq(op, STONITH_OP_LEVEL_ADD, TRUE)) {
char *device_id = NULL;
rc = stonith_level_register(request, &device_id);
do_stonith_notify_level(call_options, op, rc, device_id);
free(device_id);
} else if (crm_str_eq(op, STONITH_OP_LEVEL_DEL, TRUE)) {
char *device_id = NULL;
rc = stonith_level_remove(request, &device_id);
do_stonith_notify_level(call_options, op, rc, device_id);
} else if(safe_str_eq(op, CRM_OP_RM_NODE_CACHE)) {
int node_id = 0;
const char *name = NULL;
crm_element_value_int(request, XML_ATTR_ID, &node_id);
name = crm_element_value(request, XML_ATTR_UNAME);
reap_crm_member(node_id, name);
return pcmk_ok;
} else {
crm_err("Unknown IPC request %s from %s",
op, (client? client->name : remote_peer));
}
done:
/* Always reply unless the request is in process still.
* If in progress, a reply will happen async after the request
* processing is finished */
if (rc != -EINPROGRESS) {
crm_trace("Reply handling: %p %u %u %d %d %s", client, client?client->request_id:0,
id, is_set(call_options, st_opt_sync_call), call_options,
crm_element_value(request, F_STONITH_CALLOPTS));
if (is_set(call_options, st_opt_sync_call)) {
CRM_ASSERT(client == NULL || client->request_id == id);
}
reply = stonith_construct_reply(request, output, data, rc);
stonith_send_reply(reply, call_options, remote_peer, client_id);
}
free(output);
free_xml(data);
free_xml(reply);
return rc;
}
static void
handle_reply(pcmk__client_t *client, xmlNode *request, const char *remote_peer)
{
const char *op = crm_element_value(request, F_STONITH_OPERATION);
if (crm_str_eq(op, STONITH_OP_QUERY, TRUE)) {
process_remote_stonith_query(request);
} else if (crm_str_eq(op, T_STONITH_NOTIFY, TRUE)) {
process_remote_stonith_exec(request);
} else if (crm_str_eq(op, STONITH_OP_FENCE, TRUE)) {
/* Reply to a complex fencing op */
process_remote_stonith_exec(request);
} else {
crm_err("Unknown %s reply from %s", op, client ? client->name : remote_peer);
crm_log_xml_warn(request, "UnknownOp");
}
}
void
stonith_command(pcmk__client_t *client, uint32_t id, uint32_t flags,
xmlNode *request, const char *remote_peer)
{
int call_options = 0;
int rc = 0;
gboolean is_reply = FALSE;
/* Copy op for reporting. The original might get freed by handle_reply()
* before we use it in crm_debug():
* handle_reply()
* |- process_remote_stonith_exec()
* |-- remote_op_done()
* |--- handle_local_reply_and_notify()
* |---- crm_xml_add(...F_STONITH_OPERATION...)
* |--- free_xml(op->request)
*/
char *op = crm_element_value_copy(request, F_STONITH_OPERATION);
if (get_xpath_object("//" T_STONITH_REPLY, request, LOG_NEVER)) {
is_reply = TRUE;
}
crm_element_value_int(request, F_STONITH_CALLOPTS, &call_options);
crm_debug("Processing %s%s %u from %s (%16x)", op, is_reply ? " reply" : "",
id, client ? client->name : remote_peer, call_options);
if (is_set(call_options, st_opt_sync_call)) {
CRM_ASSERT(client == NULL || client->request_id == id);
}
if (is_reply) {
handle_reply(client, request, remote_peer);
} else {
rc = handle_request(client, id, flags, request, remote_peer);
}
crm_debug("Processed %s%s from %s: %s (%d)", op,
is_reply ? " reply" : "", client ? client->name : remote_peer,
rc > 0 ? "" : pcmk_strerror(rc), rc);
free(op);
}
diff --git a/daemons/fenced/fenced_remote.c b/daemons/fenced/fenced_remote.c
index 11125d7360..ac9bcb0692 100644
--- a/daemons/fenced/fenced_remote.c
+++ b/daemons/fenced/fenced_remote.c
@@ -1,2101 +1,2120 @@
/*
* Copyright 2009-2020 the Pacemaker project contributors
*
* The version control history for this file may have further details.
*
* This source code is licensed under the GNU General Public License version 2
* or later (GPLv2+) WITHOUT ANY WARRANTY.
*/
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#define TIMEOUT_MULTIPLY_FACTOR 1.2
/* When one fencer queries its peers for devices able to handle a fencing
* request, each peer will reply with a list of such devices available to it.
* Each reply will be parsed into a st_query_result_t, with each device's
* information kept in a device_properties_t.
*/
typedef struct device_properties_s {
/* Whether access to this device has been verified */
gboolean verified;
/* The remaining members are indexed by the operation's "phase" */
/* Whether this device has been executed in each phase */
gboolean executed[st_phase_max];
/* Whether this device is disallowed from executing in each phase */
gboolean disallowed[st_phase_max];
/* Action-specific timeout for each phase */
int custom_action_timeout[st_phase_max];
/* Action-specific maximum random delay for each phase */
int delay_max[st_phase_max];
/* Action-specific base delay for each phase */
int delay_base[st_phase_max];
} device_properties_t;
typedef struct st_query_result_s {
/* Name of peer that sent this result */
char *host;
/* Only try peers for non-topology based operations once */
gboolean tried;
/* Number of entries in the devices table */
int ndevices;
/* Devices available to this host that are capable of fencing the target */
GHashTable *devices;
} st_query_result_t;
GHashTable *stonith_remote_op_list = NULL;
void call_remote_stonith(remote_fencing_op_t * op, st_query_result_t * peer);
static void remote_op_done(remote_fencing_op_t * op, xmlNode * data, int rc, int dup);
extern xmlNode *stonith_create_op(int call_id, const char *token, const char *op, xmlNode * data,
int call_options);
static void report_timeout_period(remote_fencing_op_t * op, int op_timeout);
static int get_op_total_timeout(const remote_fencing_op_t *op,
const st_query_result_t *chosen_peer);
static gint
sort_strings(gconstpointer a, gconstpointer b)
{
return strcmp(a, b);
}
static void
free_remote_query(gpointer data)
{
if (data) {
st_query_result_t *query = data;
crm_trace("Free'ing query result from %s", query->host);
g_hash_table_destroy(query->devices);
free(query->host);
free(query);
}
}
void
free_stonith_remote_op_list()
{
if (stonith_remote_op_list != NULL) {
g_hash_table_destroy(stonith_remote_op_list);
stonith_remote_op_list = NULL;
}
}
struct peer_count_data {
const remote_fencing_op_t *op;
gboolean verified_only;
int count;
};
/*!
* \internal
* \brief Increment a counter if a device has not been executed yet
*
* \param[in] key Device ID (ignored)
* \param[in] value Device properties
* \param[in] user_data Peer count data
*/
static void
count_peer_device(gpointer key, gpointer value, gpointer user_data)
{
device_properties_t *props = (device_properties_t*)value;
struct peer_count_data *data = user_data;
if (!props->executed[data->op->phase]
&& (!data->verified_only || props->verified)) {
++(data->count);
}
}
/*!
* \internal
* \brief Check the number of available devices in a peer's query results
*
* \param[in] op Operation that results are for
* \param[in] peer Peer to count
* \param[in] verified_only Whether to count only verified devices
*
* \return Number of devices available to peer that were not already executed
*/
static int
count_peer_devices(const remote_fencing_op_t *op, const st_query_result_t *peer,
gboolean verified_only)
{
struct peer_count_data data;
data.op = op;
data.verified_only = verified_only;
data.count = 0;
if (peer) {
g_hash_table_foreach(peer->devices, count_peer_device, &data);
}
return data.count;
}
/*!
* \internal
* \brief Search for a device in a query result
*
* \param[in] op Operation that result is for
* \param[in] peer Query result for a peer
* \param[in] device Device ID to search for
*
* \return Device properties if found, NULL otherwise
*/
static device_properties_t *
find_peer_device(const remote_fencing_op_t *op, const st_query_result_t *peer,
const char *device)
{
device_properties_t *props = g_hash_table_lookup(peer->devices, device);
return (props && !props->executed[op->phase]
&& !props->disallowed[op->phase])? props : NULL;
}
/*!
* \internal
* \brief Find a device in a peer's device list and mark it as executed
*
* \param[in] op Operation that peer result is for
* \param[in,out] peer Peer with results to search
* \param[in] device ID of device to mark as done
* \param[in] verified_devices_only Only consider verified devices
*
* \return TRUE if device was found and marked, FALSE otherwise
*/
static gboolean
grab_peer_device(const remote_fencing_op_t *op, st_query_result_t *peer,
const char *device, gboolean verified_devices_only)
{
device_properties_t *props = find_peer_device(op, peer, device);
if ((props == NULL) || (verified_devices_only && !props->verified)) {
return FALSE;
}
crm_trace("Removing %s from %s (%d remaining)",
device, peer->host, count_peer_devices(op, peer, FALSE));
props->executed[op->phase] = TRUE;
return TRUE;
}
static void
clear_remote_op_timers(remote_fencing_op_t * op)
{
if (op->query_timer) {
g_source_remove(op->query_timer);
op->query_timer = 0;
}
if (op->op_timer_total) {
g_source_remove(op->op_timer_total);
op->op_timer_total = 0;
}
if (op->op_timer_one) {
g_source_remove(op->op_timer_one);
op->op_timer_one = 0;
}
}
static void
free_remote_op(gpointer data)
{
remote_fencing_op_t *op = data;
crm_trace("Free'ing op %s for %s", op->id, op->target);
crm_log_xml_debug(op->request, "Destroying");
clear_remote_op_timers(op);
free(op->id);
free(op->action);
free(op->delegate);
free(op->target);
free(op->client_id);
free(op->client_name);
free(op->originator);
if (op->query_results) {
g_list_free_full(op->query_results, free_remote_query);
}
if (op->request) {
free_xml(op->request);
op->request = NULL;
}
if (op->devices_list) {
g_list_free_full(op->devices_list, free);
op->devices_list = NULL;
}
g_list_free_full(op->automatic_list, free);
g_list_free(op->duplicates);
free(op);
}
void
init_stonith_remote_op_hash_table(GHashTable **table)
{
if (*table == NULL) {
*table = g_hash_table_new_full(crm_str_hash, g_str_equal, NULL, free_remote_op);
}
}
/*!
* \internal
* \brief Return an operation's originally requested action (before any remap)
*
* \param[in] op Operation to check
*
* \return Operation's original action
*/
static const char *
op_requested_action(const remote_fencing_op_t *op)
{
return ((op->phase > st_phase_requested)? "reboot" : op->action);
}
/*!
* \internal
* \brief Remap a "reboot" operation to the "off" phase
*
* \param[in,out] op Operation to remap
*/
static void
op_phase_off(remote_fencing_op_t *op)
{
crm_info("Remapping multiple-device reboot targeting %s (%s) to 'off'",
op->target, op->id);
op->phase = st_phase_off;
/* Happily, "off" and "on" are shorter than "reboot", so we can reuse the
* memory allocation at each phase.
*/
strcpy(op->action, "off");
}
/*!
* \internal
* \brief Advance a remapped reboot operation to the "on" phase
*
* \param[in,out] op Operation to remap
*/
static void
op_phase_on(remote_fencing_op_t *op)
{
GListPtr iter = NULL;
crm_info("Remapped 'off' targeting %s complete, "
"remapping to 'on' for %s.%.8s",
op->target, op->client_name, op->id);
op->phase = st_phase_on;
strcpy(op->action, "on");
/* Skip devices with automatic unfencing, because the cluster will handle it
* when the node rejoins.
*/
for (iter = op->automatic_list; iter != NULL; iter = iter->next) {
GListPtr match = g_list_find_custom(op->devices_list, iter->data,
sort_strings);
if (match) {
op->devices_list = g_list_remove(op->devices_list, match->data);
}
}
g_list_free_full(op->automatic_list, free);
op->automatic_list = NULL;
/* Rewind device list pointer */
op->devices = op->devices_list;
}
/*!
* \internal
* \brief Reset a remapped reboot operation
*
* \param[in,out] op Operation to reset
*/
static void
undo_op_remap(remote_fencing_op_t *op)
{
if (op->phase > 0) {
crm_info("Undoing remap of reboot targeting %s for %s.%.8s",
op->target, op->client_name, op->id);
op->phase = st_phase_requested;
strcpy(op->action, "reboot");
}
}
static xmlNode *
create_op_done_notify(remote_fencing_op_t * op, int rc)
{
xmlNode *notify_data = create_xml_node(NULL, T_STONITH_NOTIFY_FENCE);
crm_xml_add_int(notify_data, "state", op->state);
crm_xml_add_int(notify_data, F_STONITH_RC, rc);
crm_xml_add(notify_data, F_STONITH_TARGET, op->target);
crm_xml_add(notify_data, F_STONITH_ACTION, op->action);
crm_xml_add(notify_data, F_STONITH_DELEGATE, op->delegate);
crm_xml_add(notify_data, F_STONITH_REMOTE_OP_ID, op->id);
crm_xml_add(notify_data, F_STONITH_ORIGIN, op->originator);
crm_xml_add(notify_data, F_STONITH_CLIENTID, op->client_id);
crm_xml_add(notify_data, F_STONITH_CLIENTNAME, op->client_name);
return notify_data;
}
void
stonith_bcast_result_to_peers(remote_fencing_op_t * op, int rc, gboolean op_merged)
{
static int count = 0;
xmlNode *bcast = create_xml_node(NULL, T_STONITH_REPLY);
xmlNode *notify_data = create_op_done_notify(op, rc);
count++;
crm_trace("Broadcasting result to peers");
crm_xml_add(bcast, F_TYPE, T_STONITH_NOTIFY);
crm_xml_add(bcast, F_SUBTYPE, "broadcast");
crm_xml_add(bcast, F_STONITH_OPERATION, T_STONITH_NOTIFY);
crm_xml_add_int(bcast, "count", count);
if (op_merged) {
crm_xml_add(bcast, F_STONITH_MERGED, "true");
}
add_message_xml(bcast, F_STONITH_CALLDATA, notify_data);
send_cluster_message(NULL, crm_msg_stonith_ng, bcast, FALSE);
free_xml(notify_data);
free_xml(bcast);
return;
}
static void
handle_local_reply_and_notify(remote_fencing_op_t * op, xmlNode * data, int rc)
{
xmlNode *notify_data = NULL;
xmlNode *reply = NULL;
if (op->notify_sent == TRUE) {
/* nothing to do */
return;
}
/* Do notification with a clean data object */
notify_data = create_op_done_notify(op, rc);
crm_xml_add_int(data, "state", op->state);
crm_xml_add(data, F_STONITH_TARGET, op->target);
crm_xml_add(data, F_STONITH_OPERATION, op->action);
reply = stonith_construct_reply(op->request, NULL, data, rc);
crm_xml_add(reply, F_STONITH_DELEGATE, op->delegate);
/* Send fencing OP reply to local client that initiated fencing */
do_local_reply(reply, op->client_id, op->call_options & st_opt_sync_call, FALSE);
/* bcast to all local clients that the fencing operation happend */
do_stonith_notify(0, T_STONITH_NOTIFY_FENCE, rc, notify_data);
do_stonith_notify(0, T_STONITH_NOTIFY_HISTORY, 0, NULL);
/* mark this op as having notify's already sent */
op->notify_sent = TRUE;
free_xml(reply);
free_xml(notify_data);
}
static void
handle_duplicates(remote_fencing_op_t * op, xmlNode * data, int rc)
{
GListPtr iter = NULL;
for (iter = op->duplicates; iter != NULL; iter = iter->next) {
remote_fencing_op_t *other = iter->data;
if (other->state == st_duplicate) {
other->state = op->state;
crm_debug("Performing duplicate notification for %s@%s.%.8s = %s",
other->client_name, other->originator, other->id,
pcmk_strerror(rc));
remote_op_done(other, data, rc, TRUE);
} else {
// Possible if (for example) it timed out already
crm_err("Skipping duplicate notification for %s@%s - %d", other->client_name,
other->originator, other->state);
}
}
}
/*!
* \internal
* \brief Finalize a remote operation.
*
* \description This function has two code paths.
*
* Path 1. This node is the owner of the operation and needs
* to notify the cpg group via a broadcast as to the operation's
* results.
*
* Path 2. The cpg broadcast is received. All nodes notify their local
* stonith clients the operation results.
*
* So, The owner of the operation first notifies the cluster of the result,
* and once that cpg notify is received back it notifies all the local clients.
*
* Nodes that are passive watchers of the operation will receive the
* broadcast and only need to notify their local clients the operation finished.
*
* \param op, The fencing operation to finalize
* \param data, The xml msg reply (if present) of the last delegated fencing
* operation.
* \param dup, Is this operation a duplicate, if so treat it a little differently
* making sure the broadcast is not sent out.
*/
static void
remote_op_done(remote_fencing_op_t * op, xmlNode * data, int rc, int dup)
{
int level = LOG_ERR;
const char *subt = NULL;
xmlNode *local_data = NULL;
gboolean op_merged = FALSE;
op->completed = time(NULL);
clear_remote_op_timers(op);
undo_op_remap(op);
if (op->notify_sent == TRUE) {
crm_err("Already sent notifications for '%s' targeting %s on %s for "
"client %s@%s.%.8s: %s " CRM_XS " rc=%d state=%d",
op->action, op->target,
(op->delegate? op->delegate : "unknown node"),
op->client_name, op->originator, op->id, pcmk_strerror(rc),
rc, op->state);
goto remote_op_done_cleanup;
}
if (!op->delegate && data && rc != -ENODEV && rc != -EHOSTUNREACH) {
xmlNode *ndata = get_xpath_object("//@" F_STONITH_DELEGATE, data,
LOG_NEVER);
if(ndata) {
op->delegate = crm_element_value_copy(ndata, F_STONITH_DELEGATE);
} else {
op->delegate = crm_element_value_copy(data, F_ORIG);
}
}
if (data == NULL) {
data = create_xml_node(NULL, "remote-op");
local_data = data;
}
if(dup) {
op_merged = TRUE;
} else if (crm_element_value(data, F_STONITH_MERGED)) {
op_merged = TRUE;
}
/* Tell everyone the operation is done, we will continue
* with doing the local notifications once we receive
* the broadcast back. */
subt = crm_element_value(data, F_SUBTYPE);
if (dup == FALSE && safe_str_neq(subt, "broadcast")) {
/* Defer notification until the bcast message arrives */
stonith_bcast_result_to_peers(op, rc, (op_merged? TRUE: FALSE));
goto remote_op_done_cleanup;
}
if (rc == pcmk_ok || dup) {
level = LOG_NOTICE;
} else if (safe_str_neq(op->originator, stonith_our_uname)) {
level = LOG_NOTICE;
}
do_crm_log(level, "Operation '%s'%s%s on %s for %s@%s.%.8s%s: %s",
op->action, (op->target? " targeting " : ""),
(op->target? op->target : ""),
(op->delegate? op->delegate : ""),
op->client_name, op->originator, op->id,
(op_merged? " (merged)" : ""), pcmk_strerror(rc));
handle_local_reply_and_notify(op, data, rc);
if (dup == FALSE) {
handle_duplicates(op, data, rc);
}
/* Free non-essential parts of the record
* Keep the record around so we can query the history
*/
if (op->query_results) {
g_list_free_full(op->query_results, free_remote_query);
op->query_results = NULL;
}
if (op->request) {
free_xml(op->request);
op->request = NULL;
}
remote_op_done_cleanup:
free_xml(local_data);
}
static gboolean
remote_op_watchdog_done(gpointer userdata)
{
remote_fencing_op_t *op = userdata;
op->op_timer_one = 0;
crm_notice("Self-fencing (%s) by %s for %s.%8s assumed complete",
op->action, op->target, op->client_name, op->id);
op->state = st_done;
remote_op_done(op, NULL, pcmk_ok, FALSE);
return FALSE;
}
static gboolean
remote_op_timeout_one(gpointer userdata)
{
remote_fencing_op_t *op = userdata;
op->op_timer_one = 0;
crm_notice("Peer's '%s' action targeting %s for client %s timed out " CRM_XS
" id=%s", op->action, op->target, op->client_name, op->id);
call_remote_stonith(op, NULL);
return FALSE;
}
static gboolean
remote_op_timeout(gpointer userdata)
{
remote_fencing_op_t *op = userdata;
op->op_timer_total = 0;
if (op->state == st_done) {
crm_debug("Action '%s' targeting %s for client %s already completed "
CRM_XS " id=%s",
op->action, op->target, op->client_name, op->id);
return FALSE;
}
crm_debug("Action '%s' targeting %s for client %s timed out "
CRM_XS " id=%s",
op->action, op->target, op->client_name, op->id);
if (op->phase == st_phase_on) {
/* A remapped reboot operation timed out in the "on" phase, but the
* "off" phase completed successfully, so quit trying any further
* devices, and return success.
*/
remote_op_done(op, NULL, pcmk_ok, FALSE);
return FALSE;
}
op->state = st_failed;
remote_op_done(op, NULL, -ETIME, FALSE);
return FALSE;
}
static gboolean
remote_op_query_timeout(gpointer data)
{
remote_fencing_op_t *op = data;
op->query_timer = 0;
if (op->state == st_done) {
crm_debug("Operation %s targeting %s already completed",
op->id, op->target);
} else if (op->state == st_exec) {
crm_debug("Operation %s targeting %s already in progress",
op->id, op->target);
} else if (op->query_results) {
crm_debug("Query %s targeting %s complete (state=%d)",
op->id, op->target, op->state);
call_remote_stonith(op, NULL);
} else {
crm_debug("Query %s targeting %s timed out (state=%d)",
op->id, op->target, op->state);
if (op->op_timer_total) {
g_source_remove(op->op_timer_total);
op->op_timer_total = 0;
}
remote_op_timeout(op);
}
return FALSE;
}
static gboolean
topology_is_empty(stonith_topology_t *tp)
{
int i;
if (tp == NULL) {
return TRUE;
}
for (i = 0; i < ST_LEVEL_MAX; i++) {
if (tp->levels[i] != NULL) {
return FALSE;
}
}
return TRUE;
}
/*!
* \internal
* \brief Add a device to an operation's automatic unfencing list
*
* \param[in,out] op Operation to modify
* \param[in] device Device ID to add
*/
static void
add_required_device(remote_fencing_op_t *op, const char *device)
{
GListPtr match = g_list_find_custom(op->automatic_list, device,
sort_strings);
if (!match) {
op->automatic_list = g_list_prepend(op->automatic_list, strdup(device));
}
}
/*!
* \internal
* \brief Remove a device from the automatic unfencing list
*
* \param[in,out] op Operation to modify
* \param[in] device Device ID to remove
*/
static void
remove_required_device(remote_fencing_op_t *op, const char *device)
{
GListPtr match = g_list_find_custom(op->automatic_list, device,
sort_strings);
if (match) {
op->automatic_list = g_list_remove(op->automatic_list, match->data);
}
}
/* deep copy the device list */
static void
set_op_device_list(remote_fencing_op_t * op, GListPtr devices)
{
GListPtr lpc = NULL;
if (op->devices_list) {
g_list_free_full(op->devices_list, free);
op->devices_list = NULL;
}
for (lpc = devices; lpc != NULL; lpc = lpc->next) {
op->devices_list = g_list_append(op->devices_list, strdup(lpc->data));
}
op->devices = op->devices_list;
}
/*!
* \internal
* \brief Check whether a node matches a topology target
*
* \param[in] tp Topology table entry to check
* \param[in] node Name of node to check
*
* \return TRUE if node matches topology target
*/
static gboolean
topology_matches(const stonith_topology_t *tp, const char *node)
{
regex_t r_patt;
CRM_CHECK(node && tp && tp->target, return FALSE);
switch(tp->kind) {
case 2:
/* This level targets by attribute, so tp->target is a NAME=VALUE pair
* of a permanent attribute applied to targeted nodes. The test below
* relies on the locally cached copy of the CIB, so if fencing needs to
* be done before the initial CIB is received or after a malformed CIB
* is received, then the topology will be unable to be used.
*/
if (node_has_attr(node, tp->target_attribute, tp->target_value)) {
crm_notice("Matched %s with %s by attribute", node, tp->target);
return TRUE;
}
break;
case 1:
/* This level targets by name, so tp->target is a regular expression
* matching names of nodes to be targeted.
*/
if (regcomp(&r_patt, tp->target_pattern, REG_EXTENDED|REG_NOSUB)) {
crm_info("Bad regex '%s' for fencing level", tp->target);
} else {
int status = regexec(&r_patt, node, 0, NULL, 0);
regfree(&r_patt);
if (status == 0) {
crm_notice("Matched %s with %s by name", node, tp->target);
return TRUE;
}
}
break;
case 0:
crm_trace("Testing %s against %s", node, tp->target);
return safe_str_eq(tp->target, node);
}
crm_trace("No match for %s with %s", node, tp->target);
return FALSE;
}
stonith_topology_t *
find_topology_for_host(const char *host)
{
GHashTableIter tIter;
stonith_topology_t *tp = g_hash_table_lookup(topology, host);
if(tp != NULL) {
crm_trace("Found %s for %s in %d entries", tp->target, host, g_hash_table_size(topology));
return tp;
}
g_hash_table_iter_init(&tIter, topology);
while (g_hash_table_iter_next(&tIter, NULL, (gpointer *) & tp)) {
if (topology_matches(tp, host)) {
crm_trace("Found %s for %s in %d entries", tp->target, host, g_hash_table_size(topology));
return tp;
}
}
crm_trace("No matches for %s in %d topology entries", host, g_hash_table_size(topology));
return NULL;
}
/*!
* \internal
* \brief Set fencing operation's device list to target's next topology level
*
* \param[in,out] op Remote fencing operation to modify
*
* \return pcmk_ok if successful, target was not specified (i.e. queries) or
* target has no topology, or -EINVAL if no more topology levels to try
*/
static int
stonith_topology_next(remote_fencing_op_t * op)
{
stonith_topology_t *tp = NULL;
if (op->target) {
/* Queries don't have a target set */
tp = find_topology_for_host(op->target);
}
if (topology_is_empty(tp)) {
return pcmk_ok;
}
set_bit(op->call_options, st_opt_topology);
/* This is a new level, so undo any remapping left over from previous */
undo_op_remap(op);
do {
op->level++;
} while (op->level < ST_LEVEL_MAX && tp->levels[op->level] == NULL);
if (op->level < ST_LEVEL_MAX) {
crm_trace("Attempting fencing level %d targeting %s (%d devices) "
"for client %s@%s.%.8s",
op->level, op->target, g_list_length(tp->levels[op->level]),
op->client_name, op->originator, op->id);
set_op_device_list(op, tp->levels[op->level]);
+ // The enforced delay has been applied for the first fencing level
+ if (op->level > 1 && op->delay > 0) {
+ op->delay = 0;
+ }
+
if (g_list_next(op->devices_list) && safe_str_eq(op->action, "reboot")) {
/* A reboot has been requested for a topology level with multiple
* devices. Instead of rebooting the devices sequentially, we will
* turn them all off, then turn them all on again. (Think about
* switched power outlets for redundant power supplies.)
*/
op_phase_off(op);
}
return pcmk_ok;
}
crm_notice("All fencing options targeting %s for client %s@%s.%.8s failed",
op->target, op->client_name, op->originator, op->id);
return -EINVAL;
}
/*!
* \brief Check to see if this operation is a duplicate of another in flight
* operation. If so merge this operation into the inflight operation, and mark
* it as a duplicate.
*/
static void
merge_duplicates(remote_fencing_op_t * op)
{
GHashTableIter iter;
remote_fencing_op_t *other = NULL;
time_t now = time(NULL);
g_hash_table_iter_init(&iter, stonith_remote_op_list);
while (g_hash_table_iter_next(&iter, NULL, (void **)&other)) {
crm_node_t *peer = NULL;
const char *other_action = op_requested_action(other);
if (other->state > st_exec) {
/* Must be in-progress */
continue;
} else if (safe_str_neq(op->target, other->target)) {
/* Must be for the same node */
continue;
} else if (safe_str_neq(op->action, other_action)) {
crm_trace("Must be for the same action: %s vs. %s",
op->action, other_action);
continue;
} else if (safe_str_eq(op->client_name, other->client_name)) {
crm_trace("Must be for different clients: %s", op->client_name);
continue;
} else if (safe_str_eq(other->target, other->originator)) {
crm_trace("Can't be a suicide operation: %s", other->target);
continue;
}
peer = crm_get_peer(0, other->originator);
if(fencing_peer_active(peer) == FALSE) {
crm_notice("Failing action '%s' targeting %s originating from "
"client %s@%s.%.8s: Originator is dead",
other->action, other->target, other->client_name, other->originator, other->id);
other->state = st_failed;
continue;
} else if(other->total_timeout > 0 && now > (other->total_timeout + other->created)) {
crm_info("Action '%s' targeting %s originating from client "
"%s@%s.%.8s is too old: %ld vs. %ld + %d",
other->action, other->target, other->client_name, other->originator, other->id,
now, other->created, other->total_timeout);
continue;
}
/* There is another in-flight request to fence the same host
* Piggyback on that instead. If it fails, so do we.
*/
other->duplicates = g_list_append(other->duplicates, op);
if (other->total_timeout == 0) {
crm_trace("Making a best-guess as to the timeout used");
other->total_timeout = op->total_timeout =
TIMEOUT_MULTIPLY_FACTOR * get_op_total_timeout(op, NULL);
}
crm_notice("Merging stonith action '%s' targeting %s originating from "
"client %s.%.8s with identical request from %s@%s.%.8s (%ds)",
op->action, op->target, op->client_name, op->id,
other->client_name, other->originator, other->id,
other->total_timeout);
report_timeout_period(op, other->total_timeout);
op->state = st_duplicate;
}
}
static uint32_t fencing_active_peers(void)
{
uint32_t count = 0;
crm_node_t *entry;
GHashTableIter gIter;
g_hash_table_iter_init(&gIter, crm_peer_cache);
while (g_hash_table_iter_next(&gIter, NULL, (void **)&entry)) {
if(fencing_peer_active(entry)) {
count++;
}
}
return count;
}
int
stonith_manual_ack(xmlNode * msg, remote_fencing_op_t * op)
{
xmlNode *dev = get_xpath_object("//@" F_STONITH_TARGET, msg, LOG_ERR);
op->state = st_done;
op->completed = time(NULL);
op->delegate = strdup("a human");
crm_notice("Injecting manual confirmation that %s is safely off/down",
crm_element_value(dev, F_STONITH_TARGET));
remote_op_done(op, msg, pcmk_ok, FALSE);
/* Replies are sent via done_cb->stonith_send_async_reply()->do_local_reply() */
return -EINPROGRESS;
}
/*!
* \internal
* \brief Create a new remote stonith operation
*
* \param[in] client ID of local stonith client that initiated the operation
* \param[in] request The request from the client that started the operation
* \param[in] peer TRUE if this operation is owned by another stonith peer
* (an operation owned by one peer is stored on all peers,
* but only the owner executes it; all nodes get the results
* once the owner finishes execution)
*/
void *
create_remote_stonith_op(const char *client, xmlNode * request, gboolean peer)
{
remote_fencing_op_t *op = NULL;
xmlNode *dev = get_xpath_object("//@" F_STONITH_TARGET, request, LOG_NEVER);
int call_options = 0;
init_stonith_remote_op_hash_table(&stonith_remote_op_list);
/* If this operation is owned by another node, check to make
* sure we haven't already created this operation. */
if (peer && dev) {
const char *op_id = crm_element_value(dev, F_STONITH_REMOTE_OP_ID);
CRM_CHECK(op_id != NULL, return NULL);
op = g_hash_table_lookup(stonith_remote_op_list, op_id);
if (op) {
crm_debug("%s already exists", op_id);
return op;
}
}
op = calloc(1, sizeof(remote_fencing_op_t));
crm_element_value_int(request, F_STONITH_TIMEOUT, &(op->base_timeout));
+ // Default value -1 means no enforced fencing delay
+ op->delay = -1;
+ crm_element_value_int(request, F_STONITH_DELAY, &(op->delay));
+
if (peer && dev) {
op->id = crm_element_value_copy(dev, F_STONITH_REMOTE_OP_ID);
} else {
op->id = crm_generate_uuid();
}
g_hash_table_replace(stonith_remote_op_list, op->id, op);
CRM_LOG_ASSERT(g_hash_table_lookup(stonith_remote_op_list, op->id) != NULL);
crm_trace("Created %s", op->id);
op->state = st_query;
op->replies_expected = fencing_active_peers();
op->action = crm_element_value_copy(dev, F_STONITH_ACTION);
op->originator = crm_element_value_copy(dev, F_STONITH_ORIGIN);
op->delegate = crm_element_value_copy(dev, F_STONITH_DELEGATE); /* May not be set */
op->created = time(NULL);
if (op->originator == NULL) {
/* Local or relayed request */
op->originator = strdup(stonith_our_uname);
}
CRM_LOG_ASSERT(client != NULL);
if (client) {
op->client_id = strdup(client);
}
op->client_name = crm_element_value_copy(request, F_STONITH_CLIENTNAME);
op->target = crm_element_value_copy(dev, F_STONITH_TARGET);
op->request = copy_xml(request); /* TODO: Figure out how to avoid this */
crm_element_value_int(request, F_STONITH_CALLOPTS, &call_options);
op->call_options = call_options;
crm_element_value_int(request, F_STONITH_CALLID, &(op->client_callid));
crm_trace("%s new stonith op %s ('%s' targeting %s for client %s)",
(peer && dev)? "Recorded" : "Generated", op->id, op->action,
op->target, op->client_name);
if (op->call_options & st_opt_cs_nodeid) {
int nodeid = crm_atoi(op->target, NULL);
crm_node_t *node = crm_find_known_peer_full(nodeid, NULL, CRM_GET_PEER_ANY);
/* Ensure the conversion only happens once */
op->call_options &= ~st_opt_cs_nodeid;
if (node && node->uname) {
free(op->target);
op->target = strdup(node->uname);
} else {
crm_warn("Could not expand nodeid '%s' into a host name", op->target);
}
}
/* check to see if this is a duplicate operation of another in-flight operation */
merge_duplicates(op);
if (op->state != st_duplicate) {
/* kick history readers */
do_stonith_notify(0, T_STONITH_NOTIFY_HISTORY, 0, NULL);
}
/* safe to trim as long as that doesn't touch pending ops */
stonith_fence_history_trim();
return op;
}
remote_fencing_op_t *
initiate_remote_stonith_op(pcmk__client_t *client, xmlNode *request,
gboolean manual_ack)
{
int query_timeout = 0;
xmlNode *query = NULL;
const char *client_id = NULL;
remote_fencing_op_t *op = NULL;
if (client) {
client_id = client->id;
} else {
client_id = crm_element_value(request, F_STONITH_CLIENTID);
}
CRM_LOG_ASSERT(client_id != NULL);
op = create_remote_stonith_op(client_id, request, FALSE);
op->owner = TRUE;
if (manual_ack) {
crm_notice("Initiating manual confirmation for %s: %s",
op->target, op->id);
return op;
}
CRM_CHECK(op->action, return NULL);
if (stonith_topology_next(op) != pcmk_ok) {
op->state = st_failed;
}
switch (op->state) {
case st_failed:
crm_warn("Could not request peer fencing (%s) targeting %s "
CRM_XS " id=%s", op->action, op->target, op->id);
remote_op_done(op, NULL, -EINVAL, FALSE);
return op;
case st_duplicate:
crm_info("Requesting peer fencing (%s) targeting %s (duplicate) "
CRM_XS " id=%s", op->action, op->target, op->id);
return op;
default:
crm_notice("Requesting peer fencing (%s) targeting %s "
CRM_XS " id=%s state=%d",
op->action, op->target, op->id, op->state);
}
query = stonith_create_op(op->client_callid, op->id, STONITH_OP_QUERY,
NULL, op->call_options);
crm_xml_add(query, F_STONITH_REMOTE_OP_ID, op->id);
crm_xml_add(query, F_STONITH_TARGET, op->target);
crm_xml_add(query, F_STONITH_ACTION, op_requested_action(op));
crm_xml_add(query, F_STONITH_ORIGIN, op->originator);
crm_xml_add(query, F_STONITH_CLIENTID, op->client_id);
crm_xml_add(query, F_STONITH_CLIENTNAME, op->client_name);
crm_xml_add_int(query, F_STONITH_TIMEOUT, op->base_timeout);
send_cluster_message(NULL, crm_msg_stonith_ng, query, FALSE);
free_xml(query);
query_timeout = op->base_timeout * TIMEOUT_MULTIPLY_FACTOR;
op->query_timer = g_timeout_add((1000 * query_timeout), remote_op_query_timeout, op);
return op;
}
enum find_best_peer_options {
/*! Skip checking the target peer for capable fencing devices */
FIND_PEER_SKIP_TARGET = 0x0001,
/*! Only check the target peer for capable fencing devices */
FIND_PEER_TARGET_ONLY = 0x0002,
/*! Skip peers and devices that are not verified */
FIND_PEER_VERIFIED_ONLY = 0x0004,
};
static st_query_result_t *
find_best_peer(const char *device, remote_fencing_op_t * op, enum find_best_peer_options options)
{
GListPtr iter = NULL;
gboolean verified_devices_only = (options & FIND_PEER_VERIFIED_ONLY) ? TRUE : FALSE;
if (!device && is_set(op->call_options, st_opt_topology)) {
return NULL;
}
for (iter = op->query_results; iter != NULL; iter = iter->next) {
st_query_result_t *peer = iter->data;
crm_trace("Testing result from %s targeting %s with %d devices: %d %x",
peer->host, op->target, peer->ndevices, peer->tried, options);
if ((options & FIND_PEER_SKIP_TARGET) && safe_str_eq(peer->host, op->target)) {
continue;
}
if ((options & FIND_PEER_TARGET_ONLY) && safe_str_neq(peer->host, op->target)) {
continue;
}
if (is_set(op->call_options, st_opt_topology)) {
if (grab_peer_device(op, peer, device, verified_devices_only)) {
return peer;
}
} else if ((peer->tried == FALSE)
&& count_peer_devices(op, peer, verified_devices_only)) {
/* No topology: Use the current best peer */
crm_trace("Simple fencing");
return peer;
}
}
return NULL;
}
static st_query_result_t *
stonith_choose_peer(remote_fencing_op_t * op)
{
const char *device = NULL;
st_query_result_t *peer = NULL;
uint32_t active = fencing_active_peers();
do {
if (op->devices) {
device = op->devices->data;
crm_trace("Checking for someone to fence (%s) %s with %s",
op->action, op->target, device);
} else {
crm_trace("Checking for someone to fence (%s) %s",
op->action, op->target);
}
/* Best choice is a peer other than the target with verified access */
peer = find_best_peer(device, op, FIND_PEER_SKIP_TARGET|FIND_PEER_VERIFIED_ONLY);
if (peer) {
crm_trace("Found verified peer %s for %s", peer->host, device?device:"");
return peer;
}
if(op->query_timer != 0 && op->replies < QB_MIN(op->replies_expected, active)) {
crm_trace("Waiting before looking for unverified devices to fence %s", op->target);
return NULL;
}
/* If no other peer has verified access, next best is unverified access */
peer = find_best_peer(device, op, FIND_PEER_SKIP_TARGET);
if (peer) {
crm_trace("Found best unverified peer %s", peer->host);
return peer;
}
/* If no other peer can do it, last option is self-fencing
* (which is never allowed for the "on" phase of a remapped reboot)
*/
if (op->phase != st_phase_on) {
peer = find_best_peer(device, op, FIND_PEER_TARGET_ONLY);
if (peer) {
crm_trace("%s will fence itself", peer->host);
return peer;
}
}
/* Try the next fencing level if there is one (unless we're in the "on"
* phase of a remapped "reboot", because we ignore errors in that case)
*/
} while ((op->phase != st_phase_on)
&& is_set(op->call_options, st_opt_topology)
&& stonith_topology_next(op) == pcmk_ok);
crm_notice("Couldn't find anyone to fence (%s) %s with %s",
op->action, op->target, (device? device : "any device"));
return NULL;
}
static int
get_device_timeout(const remote_fencing_op_t *op, const st_query_result_t *peer,
const char *device)
{
device_properties_t *props;
if (!peer || !device) {
return op->base_timeout;
}
props = g_hash_table_lookup(peer->devices, device);
if (!props) {
return op->base_timeout;
}
return (props->custom_action_timeout[op->phase]?
props->custom_action_timeout[op->phase] : op->base_timeout)
+ props->delay_max[op->phase];
}
struct timeout_data {
const remote_fencing_op_t *op;
const st_query_result_t *peer;
int total_timeout;
};
/*!
* \internal
* \brief Add timeout to a total if device has not been executed yet
*
* \param[in] key GHashTable key (device ID)
* \param[in] value GHashTable value (device properties)
* \param[in] user_data Timeout data
*/
static void
add_device_timeout(gpointer key, gpointer value, gpointer user_data)
{
const char *device_id = key;
device_properties_t *props = value;
struct timeout_data *timeout = user_data;
if (!props->executed[timeout->op->phase]
&& !props->disallowed[timeout->op->phase]) {
timeout->total_timeout += get_device_timeout(timeout->op,
timeout->peer, device_id);
}
}
static int
get_peer_timeout(const remote_fencing_op_t *op, const st_query_result_t *peer)
{
struct timeout_data timeout;
timeout.op = op;
timeout.peer = peer;
timeout.total_timeout = 0;
g_hash_table_foreach(peer->devices, add_device_timeout, &timeout);
return (timeout.total_timeout? timeout.total_timeout : op->base_timeout);
}
static int
get_op_total_timeout(const remote_fencing_op_t *op,
const st_query_result_t *chosen_peer)
{
int total_timeout = 0;
stonith_topology_t *tp = find_topology_for_host(op->target);
if (is_set(op->call_options, st_opt_topology) && tp) {
int i;
GListPtr device_list = NULL;
GListPtr iter = NULL;
/* Yep, this looks scary, nested loops all over the place.
* Here is what is going on.
* Loop1: Iterate through fencing levels.
* Loop2: If a fencing level has devices, loop through each device
* Loop3: For each device in a fencing level, see what peer owns it
* and what that peer has reported the timeout is for the device.
*/
for (i = 0; i < ST_LEVEL_MAX; i++) {
if (!tp->levels[i]) {
continue;
}
for (device_list = tp->levels[i]; device_list; device_list = device_list->next) {
for (iter = op->query_results; iter != NULL; iter = iter->next) {
const st_query_result_t *peer = iter->data;
if (find_peer_device(op, peer, device_list->data)) {
total_timeout += get_device_timeout(op, peer,
device_list->data);
break;
}
} /* End Loop3: match device with peer that owns device, find device's timeout period */
} /* End Loop2: iterate through devices at a specific level */
} /*End Loop1: iterate through fencing levels */
} else if (chosen_peer) {
total_timeout = get_peer_timeout(op, chosen_peer);
} else {
total_timeout = op->base_timeout;
}
return total_timeout ? total_timeout : op->base_timeout;
}
static void
report_timeout_period(remote_fencing_op_t * op, int op_timeout)
{
GListPtr iter = NULL;
xmlNode *update = NULL;
const char *client_node = NULL;
const char *client_id = NULL;
const char *call_id = NULL;
if (op->call_options & st_opt_sync_call) {
/* There is no reason to report the timeout for a synchronous call. It
* is impossible to use the reported timeout to do anything when the client
* is blocking for the response. This update is only important for
* async calls that require a callback to report the results in. */
return;
} else if (!op->request) {
return;
}
crm_trace("Reporting timeout for %s.%.8s", op->client_name, op->id);
client_node = crm_element_value(op->request, F_STONITH_CLIENTNODE);
call_id = crm_element_value(op->request, F_STONITH_CALLID);
client_id = crm_element_value(op->request, F_STONITH_CLIENTID);
if (!client_node || !call_id || !client_id) {
return;
}
if (safe_str_eq(client_node, stonith_our_uname)) {
/* The client is connected to this node, send the update direclty to them */
do_stonith_async_timeout_update(client_id, call_id, op_timeout);
return;
}
/* The client is connected to another node, relay this update to them */
update = stonith_create_op(op->client_callid, op->id, STONITH_OP_TIMEOUT_UPDATE, NULL, 0);
crm_xml_add(update, F_STONITH_REMOTE_OP_ID, op->id);
crm_xml_add(update, F_STONITH_CLIENTID, client_id);
crm_xml_add(update, F_STONITH_CALLID, call_id);
crm_xml_add_int(update, F_STONITH_TIMEOUT, op_timeout);
send_cluster_message(crm_get_peer(0, client_node), crm_msg_stonith_ng, update, FALSE);
free_xml(update);
for (iter = op->duplicates; iter != NULL; iter = iter->next) {
remote_fencing_op_t *dup = iter->data;
crm_trace("Reporting timeout for duplicate %s.%.8s", dup->client_name, dup->id);
report_timeout_period(iter->data, op_timeout);
}
}
/*!
* \internal
* \brief Advance an operation to the next device in its topology
*
* \param[in,out] op Operation to advance
* \param[in] device ID of device just completed
* \param[in] msg XML reply that contained device result (if available)
* \param[in] rc Return code of device's execution
*/
static void
advance_op_topology(remote_fencing_op_t *op, const char *device, xmlNode *msg,
int rc)
{
/* Advance to the next device at this topology level, if any */
if (op->devices) {
op->devices = op->devices->next;
}
/* Handle automatic unfencing if an "on" action was requested */
if ((op->phase == st_phase_requested) && safe_str_eq(op->action, "on")) {
/* If the device we just executed was required, it's not anymore */
remove_required_device(op, device);
/* If there are no more devices at this topology level, run through any
* remaining devices with automatic unfencing
*/
if (op->devices == NULL) {
op->devices = op->automatic_list;
}
}
if ((op->devices == NULL) && (op->phase == st_phase_off)) {
/* We're done with this level and with required devices, but we had
* remapped "reboot" to "off", so start over with "on". If any devices
* need to be turned back on, op->devices will be non-NULL after this.
*/
op_phase_on(op);
}
if (op->devices) {
/* Necessary devices remain, so execute the next one */
crm_trace("Next targeting %s on behalf of %s@%s (rc was %d)",
op->target, op->originator, op->client_name, rc);
+
+ // The enforced delay has been applied for the first device
+ if (op->delay > 0) {
+ op->delay = 0;
+ }
+
call_remote_stonith(op, NULL);
} else {
/* We're done with all devices and phases, so finalize operation */
crm_trace("Marking complex fencing op targeting %s as complete",
op->target);
op->state = st_done;
remote_op_done(op, msg, rc, FALSE);
}
}
void
call_remote_stonith(remote_fencing_op_t * op, st_query_result_t * peer)
{
const char *device = NULL;
int timeout = op->base_timeout;
crm_trace("State for %s.%.8s: %s %d", op->target, op->client_name, op->id, op->state);
if (peer == NULL && !is_set(op->call_options, st_opt_topology)) {
peer = stonith_choose_peer(op);
}
if (!op->op_timer_total) {
int total_timeout = get_op_total_timeout(op, peer);
op->total_timeout = TIMEOUT_MULTIPLY_FACTOR * total_timeout;
op->op_timer_total = g_timeout_add(1000 * op->total_timeout, remote_op_timeout, op);
report_timeout_period(op, op->total_timeout);
crm_info("Total timeout set to %d for peer's fencing targeting %s for %s"
CRM_XS "id=%s",
total_timeout, op->target, op->client_name, op->id);
}
if (is_set(op->call_options, st_opt_topology) && op->devices) {
/* Ignore any peer preference, they might not have the device we need */
/* When using topology, stonith_choose_peer() removes the device from
* further consideration, so be sure to calculate timeout beforehand */
peer = stonith_choose_peer(op);
device = op->devices->data;
timeout = get_device_timeout(op, peer, device);
}
if (peer) {
int timeout_one = 0;
xmlNode *remote_op = stonith_create_op(op->client_callid, op->id, STONITH_OP_FENCE, NULL, 0);
crm_xml_add(remote_op, F_STONITH_REMOTE_OP_ID, op->id);
crm_xml_add(remote_op, F_STONITH_TARGET, op->target);
crm_xml_add(remote_op, F_STONITH_ACTION, op->action);
crm_xml_add(remote_op, F_STONITH_ORIGIN, op->originator);
crm_xml_add(remote_op, F_STONITH_CLIENTID, op->client_id);
crm_xml_add(remote_op, F_STONITH_CLIENTNAME, op->client_name);
crm_xml_add_int(remote_op, F_STONITH_TIMEOUT, timeout);
crm_xml_add_int(remote_op, F_STONITH_CALLOPTS, op->call_options);
+ if (op->delay >= 0) {
+ crm_xml_add_int(remote_op, F_STONITH_DELAY, op->delay);
+ }
+
if (device) {
timeout_one = TIMEOUT_MULTIPLY_FACTOR *
get_device_timeout(op, peer, device);
crm_notice("Requesting that %s perform '%s' action targeting %s "
"using '%s' " CRM_XS " for client %s (%ds)",
peer->host, op->action, op->target, device,
op->client_name, timeout_one);
crm_xml_add(remote_op, F_STONITH_DEVICE, device);
crm_xml_add(remote_op, F_STONITH_MODE, "slave");
} else {
timeout_one = TIMEOUT_MULTIPLY_FACTOR * get_peer_timeout(op, peer);
crm_notice("Requesting that %s perform '%s' action targeting %s "
CRM_XS " for client %s (%ds, %lds)",
peer->host, op->action, op->target, op->client_name,
timeout_one, stonith_watchdog_timeout_ms);
crm_xml_add(remote_op, F_STONITH_MODE, "smart");
}
op->state = st_exec;
if (op->op_timer_one) {
g_source_remove(op->op_timer_one);
}
if(stonith_watchdog_timeout_ms > 0 && device && safe_str_eq(device, "watchdog")) {
crm_notice("Waiting %lds for %s to self-fence (%s) for client %s.%.8s",
stonith_watchdog_timeout_ms/1000, op->target, op->action,
op->client_name, op->id);
op->op_timer_one = g_timeout_add(stonith_watchdog_timeout_ms, remote_op_watchdog_done, op);
/* TODO check devices to verify watchdog will be in use */
} else if(stonith_watchdog_timeout_ms > 0
&& safe_str_eq(peer->host, op->target)
&& safe_str_neq(op->action, "on")) {
crm_notice("Waiting %lds for %s to self-fence (%s) for client %s.%.8s",
stonith_watchdog_timeout_ms/1000, op->target, op->action,
op->client_name, op->id);
op->op_timer_one = g_timeout_add(stonith_watchdog_timeout_ms, remote_op_watchdog_done, op);
} else {
op->op_timer_one = g_timeout_add((1000 * timeout_one), remote_op_timeout_one, op);
}
send_cluster_message(crm_get_peer(0, peer->host), crm_msg_stonith_ng, remote_op, FALSE);
peer->tried = TRUE;
free_xml(remote_op);
return;
} else if (op->phase == st_phase_on) {
/* A remapped "on" cannot be executed, but the node was already
* turned off successfully, so ignore the error and continue.
*/
crm_warn("Ignoring %s 'on' failure (no capable peers) targeting %s "
"after successful 'off'", device, op->target);
advance_op_topology(op, device, NULL, pcmk_ok);
return;
} else if (op->owner == FALSE) {
crm_err("Fencing (%s) targeting %s for client %s is not ours to control",
op->action, op->target, op->client_name);
} else if (op->query_timer == 0) {
/* We've exhausted all available peers */
crm_info("No remaining peers capable of fencing (%s) %s for client %s "
CRM_XS " state=%d",
op->action, op->target, op->client_name, op->state);
CRM_LOG_ASSERT(op->state < st_done);
remote_op_timeout(op);
} else if(op->replies >= op->replies_expected || op->replies >= fencing_active_peers()) {
int rc = -EHOSTUNREACH;
/* if the operation never left the query state,
* but we have all the expected replies, then no devices
* are available to execute the fencing operation. */
if(stonith_watchdog_timeout_ms && (device == NULL || safe_str_eq(device, "watchdog"))) {
crm_notice("Waiting %lds for %s to self-fence (%s) for client %s.%.8s",
stonith_watchdog_timeout_ms/1000, op->target,
op->action, op->client_name, op->id);
op->op_timer_one = g_timeout_add(stonith_watchdog_timeout_ms, remote_op_watchdog_done, op);
return;
}
if (op->state == st_query) {
crm_info("No peers (out of %d) have devices capable of fencing "
"(%s) %s for client %s " CRM_XS " state=%d",
op->replies, op->action, op->target, op->client_name,
op->state);
rc = -ENODEV;
} else {
crm_info("No peers (out of %d) are capable of fencing (%s) %s "
"for client %s " CRM_XS " state=%d",
op->replies, op->action, op->target, op->client_name,
op->state);
}
op->state = st_failed;
remote_op_done(op, NULL, rc, FALSE);
} else {
crm_info("Waiting for additional peers capable of fencing (%s) %s%s%s "
"for client %s%.8s",
op->action, op->target, (device? " with " : ""),
(device? device : ""), op->client_name, op->id);
}
}
/*!
* \internal
* \brief Comparison function for sorting query results
*
* \param[in] a GList item to compare
* \param[in] b GList item to compare
*
* \return Per the glib documentation, "a negative integer if the first value
* comes before the second, 0 if they are equal, or a positive integer
* if the first value comes after the second."
*/
static gint
sort_peers(gconstpointer a, gconstpointer b)
{
const st_query_result_t *peer_a = a;
const st_query_result_t *peer_b = b;
return (peer_b->ndevices - peer_a->ndevices);
}
/*!
* \internal
* \brief Determine if all the devices in the topology are found or not
*/
static gboolean
all_topology_devices_found(remote_fencing_op_t * op)
{
GListPtr device = NULL;
GListPtr iter = NULL;
device_properties_t *match = NULL;
stonith_topology_t *tp = NULL;
gboolean skip_target = FALSE;
int i;
tp = find_topology_for_host(op->target);
if (!tp) {
return FALSE;
}
if (safe_str_eq(op->action, "off") || safe_str_eq(op->action, "reboot")) {
/* Don't count the devices on the target node if we are killing
* the target node. */
skip_target = TRUE;
}
for (i = 0; i < ST_LEVEL_MAX; i++) {
for (device = tp->levels[i]; device; device = device->next) {
match = NULL;
for (iter = op->query_results; iter && !match; iter = iter->next) {
st_query_result_t *peer = iter->data;
if (skip_target && safe_str_eq(peer->host, op->target)) {
continue;
}
match = find_peer_device(op, peer, device->data);
}
if (!match) {
return FALSE;
}
}
}
return TRUE;
}
/*!
* \internal
* \brief Parse action-specific device properties from XML
*
* \param[in] msg XML element containing the properties
* \param[in] peer Name of peer that sent XML (for logs)
* \param[in] device Device ID (for logs)
* \param[in] action Action the properties relate to (for logs)
* \param[in] phase Phase the properties relate to
* \param[in,out] props Device properties to update
*/
static void
parse_action_specific(xmlNode *xml, const char *peer, const char *device,
const char *action, remote_fencing_op_t *op,
enum st_remap_phase phase, device_properties_t *props)
{
props->custom_action_timeout[phase] = 0;
crm_element_value_int(xml, F_STONITH_ACTION_TIMEOUT,
&props->custom_action_timeout[phase]);
if (props->custom_action_timeout[phase]) {
crm_trace("Peer %s with device %s returned %s action timeout %d",
peer, device, action, props->custom_action_timeout[phase]);
}
props->delay_max[phase] = 0;
crm_element_value_int(xml, F_STONITH_DELAY_MAX, &props->delay_max[phase]);
if (props->delay_max[phase]) {
crm_trace("Peer %s with device %s returned maximum of random delay %d for %s",
peer, device, props->delay_max[phase], action);
}
props->delay_base[phase] = 0;
crm_element_value_int(xml, F_STONITH_DELAY_BASE, &props->delay_base[phase]);
if (props->delay_base[phase]) {
crm_trace("Peer %s with device %s returned base delay %d for %s",
peer, device, props->delay_base[phase], action);
}
/* Handle devices with automatic unfencing */
if (safe_str_eq(action, "on")) {
int required = 0;
crm_element_value_int(xml, F_STONITH_DEVICE_REQUIRED, &required);
if (required) {
crm_trace("Peer %s requires device %s to execute for action %s",
peer, device, action);
add_required_device(op, device);
}
}
/* If a reboot is remapped to off+on, it's possible that a node is allowed
* to perform one action but not another.
*/
if (crm_is_true(crm_element_value(xml, F_STONITH_ACTION_DISALLOWED))) {
props->disallowed[phase] = TRUE;
crm_trace("Peer %s is disallowed from executing %s for device %s",
peer, action, device);
}
}
/*!
* \internal
* \brief Parse one device's properties from peer's XML query reply
*
* \param[in] xml XML node containing device properties
* \param[in,out] op Operation that query and reply relate to
* \param[in,out] result Peer's results
* \param[in] device ID of device being parsed
*/
static void
add_device_properties(xmlNode *xml, remote_fencing_op_t *op,
st_query_result_t *result, const char *device)
{
xmlNode *child;
int verified = 0;
device_properties_t *props = calloc(1, sizeof(device_properties_t));
/* Add a new entry to this result's devices list */
CRM_ASSERT(props != NULL);
g_hash_table_insert(result->devices, strdup(device), props);
/* Peers with verified (monitored) access will be preferred */
crm_element_value_int(xml, F_STONITH_DEVICE_VERIFIED, &verified);
if (verified) {
crm_trace("Peer %s has confirmed a verified device %s",
result->host, device);
props->verified = TRUE;
}
/* Parse action-specific device properties */
parse_action_specific(xml, result->host, device, op_requested_action(op),
op, st_phase_requested, props);
for (child = __xml_first_child(xml); child != NULL; child = __xml_next(child)) {
/* Replies for "reboot" operations will include the action-specific
* values for "off" and "on" in child elements, just in case the reboot
* winds up getting remapped.
*/
if (safe_str_eq(ID(child), "off")) {
parse_action_specific(child, result->host, device, "off",
op, st_phase_off, props);
} else if (safe_str_eq(ID(child), "on")) {
parse_action_specific(child, result->host, device, "on",
op, st_phase_on, props);
}
}
}
/*!
* \internal
* \brief Parse a peer's XML query reply and add it to operation's results
*
* \param[in,out] op Operation that query and reply relate to
* \param[in] host Name of peer that sent this reply
* \param[in] ndevices Number of devices expected in reply
* \param[in] xml XML node containing device list
*
* \return Newly allocated result structure with parsed reply
*/
static st_query_result_t *
add_result(remote_fencing_op_t *op, const char *host, int ndevices, xmlNode *xml)
{
st_query_result_t *result = calloc(1, sizeof(st_query_result_t));
xmlNode *child;
CRM_CHECK(result != NULL, return NULL);
result->host = strdup(host);
result->devices = crm_str_table_new();
/* Each child element describes one capable device available to the peer */
for (child = __xml_first_child(xml); child != NULL; child = __xml_next(child)) {
const char *device = ID(child);
if (device) {
add_device_properties(child, op, result, device);
}
}
result->ndevices = g_hash_table_size(result->devices);
CRM_CHECK(ndevices == result->ndevices,
crm_err("Query claimed to have %d devices but %d found",
ndevices, result->ndevices));
op->query_results = g_list_insert_sorted(op->query_results, result, sort_peers);
return result;
}
/*!
* \internal
* \brief Handle a peer's reply to our fencing query
*
* Parse a query result from XML and store it in the remote operation
* table, and when enough replies have been received, issue a fencing request.
*
* \param[in] msg XML reply received
*
* \return pcmk_ok on success, -errno on error
*
* \note See initiate_remote_stonith_op() for how the XML query was initially
* formed, and stonith_query() for how the peer formed its XML reply.
*/
int
process_remote_stonith_query(xmlNode * msg)
{
int ndevices = 0;
gboolean host_is_target = FALSE;
gboolean have_all_replies = FALSE;
const char *id = NULL;
const char *host = NULL;
remote_fencing_op_t *op = NULL;
st_query_result_t *result = NULL;
uint32_t replies_expected;
xmlNode *dev = get_xpath_object("//@" F_STONITH_REMOTE_OP_ID, msg, LOG_ERR);
CRM_CHECK(dev != NULL, return -EPROTO);
id = crm_element_value(dev, F_STONITH_REMOTE_OP_ID);
CRM_CHECK(id != NULL, return -EPROTO);
dev = get_xpath_object("//@" F_STONITH_AVAILABLE_DEVICES, msg, LOG_ERR);
CRM_CHECK(dev != NULL, return -EPROTO);
crm_element_value_int(dev, F_STONITH_AVAILABLE_DEVICES, &ndevices);
op = g_hash_table_lookup(stonith_remote_op_list, id);
if (op == NULL) {
crm_debug("Received query reply for unknown or expired operation %s",
id);
return -EOPNOTSUPP;
}
replies_expected = fencing_active_peers();
if (op->replies_expected < replies_expected) {
replies_expected = op->replies_expected;
}
if ((++op->replies >= replies_expected) && (op->state == st_query)) {
have_all_replies = TRUE;
}
host = crm_element_value(msg, F_ORIG);
host_is_target = safe_str_eq(host, op->target);
crm_info("Query result %d of %d from %s for %s/%s (%d devices) %s",
op->replies, replies_expected, host,
op->target, op->action, ndevices, id);
if (ndevices > 0) {
result = add_result(op, host, ndevices, dev);
}
if (is_set(op->call_options, st_opt_topology)) {
/* If we start the fencing before all the topology results are in,
* it is possible fencing levels will be skipped because of the missing
* query results. */
if (op->state == st_query && all_topology_devices_found(op)) {
/* All the query results are in for the topology, start the fencing ops. */
crm_trace("All topology devices found");
call_remote_stonith(op, result);
} else if (have_all_replies) {
crm_info("All topology query replies have arrived, continuing (%d expected/%d received) ",
replies_expected, op->replies);
call_remote_stonith(op, NULL);
}
} else if (op->state == st_query) {
int nverified = count_peer_devices(op, result, TRUE);
/* We have a result for a non-topology fencing op that looks promising,
* go ahead and start fencing before query timeout */
if (result && (host_is_target == FALSE) && nverified) {
/* we have a verified device living on a peer that is not the target */
crm_trace("Found %d verified devices", nverified);
call_remote_stonith(op, result);
} else if (have_all_replies) {
crm_info("All query replies have arrived, continuing (%d expected/%d received) ",
replies_expected, op->replies);
call_remote_stonith(op, NULL);
} else {
crm_trace("Waiting for more peer results before launching fencing operation");
}
} else if (result && (op->state == st_done)) {
crm_info("Discarding query result from %s (%d devices): Operation is in state %d",
result->host, result->ndevices, op->state);
}
return pcmk_ok;
}
/*!
* \internal
* \brief Handle a peer's reply to a fencing request
*
* Parse a fencing reply from XML, and either finalize the operation
* or attempt another device as appropriate.
*
* \param[in] msg XML reply received
*
* \return pcmk_ok on success, -errno on error
*/
int
process_remote_stonith_exec(xmlNode * msg)
{
int rc = 0;
const char *id = NULL;
const char *device = NULL;
remote_fencing_op_t *op = NULL;
xmlNode *dev = get_xpath_object("//@" F_STONITH_REMOTE_OP_ID, msg, LOG_ERR);
CRM_CHECK(dev != NULL, return -EPROTO);
id = crm_element_value(dev, F_STONITH_REMOTE_OP_ID);
CRM_CHECK(id != NULL, return -EPROTO);
dev = get_xpath_object("//@" F_STONITH_RC, msg, LOG_ERR);
CRM_CHECK(dev != NULL, return -EPROTO);
crm_element_value_int(dev, F_STONITH_RC, &rc);
device = crm_element_value(dev, F_STONITH_DEVICE);
if (stonith_remote_op_list) {
op = g_hash_table_lookup(stonith_remote_op_list, id);
}
if (op == NULL && rc == pcmk_ok) {
/* Record successful fencing operations */
const char *client_id = crm_element_value(dev, F_STONITH_CLIENTID);
op = create_remote_stonith_op(client_id, dev, TRUE);
}
if (op == NULL) {
/* Could be for an event that began before we started */
/* TODO: Record the op for later querying */
crm_info("Received peer result of unknown or expired operation %s", id);
return -EOPNOTSUPP;
}
if (op->devices && device && safe_str_neq(op->devices->data, device)) {
crm_err("Received outdated reply for device %s (instead of %s) to "
"fence (%s) %s. Operation already timed out at peer level.",
device, (const char *) op->devices->data, op->action, op->target);
return rc;
}
if (safe_str_eq(crm_element_value(msg, F_SUBTYPE), "broadcast")) {
crm_debug("Marking call to %s for %s on behalf of %s@%s.%.8s: %s (%d)",
op->action, op->target, op->client_name, op->id, op->originator,
pcmk_strerror(rc), rc);
if (rc == pcmk_ok) {
op->state = st_done;
} else {
op->state = st_failed;
}
remote_op_done(op, msg, rc, FALSE);
return pcmk_ok;
} else if (safe_str_neq(op->originator, stonith_our_uname)) {
/* If this isn't a remote level broadcast, and we are not the
* originator of the operation, we should not be receiving this msg. */
crm_err
("%s received non-broadcast fencing result for operation it does not own (device %s targeting %s)",
stonith_our_uname, device, op->target);
return rc;
}
if (is_set(op->call_options, st_opt_topology)) {
const char *device = crm_element_value(msg, F_STONITH_DEVICE);
crm_notice("Action '%s' targeting %s using %s on behalf of %s@%s: %s "
CRM_XS " rc=%d",
op->action, op->target, device, op->client_name,
op->originator, pcmk_strerror(rc), rc);
/* We own the op, and it is complete. broadcast the result to all nodes
* and notify our local clients. */
if (op->state == st_done) {
remote_op_done(op, msg, rc, FALSE);
return rc;
}
if ((op->phase == 2) && (rc != pcmk_ok)) {
/* A remapped "on" failed, but the node was already turned off
* successfully, so ignore the error and continue.
*/
crm_warn("Ignoring %s 'on' failure (exit code %d) targeting %s "
"after successful 'off'", device, rc, op->target);
rc = pcmk_ok;
}
if (rc == pcmk_ok) {
/* An operation completed successfully. Try another device if
* necessary, otherwise mark the operation as done. */
advance_op_topology(op, device, msg, rc);
return rc;
} else {
/* This device failed, time to try another topology level. If no other
* levels are available, mark this operation as failed and report results. */
if (stonith_topology_next(op) != pcmk_ok) {
op->state = st_failed;
remote_op_done(op, msg, rc, FALSE);
return rc;
}
}
} else if (rc == pcmk_ok && op->devices == NULL) {
crm_trace("All done for %s", op->target);
op->state = st_done;
remote_op_done(op, msg, rc, FALSE);
return rc;
} else if (rc == -ETIME && op->devices == NULL) {
/* If the operation timed out don't bother retrying other peers. */
op->state = st_failed;
remote_op_done(op, msg, rc, FALSE);
return rc;
} else {
/* fall-through and attempt other fencing action using another peer */
}
/* Retry on failure */
crm_trace("Next for %s on behalf of %s@%s (rc was %d)", op->target, op->originator,
op->client_name, rc);
call_remote_stonith(op, NULL);
return rc;
}
gboolean
stonith_check_fence_tolerance(int tolerance, const char *target, const char *action)
{
GHashTableIter iter;
time_t now = time(NULL);
remote_fencing_op_t *rop = NULL;
crm_trace("tolerance=%d, stonith_remote_op_list=%p", tolerance,
stonith_remote_op_list);
if (tolerance <= 0 || !stonith_remote_op_list || target == NULL ||
action == NULL) {
return FALSE;
}
g_hash_table_iter_init(&iter, stonith_remote_op_list);
while (g_hash_table_iter_next(&iter, NULL, (void **)&rop)) {
if (strcmp(rop->target, target) != 0) {
continue;
} else if (rop->state != st_done) {
continue;
/* We don't have to worry about remapped reboots here
* because if state is done, any remapping has been undone
*/
} else if (strcmp(rop->action, action) != 0) {
continue;
} else if ((rop->completed + tolerance) < now) {
continue;
}
crm_notice("Target %s was fenced (%s) less than %ds ago by %s on behalf of %s",
target, action, tolerance, rop->delegate, rop->originator);
return TRUE;
}
return FALSE;
}
diff --git a/daemons/fenced/pacemaker-fenced.h b/daemons/fenced/pacemaker-fenced.h
index cc7f9105c5..73cd28e340 100644
--- a/daemons/fenced/pacemaker-fenced.h
+++ b/daemons/fenced/pacemaker-fenced.h
@@ -1,272 +1,276 @@
/*
* Copyright 2009-2020 the Pacemaker project contributors
*
* This source code is licensed under the GNU General Public License version 2
* or later (GPLv2+) WITHOUT ANY WARRANTY.
*/
#include
/*!
* \internal
* \brief Check to see if target was fenced in the last few seconds.
* \param tolerance, The number of seconds to look back in time
* \param target, The node to search for
* \param action, The action we want to match.
*
* \retval FALSE, not match
* \retval TRUE, fencing operation took place in the last 'tolerance' number of seconds.
*/
gboolean stonith_check_fence_tolerance(int tolerance, const char *target, const char *action);
enum st_device_flags
{
st_device_supports_list = 0x0001,
st_device_supports_status = 0x0002,
st_device_supports_reboot = 0x0004,
};
typedef struct stonith_device_s {
char *id;
char *agent;
char *namespace;
/*! list of actions that must execute on the target node. Used for unfencing */
char *on_target_actions;
GListPtr targets;
time_t targets_age;
gboolean has_attr_map;
/* should nodeid parameter for victim be included in agent arguments */
gboolean include_nodeid;
/* whether the cluster should automatically unfence nodes with the device */
gboolean automatic_unfencing;
guint priority;
enum st_device_flags flags;
GHashTable *params;
GHashTable *aliases;
GList *pending_ops;
crm_trigger_t *work;
xmlNode *agent_metadata;
/*! A verified device is one that has contacted the
* agent successfully to perform a monitor operation */
gboolean verified;
gboolean cib_registered;
gboolean api_registered;
} stonith_device_t;
/* These values are used to index certain arrays by "phase". Usually an
* operation has only one "phase", so phase is always zero. However, some
* reboots are remapped to "off" then "on", in which case "reboot" will be
* phase 0, "off" will be phase 1 and "on" will be phase 2.
*/
enum st_remap_phase {
st_phase_requested = 0,
st_phase_off = 1,
st_phase_on = 2,
st_phase_max = 3
};
/* These values provide additional information for STONITH's asynchronous reply response.
* The st_reply_opt_merged value indicates an operation that has been merged and completed without being executed.
*/
enum st_replay_option {
st_reply_opt_none = 0x00000000,
st_reply_opt_merged = 0x00000001,
};
typedef struct remote_fencing_op_s {
/* The unique id associated with this operation */
char *id;
/*! The node this operation will fence */
char *target;
/*! The fencing action to perform on the target. (reboot, on, off) */
char *action;
/*! When was the fencing action recorded (seconds since epoch) */
time_t created;
/*! Marks if the final notifications have been sent to local stonith clients. */
gboolean notify_sent;
/*! The number of query replies received */
guint replies;
/*! The number of query replies expected */
guint replies_expected;
/*! Does this node own control of this operation */
gboolean owner;
/*! After query is complete, This the high level timer that expires the entire operation */
guint op_timer_total;
/*! This timer expires the current fencing request. Many fencing
* requests may exist in a single operation */
guint op_timer_one;
/*! This timer expires the query request sent out to determine
* what nodes are contain what devices, and who those devices can fence */
guint query_timer;
/*! This is the default timeout to use for each fencing device if no
* custom timeout is received in the query. */
gint base_timeout;
/*! This is the calculated total timeout an operation can take before
* expiring. This is calculated by adding together all the timeout
* values associated with the devices this fencing operation may call */
gint total_timeout;
+ /*! Enforced fencing delay.
+ * Default value -1 means no enforced fencing delay. */
+ int delay;
+
/*! Delegate is the node being asked to perform a fencing action
* on behalf of the node that owns the remote operation. Some operations
* will involve multiple delegates. This value represents the final delegate
* that is used. */
char *delegate;
/*! The point at which the remote operation completed */
time_t completed;
/*! The stonith_call_options associated with this remote operation */
long long call_options;
/*! The current state of the remote operation. This indicates
* what stage the op is in, query, exec, done, duplicate, failed. */
enum op_state state;
/*! The node that owns the remote operation */
char *originator;
/*! The local client id that initiated the fencing request */
char *client_id;
/*! The client's call_id that initiated the fencing request */
int client_callid;
/*! The name of client that initiated the fencing request */
char *client_name;
/*! List of the received query results for all the nodes in the cpg group */
GListPtr query_results;
/*! The original request that initiated the remote stonith operation */
xmlNode *request;
/*! The current topology level being executed */
guint level;
/*! The current operation phase being executed */
enum st_remap_phase phase;
/*! Devices with automatic unfencing (always run if "on" requested, never if remapped) */
GListPtr automatic_list;
/*! List of all devices at the currently executing topology level */
GListPtr devices_list;
/*! Current entry in the topology device list */
GListPtr devices;
/*! List of duplicate operations attached to this operation. Once this operation
* completes, the duplicate operations will be closed out as well. */
GListPtr duplicates;
} remote_fencing_op_t;
/*!
* \internal
* \brief Broadcast the result of an operation to the peers.
* \param op, Operation whose result should be broadcast
* \param rc, Result of the operation
*/
void stonith_bcast_result_to_peers(remote_fencing_op_t * op, int rc, gboolean op_merged);
enum st_callback_flags {
st_callback_unknown = 0x0000,
st_callback_notify_fence = 0x0001,
st_callback_device_add = 0x0004,
st_callback_device_del = 0x0010,
st_callback_notify_history = 0x0020,
st_callback_notify_history_synced = 0x0040
};
/*
* Complex fencing requirements are specified via fencing topologies.
* A topology consists of levels; each level is a list of fencing devices.
* Topologies are stored in a hash table by node name. When a node needs to be
* fenced, if it has an entry in the topology table, the levels are tried
* sequentially, and the devices in each level are tried sequentially.
* Fencing is considered successful as soon as any level succeeds;
* a level is considered successful if all its devices succeed.
* Essentially, all devices at a given level are "and-ed" and the
* levels are "or-ed".
*
* This structure is used for the topology table entries.
* Topology levels start from 1, so levels[0] is unused and always NULL.
*/
typedef struct stonith_topology_s {
int kind;
/*! Node name regex or attribute name=value for which topology applies */
char *target;
char *target_value;
char *target_pattern;
char *target_attribute;
/*! Names of fencing devices at each topology level */
GListPtr levels[ST_LEVEL_MAX];
} stonith_topology_t;
void init_device_list(void);
void free_device_list(void);
void init_topology_list(void);
void free_topology_list(void);
void free_stonith_remote_op_list(void);
void init_stonith_remote_op_hash_table(GHashTable **table);
void free_metadata_cache(void);
long long get_stonith_flag(const char *name);
void stonith_command(pcmk__client_t *client, uint32_t id, uint32_t flags,
xmlNode *op_request, const char *remote_peer);
int stonith_device_register(xmlNode * msg, const char **desc, gboolean from_cib);
int stonith_device_remove(const char *id, gboolean from_cib);
char *stonith_level_key(xmlNode * msg, int mode);
int stonith_level_kind(xmlNode * msg);
int stonith_level_register(xmlNode * msg, char **desc);
int stonith_level_remove(xmlNode * msg, char **desc);
stonith_topology_t *find_topology_for_host(const char *host);
void do_local_reply(xmlNode * notify_src, const char *client_id, gboolean sync_reply,
gboolean from_peer);
xmlNode *stonith_construct_reply(xmlNode * request, const char *output, xmlNode * data,
int rc);
void
do_stonith_async_timeout_update(const char *client, const char *call_id, int timeout);
void do_stonith_notify(int options, const char *type, int result, xmlNode * data);
void do_stonith_notify_device(int options, const char *op, int rc, const char *desc);
void do_stonith_notify_level(int options, const char *op, int rc, const char *desc);
remote_fencing_op_t *initiate_remote_stonith_op(pcmk__client_t *client,
xmlNode *request,
gboolean manual_ack);
int process_remote_stonith_exec(xmlNode * msg);
int process_remote_stonith_query(xmlNode * msg);
void *create_remote_stonith_op(const char *client, xmlNode * request, gboolean peer);
int stonith_fence_history(xmlNode *msg, xmlNode **output,
const char *remote_peer, int options);
void stonith_fence_history_trim(void);
bool fencing_peer_active(crm_node_t *peer);
int stonith_manual_ack(xmlNode * msg, remote_fencing_op_t * op);
gboolean string_in_list(GListPtr list, const char *item);
gboolean node_has_attr(const char *node, const char *name, const char *value);
extern char *stonith_our_uname;
extern gboolean stand_alone;
extern GHashTable *device_list;
extern GHashTable *topology;
extern long stonith_watchdog_timeout_ms;
extern GHashTable *stonith_remote_op_list;
diff --git a/doc/Pacemaker_Explained/en-US/Ch-Options.txt b/doc/Pacemaker_Explained/en-US/Ch-Options.txt
index 35856aaf64..a6b4fdd2eb 100644
--- a/doc/Pacemaker_Explained/en-US/Ch-Options.txt
+++ b/doc/Pacemaker_Explained/en-US/Ch-Options.txt
@@ -1,465 +1,481 @@
:compat-mode: legacy
= Cluster-Wide Configuration =
== Configuration Layout ==
The cluster is defined by the Cluster Information Base (CIB),
which uses XML notation. The simplest CIB, an empty one, looks like this:
.An empty configuration
======
[source,XML]
-------
-------
======
The empty configuration above contains the major sections that make up a CIB:
* +cib+: The entire CIB is enclosed with a +cib+ tag. Certain fundamental settings
are defined as attributes of this tag.
** +configuration+: This section -- the primary focus of this document --
contains traditional configuration information such as what resources the
cluster serves and the relationships among them.
*** +crm_config+: cluster-wide configuration options
*** +nodes+: the machines that host the cluster
*** +resources+: the services run by the cluster
*** +constraints+: indications of how resources should be placed
** +status+: This section contains the history of each resource on each node.
Based on this data, the cluster can construct the complete current
state of the cluster. The authoritative source for this section
is the local executor (pacemaker-execd process) on each cluster node, and
the cluster will occasionally repopulate the entire section. For this
reason, it is never written to disk, and administrators are advised
against modifying it in any way.
In this document, configuration settings will be described as 'properties' or 'options'
based on how they are defined in the CIB:
* Properties are XML attributes of an XML element.
* Options are name-value pairs expressed as +nvpair+ child elements of an XML element.
Normally, you will use command-line tools that abstract the XML, so the
distinction will be unimportant; both properties and options are
cluster settings you can tweak.
== CIB Properties ==
Certain settings are defined by CIB properties (that is, attributes of the
+cib+ tag) rather than with the rest of the cluster configuration in the
+configuration+ section.
The reason is simply a matter of parsing. These options are used by the
configuration database which is, by design, mostly ignorant of the content it
holds. So the decision was made to place them in an easy-to-find location.
.CIB Properties
[width="95%",cols="2m,<5",options="header",align="center"]
|=========================================================
|Field |Description
| admin_epoch |
indexterm:[Configuration Version,Cluster]
indexterm:[Cluster,Option,Configuration Version]
indexterm:[admin_epoch,Cluster Option]
indexterm:[Cluster,Option,admin_epoch]
When a node joins the cluster, the cluster performs a check to see
which node has the best configuration. It asks the node with the highest
(+admin_epoch+, +epoch+, +num_updates+) tuple to replace the configuration on
all the nodes -- which makes setting them, and setting them correctly, very
important. +admin_epoch+ is never modified by the cluster; you can use this
to make the configurations on any inactive nodes obsolete. _Never set this
value to zero_. In such cases, the cluster cannot tell the difference between
your configuration and the "empty" one used when nothing is found on disk.
| epoch |
indexterm:[epoch,Cluster Option]
indexterm:[Cluster,Option,epoch]
The cluster increments this every time the configuration is updated (usually by
the administrator).
| num_updates |
indexterm:[num_updates,Cluster Option]
indexterm:[Cluster,Option,num_updates]
The cluster increments this every time the configuration or status is updated
(usually by the cluster) and resets it to 0 when epoch changes.
| validate-with |
indexterm:[validate-with,Cluster Option]
indexterm:[Cluster,Option,validate-with]
Determines the type of XML validation that will be done on the configuration.
If set to +none+, the cluster will not verify that updates conform to the
DTD (nor reject ones that don't). This option can be useful when
operating a mixed-version cluster during an upgrade.
|cib-last-written |
indexterm:[cib-last-written,Cluster Property]
indexterm:[Cluster,Property,cib-last-written]
Indicates when the configuration was last written to disk. Maintained by the
cluster; for informational purposes only.
|have-quorum |
indexterm:[have-quorum,Cluster Property]
indexterm:[Cluster,Property,have-quorum]
Indicates if the cluster has quorum. If false, this may mean that the
cluster cannot start resources or fence other nodes (see
+no-quorum-policy+ below). Maintained by the cluster.
|dc-uuid |
indexterm:[dc-uuid,Cluster Property]
indexterm:[Cluster,Property,dc-uuid]
Indicates which cluster node is the current leader. Used by the
cluster when placing resources and determining the order of some
events. Maintained by the cluster.
|=========================================================
[[s-cluster-options]]
== Cluster Options ==
Cluster options, as you might expect, control how the cluster behaves
when confronted with certain situations.
They are grouped into sets within the +crm_config+ section, and, in advanced
configurations, there may be more than one set. (This will be described later
in the section on <> where we will show how to have the cluster use
different sets of options during working hours than during weekends.) For now,
we will describe the simple case where each option is present at most once.
You can obtain an up-to-date list of cluster options, including
their default values, by running the `man pacemaker-schedulerd` and
`man pacemaker-controld` commands.
.Cluster Options
[width="95%",cols="5m,2,<11",options="header",align="center"]
|=========================================================
|Option |Default |Description
| cluster-name | |
indexterm:[cluster-name,Cluster Property]
indexterm:[Cluster,Property,cluster-name]
An (optional) name for the cluster as a whole. This is mostly for users'
convenience for use as desired in administration, but this can be used
in the Pacemaker configuration in <> (as the
+#cluster-name+ <>). It may
also be used by higher-level tools when displaying cluster information, and by
certain resource agents (for example, the +ocf:heartbeat:GFS2+ agent stores the
cluster name in filesystem meta-data).
| dc-version | |
indexterm:[dc-version,Cluster Property]
indexterm:[Cluster,Property,dc-version]
Version of Pacemaker on the cluster's DC.
Determined automatically by the cluster.
Often includes the hash which identifies the exact Git changeset it was built
from. Used for diagnostic purposes.
| cluster-infrastructure | |
indexterm:[cluster-infrastructure,Cluster Property]
indexterm:[Cluster,Property,cluster-infrastructure]
The messaging stack on which Pacemaker is currently running.
Determined automatically by the cluster.
Used for informational and diagnostic purposes.
| no-quorum-policy | stop
a|
indexterm:[no-quorum-policy,Cluster Option]
indexterm:[Cluster,Option,no-quorum-policy]
What to do when the cluster does not have quorum. Allowed values:
* +ignore:+ continue all resource management
* +freeze:+ continue resource management, but don't recover resources from nodes not in the affected partition
* +stop:+ stop all resources in the affected cluster partition
* +suicide:+ fence all nodes in the affected cluster partition
| batch-limit | 0 |
indexterm:[batch-limit,Cluster Option]
indexterm:[Cluster,Option,batch-limit]
The maximum number of actions that the cluster may execute in parallel across
all nodes. The "correct" value will depend on the speed and load of your
network and cluster nodes. If zero, the cluster will impose a dynamically
calculated limit only when any node has high load.
| migration-limit | -1 |
indexterm:[migration-limit,Cluster Option]
indexterm:[Cluster,Option,migration-limit]
The number of <> actions that the cluster
is allowed to execute in parallel on a node. A value of -1 means unlimited.
| symmetric-cluster | TRUE |
indexterm:[symmetric-cluster,Cluster Option]
indexterm:[Cluster,Option,symmetric-cluster]
Can all resources run on any node by default?
| stop-all-resources | FALSE |
indexterm:[stop-all-resources,Cluster Option]
indexterm:[Cluster,Option,stop-all-resources]
Should the cluster stop all resources?
| stop-orphan-resources | TRUE |
indexterm:[stop-orphan-resources,Cluster Option]
indexterm:[Cluster,Option,stop-orphan-resources]
Should deleted resources be stopped? This value takes precedence over
+is-managed+ (i.e. even unmanaged resources will be stopped if deleted from
the configuration when this value is TRUE).
| stop-orphan-actions | TRUE |
indexterm:[stop-orphan-actions,Cluster Option]
indexterm:[Cluster,Option,stop-orphan-actions]
Should deleted actions be cancelled?
| start-failure-is-fatal | TRUE |
indexterm:[start-failure-is-fatal,Cluster Option]
indexterm:[Cluster,Option,start-failure-is-fatal]
Should a failure to start a resource on a particular node prevent further start
attempts on that node? If FALSE, the cluster will decide whether the same
node is still eligible based on the resource's current failure count
and +migration-threshold+ (see <>).
| enable-startup-probes | TRUE |
indexterm:[enable-startup-probes,Cluster Option]
indexterm:[Cluster,Option,enable-startup-probes]
Should the cluster check for active resources during startup?
| maintenance-mode | FALSE |
indexterm:[maintenance-mode,Cluster Option]
indexterm:[Cluster,Option,maintenance-mode]
Should the cluster refrain from monitoring, starting and stopping resources?
| stonith-enabled | TRUE |
indexterm:[stonith-enabled,Cluster Option]
indexterm:[Cluster,Option,stonith-enabled]
Should failed nodes and nodes with resources that can't be stopped be
shot? If you value your data, set up a STONITH device and enable this.
If true, or unset, the cluster will refuse to start resources unless
one or more STONITH resources have been configured.
If false, unresponsive nodes are immediately assumed to be running no
resources, and resource takeover to online nodes starts without any
further protection (which means _data loss_ if the unresponsive node
still accesses shared storage, for example). See also the +requires+
meta-attribute in <>.
| stonith-action | reboot |
indexterm:[stonith-action,Cluster Option]
indexterm:[Cluster,Option,stonith-action]
Action to send to STONITH device. Allowed values are +reboot+ and +off+.
The value +poweroff+ is also allowed, but is only used for
legacy devices.
| stonith-timeout | 60s |
indexterm:[stonith-timeout,Cluster Option]
indexterm:[Cluster,Option,stonith-timeout]
How long to wait for STONITH actions (reboot, on, off) to complete
| stonith-max-attempts | 10 |
indexterm:[stonith-max-attempts,Cluster Option]
indexterm:[Cluster,Option,stonith-max-attempts]
How many times fencing can fail for a target before the cluster will no longer
immediately re-attempt it.
| stonith-watchdog-timeout | 0 |
indexterm:[stonith-watchdog-timeout,Cluster Option]
indexterm:[Cluster,Option,stonith-watchdog-timeout]
If nonzero, rely on hardware watchdog self-fencing. If positive, assume unseen
nodes self-fence within this much time. If negative, and the
SBD_WATCHDOG_TIMEOUT environment variable is set, use twice that value.
| concurrent-fencing | FALSE |
indexterm:[concurrent-fencing,Cluster Option]
indexterm:[Cluster,Option,concurrent-fencing]
Is the cluster allowed to initiate multiple fence actions concurrently?
| fence-reaction | stop |
indexterm:[fence-reaction,Cluster Option]
indexterm:[Cluster,Option,fence-reaction]
How should a cluster node react if notified of its own fencing? A cluster node
may receive notification of its own fencing if fencing is misconfigured, or if
fabric fencing is in use that doesn't cut cluster communication. Allowed values
are +stop+ to attempt to immediately stop pacemaker and stay stopped, or
+panic+ to attempt to immediately reboot the local node, falling back to stop
on failure. The default is likely to be changed to +panic+ in a future release.
'(since 2.0.3)'
+| priority-fencing-delay | |
+indexterm:[priority-fencing-delay,Cluster Option]
+indexterm:[Cluster,Option,priority-fencing-delay]
+Enforce specified delay for the fencings that are targeting the lost
+nodes with the highest total resource priority in case we don't
+have the majority of the nodes in our cluster partition, so that
+the more significant nodes potentially win any fencing match,
+which is especially meaningful under split-brain of 2-node
+cluster. A promoted resource instance takes the base priority + 1
+on calculation if the base priority is not 0. If all the nodes
+have equal priority, then any pcmk_delay_base/max configured for
+the corresponding fencing resources will be applied. Otherwise as
+long as it's set, even if to 0, it takes precedence over any
+configured pcmk_delay_base/max. By default, priority fencing
+delay is disabled.
+
| cluster-delay | 60s |
indexterm:[cluster-delay,Cluster Option]
indexterm:[Cluster,Option,cluster-delay]
Estimated maximum round-trip delay over the network (excluding action
execution). If the DC requires an action to be executed on another
node, it will consider the action failed if it does not get a response
from the other node in this time (after considering the action's
own timeout). The "correct" value will depend on the speed and load of your
network and cluster nodes.
| dc-deadtime | 20s |
indexterm:[dc-deadtime,Cluster Option]
indexterm:[Cluster,Option,dc-deadtime]
How long to wait for a response from other nodes during startup.
The "correct" value will depend on the speed/load of your network and the type of switches used.
| cluster-ipc-limit | 500 |
indexterm:[cluster-ipc-limit,Cluster Option]
indexterm:[Cluster,Option,cluster-ipc-limit]
The maximum IPC message backlog before one cluster daemon will disconnect
another. This is of use in large clusters, for which a good value is the number
of resources in the cluster multiplied by the number of nodes. The default of
500 is also the minimum. Raise this if you see "Evicting client" messages for
cluster daemon PIDs in the logs.
| pe-error-series-max | -1 |
indexterm:[pe-error-series-max,Cluster Option]
indexterm:[Cluster,Option,pe-error-series-max]
The number of PE inputs resulting in ERRORs to save. Used when reporting problems.
A value of -1 means unlimited (report all).
| pe-warn-series-max | -1 |
indexterm:[pe-warn-series-max,Cluster Option]
indexterm:[Cluster,Option,pe-warn-series-max]
The number of PE inputs resulting in WARNINGs to save. Used when reporting problems.
A value of -1 means unlimited (report all).
| pe-input-series-max | -1 |
indexterm:[pe-input-series-max,Cluster Option]
indexterm:[Cluster,Option,pe-input-series-max]
The number of "normal" PE inputs to save. Used when reporting problems.
A value of -1 means unlimited (report all).
| placement-strategy | default |
indexterm:[placement-strategy,Cluster Option]
indexterm:[Cluster,Option,placement-strategy]
How the cluster should allocate resources to nodes (see <>).
Allowed values are +default+, +utilization+, +balanced+, and +minimal+.
| node-health-strategy | none |
indexterm:[node-health-strategy,Cluster Option]
indexterm:[Cluster,Option,node-health-strategy]
How the cluster should react to node health attributes (see <>).
Allowed values are +none+, +migrate-on-red+, +only-green+, +progressive+, and
+custom+.
| enable-acl | FALSE |
indexterm:[enable-acl,Cluster Option]
indexterm:[Cluster,Option,enable-acl]
Whether access control lists (ACLs) (see <>) can be used to authorize
modifications to the CIB.
| node-health-base | 0 |
indexterm:[node-health-base,Cluster Option]
indexterm:[Cluster,Option,node-health-base]
The base health score assigned to a node. Only used when
+node-health-strategy+ is +progressive+.
| node-health-green | 0 |
indexterm:[node-health-green,Cluster Option]
indexterm:[Cluster,Option,node-health-green]
The score to use for a node health attribute whose value is +green+.
Only used when +node-health-strategy+ is +progressive+ or +custom+.
| node-health-yellow | 0 |
indexterm:[node-health-yellow,Cluster Option]
indexterm:[Cluster,Option,node-health-yellow]
The score to use for a node health attribute whose value is +yellow+.
Only used when +node-health-strategy+ is +progressive+ or +custom+.
| node-health-red | 0 |
indexterm:[node-health-red,Cluster Option]
indexterm:[Cluster,Option,node-health-red]
The score to use for a node health attribute whose value is +red+.
Only used when +node-health-strategy+ is +progressive+ or +custom+.
| cluster-recheck-interval | 15min |
indexterm:[cluster-recheck-interval,Cluster Option]
indexterm:[Cluster,Option,cluster-recheck-interval]
Pacemaker is primarily event-driven, and looks ahead to know when to recheck
the cluster for failure timeouts and most time-based rules. However, it will
also recheck the cluster after this amount of inactivity. This has two goals:
rules with +date_spec+ are only guaranteed to be checked this often, and it
also serves as a fail-safe for certain classes of scheduler bugs. A value of 0
disables this polling; positive values are a time interval.
| shutdown-lock | false |
The default of false allows active resources to be recovered elsewhere when
their node is cleanly shut down, which is what the vast majority of users will
want. However, some users prefer to make resources highly available only for
failures, with no recovery for clean shutdowns. If this option is true,
resources active on a node when it is cleanly shut down are kept "locked" to
that node (not allowed to run elsewhere) until they start again on that node
after it rejoins (or for at most shutdown-lock-limit, if set). Stonith
resources and Pacemaker Remote connections are never locked. Clone and bundle
instances and the master role of promotable clones are currently never locked,
though support could be added in a future release. Locks may be manually
cleared using the `--refresh` option of `crm_resource` (both the resource and
node must be specified; this works with remote nodes if their connection
resource's target-role is set to Stopped, but not if Pacemaker Remote is
stopped on the remote node without disabling the connection resource).
indexterm:[shutdown-lock,Cluster Option]
indexterm:[Cluster,Option,shutdown-lock]
| shutdown-lock-limit | 0 |
If shutdown-lock is true, and this is set to a nonzero time duration, locked
resources will be allowed to start after this much time has passed since the
node shutdown was initiated, even if the node has not rejoined. (This works
with remote nodes only if their connection resource's target-role is set to
Stopped.)
indexterm:[shutdown-lock-limit,Cluster Option]
indexterm:[Cluster,Option,shutdown-lock-limit]
| remove-after-stop | FALSE |
indexterm:[remove-after-stop,Cluster Option]
indexterm:[Cluster,Option,remove-after-stop]
_Advanced Use Only:_ Should the cluster remove resources from the LRM after
they are stopped? Values other than the default are, at best, poorly tested and
potentially dangerous.
| startup-fencing | TRUE |
indexterm:[startup-fencing,Cluster Option]
indexterm:[Cluster,Option,startup-fencing]
_Advanced Use Only:_ Should the cluster shoot unseen nodes?
Not using the default is very unsafe!
| election-timeout | 2min |
indexterm:[election-timeout,Cluster Option]
indexterm:[Cluster,Option,election-timeout]
_Advanced Use Only:_ If you need to adjust this value, it probably indicates
the presence of a bug.
| shutdown-escalation | 20min |
indexterm:[shutdown-escalation,Cluster Option]
indexterm:[Cluster,Option,shutdown-escalation]
_Advanced Use Only:_ If you need to adjust this value, it probably indicates
the presence of a bug.
| join-integration-timeout | 3min |
indexterm:[join-integration-timeout,Cluster Option]
indexterm:[Cluster,Option,join-integration-timeout]
_Advanced Use Only:_ If you need to adjust this value, it probably indicates
the presence of a bug.
| join-finalization-timeout | 30min |
indexterm:[join-finalization-timeout,Cluster Option]
indexterm:[Cluster,Option,join-finalization-timeout]
_Advanced Use Only:_ If you need to adjust this value, it probably indicates
the presence of a bug.
| transition-delay | 0s |
indexterm:[transition-delay,Cluster Option]
indexterm:[Cluster,Option,transition-delay]
_Advanced Use Only:_ Delay cluster recovery for the configured interval to
allow for additional/related events to occur. Useful if your configuration is
sensitive to the order in which ping updates arrive.
Enabling this option will slow down cluster recovery under
all conditions.
|=========================================================
diff --git a/include/crm/fencing/internal.h b/include/crm/fencing/internal.h
index 82389b4181..3f2c090602 100644
--- a/include/crm/fencing/internal.h
+++ b/include/crm/fencing/internal.h
@@ -1,172 +1,173 @@
/*
* Copyright 2011-2019 the Pacemaker project contributors
*
* The version control history for this file may have further details.
*
* This source code is licensed under the GNU Lesser General Public License
* version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY.
*/
#ifndef STONITH_NG_INTERNAL__H
# define STONITH_NG_INTERNAL__H
# include
# include
# include
# include
struct stonith_action_s;
typedef struct stonith_action_s stonith_action_t;
stonith_action_t *stonith_action_create(const char *agent,
const char *_action,
const char *victim,
uint32_t victim_nodeid,
int timeout,
GHashTable * device_args, GHashTable * port_map);
void stonith__destroy_action(stonith_action_t *action);
void stonith__action_result(stonith_action_t *action, int *rc, char **output,
char **error_output);
int
stonith_action_execute_async(stonith_action_t * action,
void *userdata,
void (*done) (GPid pid, int rc, const char *output,
gpointer user_data),
void (*fork_cb) (GPid pid, gpointer user_data));
int stonith__execute(stonith_action_t *action);
xmlNode *create_level_registration_xml(const char *node, const char *pattern,
const char *attr, const char *value,
int level,
stonith_key_value_t *device_list);
xmlNode *create_device_registration_xml(const char *id,
enum stonith_namespace namespace,
const char *agent,
stonith_key_value_t *params,
const char *rsc_provides);
void stonith__register_messages(pcmk__output_t *out);
GList *stonith__parse_targets(const char *hosts);
gboolean stonith__later_succeeded(stonith_history_t *event, stonith_history_t *top_history);
stonith_history_t *stonith__sort_history(stonith_history_t *history);
# define ST_LEVEL_MAX 10
# define F_STONITH_CLIENTID "st_clientid"
# define F_STONITH_CALLOPTS "st_callopt"
# define F_STONITH_CALLID "st_callid"
# define F_STONITH_CALLDATA "st_calldata"
# define F_STONITH_OPERATION "st_op"
# define F_STONITH_TARGET "st_target"
# define F_STONITH_REMOTE_OP_ID "st_remote_op"
# define F_STONITH_RC "st_rc"
/*! Timeout period per a device execution */
# define F_STONITH_TIMEOUT "st_timeout"
# define F_STONITH_TOLERANCE "st_tolerance"
+# define F_STONITH_DELAY "st_delay"
/*! Action specific timeout period returned in query of fencing devices. */
# define F_STONITH_ACTION_TIMEOUT "st_action_timeout"
/*! Host in query result is not allowed to run this action */
# define F_STONITH_ACTION_DISALLOWED "st_action_disallowed"
/*! Maximum of random fencing delay for a device */
# define F_STONITH_DELAY_MAX "st_delay_max"
/*! Base delay used for a fencing delay */
# define F_STONITH_DELAY_BASE "st_delay_base"
/*! Has this device been verified using a monitor type
* operation (monitor, list, status) */
# define F_STONITH_DEVICE_VERIFIED "st_monitor_verified"
/*! device is required for this action */
# define F_STONITH_DEVICE_REQUIRED "st_required"
/*! number of available devices in query result */
# define F_STONITH_AVAILABLE_DEVICES "st-available-devices"
# define F_STONITH_CALLBACK_TOKEN "st_async_id"
# define F_STONITH_CLIENTNAME "st_clientname"
# define F_STONITH_CLIENTNODE "st_clientnode"
# define F_STONITH_NOTIFY_ACTIVATE "st_notify_activate"
# define F_STONITH_NOTIFY_DEACTIVATE "st_notify_deactivate"
# define F_STONITH_DELEGATE "st_delegate"
/*! The node initiating the stonith operation. If an operation
* is relayed, this is the last node the operation lands on. When
* in standalone mode, origin is the client's id that originated the
* operation. */
# define F_STONITH_ORIGIN "st_origin"
# define F_STONITH_HISTORY_LIST "st_history"
# define F_STONITH_DATE "st_date"
# define F_STONITH_STATE "st_state"
# define F_STONITH_ACTIVE "st_active"
# define F_STONITH_DIFFERENTIAL "st_differential"
# define F_STONITH_DEVICE "st_device_id"
# define F_STONITH_ACTION "st_device_action"
# define F_STONITH_MODE "st_mode"
# define F_STONITH_MERGED "st_op_merged"
# define T_STONITH_NG "stonith-ng"
# define T_STONITH_REPLY "st-reply"
/*! For async operations, an event from the server containing
* the total amount of time the server is allowing for the operation
* to take place is returned to the client. */
# define T_STONITH_TIMEOUT_VALUE "st-async-timeout-value"
# define T_STONITH_NOTIFY "st_notify"
# define STONITH_ATTR_HOSTARG "pcmk_host_argument"
# define STONITH_ATTR_HOSTMAP "pcmk_host_map"
# define STONITH_ATTR_HOSTLIST "pcmk_host_list"
# define STONITH_ATTR_HOSTCHECK "pcmk_host_check"
# define STONITH_ATTR_DELAY_MAX "pcmk_delay_max"
# define STONITH_ATTR_DELAY_BASE "pcmk_delay_base"
# define STONITH_ATTR_ACTION_LIMIT "pcmk_action_limit"
# define STONITH_ATTR_ACTION_OP "action"
# define STONITH_OP_EXEC "st_execute"
# define STONITH_OP_TIMEOUT_UPDATE "st_timeout_update"
# define STONITH_OP_QUERY "st_query"
# define STONITH_OP_FENCE "st_fence"
# define STONITH_OP_RELAY "st_relay"
# define STONITH_OP_DEVICE_ADD "st_device_register"
# define STONITH_OP_DEVICE_DEL "st_device_remove"
# define STONITH_OP_FENCE_HISTORY "st_fence_history"
# define STONITH_OP_LEVEL_ADD "st_level_add"
# define STONITH_OP_LEVEL_DEL "st_level_remove"
# define STONITH_WATCHDOG_AGENT "#watchdog"
# ifdef HAVE_STONITH_STONITH_H
// utilities from st_lha.c
int stonith__list_lha_agents(stonith_key_value_t **devices);
int stonith__lha_metadata(const char *agent, int timeout, char **output);
bool stonith__agent_is_lha(const char *agent);
int stonith__lha_validate(stonith_t *st, int call_options, const char *target,
const char *agent, GHashTable *params,
int timeout, char **output, char **error_output);
# endif
// utilities from st_rhcs.c
int stonith__list_rhcs_agents(stonith_key_value_t **devices);
int stonith__rhcs_metadata(const char *agent, int timeout, char **output);
bool stonith__agent_is_rhcs(const char *agent);
int stonith__rhcs_validate(stonith_t *st, int call_options, const char *target,
const char *agent, GHashTable *params,
int timeout, char **output, char **error_output);
int stonith__failed_history(pcmk__output_t *out, va_list args);
int stonith__history(pcmk__output_t *out, va_list args);
int stonith__full_history(pcmk__output_t *out, va_list args);
int stonith__full_history_xml(pcmk__output_t *out, va_list args);
int stonith__last_fenced_html(pcmk__output_t *out, va_list args);
int stonith__last_fenced_text(pcmk__output_t *out, va_list args);
int stonith__last_fenced_xml(pcmk__output_t *out, va_list args);
int stonith__pending_actions(pcmk__output_t *out, va_list args);
int stonith__event_html(pcmk__output_t *out, va_list args);
int stonith__event_text(pcmk__output_t *out, va_list args);
int stonith__event_xml(pcmk__output_t *out, va_list args);
int stonith__validate_agent_html(pcmk__output_t *out, va_list args);
int stonith__validate_agent_text(pcmk__output_t *out, va_list args);
int stonith__validate_agent_xml(pcmk__output_t *out, va_list args);
#endif
diff --git a/include/crm/msg_xml.h b/include/crm/msg_xml.h
index d0cdf6cd3d..af3f33e883 100644
--- a/include/crm/msg_xml.h
+++ b/include/crm/msg_xml.h
@@ -1,424 +1,425 @@
/*
* Copyright 2004-2020 the Pacemaker project contributors
*
* The version control history for this file may have further details.
*
* This source code is licensed under the GNU Lesser General Public License
* version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY.
*/
#ifndef XML_TAGS__H
# define XML_TAGS__H
#ifdef __cplusplus
extern "C" {
#endif
# ifndef F_ORIG
# define F_ORIG "src"
# endif
# ifndef F_SEQ
# define F_SEQ "seq"
# endif
# ifndef F_SUBTYPE
# define F_SUBTYPE "subt"
# endif
# ifndef F_TYPE
# define F_TYPE "t"
# endif
# ifndef F_CLIENTNAME
# define F_CLIENTNAME "cn"
# endif
# ifndef F_XML_TAGNAME
# define F_XML_TAGNAME "__name__"
# endif
# ifndef T_CRM
# define T_CRM "crmd"
# endif
# ifndef T_ATTRD
# define T_ATTRD "attrd"
# endif
# define CIB_OPTIONS_FIRST "cib-bootstrap-options"
# define F_CRM_DATA "crm_xml"
# define F_CRM_TASK "crm_task"
# define F_CRM_HOST_TO "crm_host_to"
# define F_CRM_MSG_TYPE F_SUBTYPE
# define F_CRM_SYS_TO "crm_sys_to"
# define F_CRM_SYS_FROM "crm_sys_from"
# define F_CRM_HOST_FROM F_ORIG
# define F_CRM_REFERENCE XML_ATTR_REFERENCE
# define F_CRM_VERSION XML_ATTR_VERSION
# define F_CRM_ORIGIN "origin"
# define F_CRM_USER "crm_user"
# define F_CRM_JOIN_ID "join_id"
# define F_CRM_DC_LEAVING "dc-leaving"
# define F_CRM_ELECTION_ID "election-id"
# define F_CRM_ELECTION_AGE_S "election-age-sec"
# define F_CRM_ELECTION_AGE_US "election-age-nano-sec"
# define F_CRM_ELECTION_OWNER "election-owner"
# define F_CRM_TGRAPH "crm-tgraph-file"
# define F_CRM_TGRAPH_INPUT "crm-tgraph-in"
# define F_CRM_THROTTLE_MODE "crm-limit-mode"
# define F_CRM_THROTTLE_MAX "crm-limit-max"
/*---- Common tags/attrs */
# define XML_DIFF_MARKER "__crm_diff_marker__"
# define XML_TAG_CIB "cib"
# define XML_TAG_FAILED "failed"
# define XML_ATTR_CRM_VERSION "crm_feature_set"
# define XML_ATTR_DIGEST "digest"
# define XML_ATTR_VALIDATION "validate-with"
# define XML_ATTR_RA_VERSION "ra-version"
# define XML_ATTR_QUORUM_PANIC "no-quorum-panic"
# define XML_ATTR_HAVE_QUORUM "have-quorum"
# define XML_ATTR_HAVE_WATCHDOG "have-watchdog"
# define XML_ATTR_GENERATION "epoch"
# define XML_ATTR_GENERATION_ADMIN "admin_epoch"
# define XML_ATTR_NUMUPDATES "num_updates"
# define XML_ATTR_TIMEOUT "timeout"
# define XML_ATTR_ORIGIN "crm-debug-origin"
# define XML_ATTR_TSTAMP "crm-timestamp"
# define XML_CIB_ATTR_WRITTEN "cib-last-written"
# define XML_ATTR_VERSION "version"
# define XML_ATTR_DESC "description"
# define XML_ATTR_ID "id"
# define XML_ATTR_IDREF "id-ref"
# define XML_ATTR_ID_LONG "long-id"
# define XML_ATTR_TYPE "type"
# define XML_ATTR_VERBOSE "verbose"
# define XML_ATTR_OP "op"
# define XML_ATTR_DC_UUID "dc-uuid"
# define XML_ATTR_UPDATE_ORIG "update-origin"
# define XML_ATTR_UPDATE_CLIENT "update-client"
# define XML_ATTR_UPDATE_USER "update-user"
# define XML_BOOLEAN_TRUE "true"
# define XML_BOOLEAN_FALSE "false"
# define XML_BOOLEAN_YES XML_BOOLEAN_TRUE
# define XML_BOOLEAN_NO XML_BOOLEAN_FALSE
# define XML_TAG_OPTIONS "options"
/*---- top level tags/attrs */
# define XML_ATTR_REQUEST "request"
# define XML_ATTR_RESPONSE "response"
# define XML_ATTR_UNAME "uname"
# define XML_ATTR_UUID "id"
# define XML_ATTR_REFERENCE "reference"
# define XML_CRM_TAG_PING "ping_response"
# define XML_PING_ATTR_STATUS "result"
# define XML_PING_ATTR_SYSFROM "crm_subsystem"
# define XML_PING_ATTR_CRMDSTATE "crmd_state"
# define XML_TAG_FRAGMENT "cib_fragment"
# define XML_FAIL_TAG_CIB "failed_update"
# define XML_FAILCIB_ATTR_ID "id"
# define XML_FAILCIB_ATTR_OBJTYPE "object_type"
# define XML_FAILCIB_ATTR_OP "operation"
# define XML_FAILCIB_ATTR_REASON "reason"
/*---- CIB specific tags/attrs */
# define XML_CIB_TAG_SECTION_ALL "all"
# define XML_CIB_TAG_CONFIGURATION "configuration"
# define XML_CIB_TAG_STATUS "status"
# define XML_CIB_TAG_RESOURCES "resources"
# define XML_CIB_TAG_NODES "nodes"
# define XML_CIB_TAG_DOMAINS "domains"
# define XML_CIB_TAG_CONSTRAINTS "constraints"
# define XML_CIB_TAG_CRMCONFIG "crm_config"
# define XML_CIB_TAG_OPCONFIG "op_defaults"
# define XML_CIB_TAG_RSCCONFIG "rsc_defaults"
# define XML_CIB_TAG_ACLS "acls"
# define XML_CIB_TAG_ALERTS "alerts"
# define XML_CIB_TAG_ALERT "alert"
# define XML_CIB_TAG_ALERT_RECIPIENT "recipient"
# define XML_CIB_TAG_ALERT_SELECT "select"
# define XML_CIB_TAG_ALERT_ATTRIBUTES "select_attributes"
# define XML_CIB_TAG_ALERT_FENCING "select_fencing"
# define XML_CIB_TAG_ALERT_NODES "select_nodes"
# define XML_CIB_TAG_ALERT_RESOURCES "select_resources"
# define XML_CIB_TAG_ALERT_ATTR "attribute"
# define XML_CIB_TAG_STATE "node_state"
# define XML_CIB_TAG_NODE "node"
# define XML_CIB_TAG_NVPAIR "nvpair"
# define XML_CIB_TAG_PROPSET "cluster_property_set"
# define XML_TAG_ATTR_SETS "instance_attributes"
# define XML_TAG_META_SETS "meta_attributes"
# define XML_TAG_ATTRS "attributes"
# define XML_TAG_RSC_VER_ATTRS "rsc_versioned_attrs"
# define XML_TAG_OP_VER_ATTRS "op_versioned_attrs"
# define XML_TAG_OP_VER_META "op_versioned_meta"
# define XML_TAG_PARAMS "parameters"
# define XML_TAG_PARAM "param"
# define XML_TAG_UTILIZATION "utilization"
# define XML_TAG_RESOURCE_REF "resource_ref"
# define XML_CIB_TAG_RESOURCE "primitive"
# define XML_CIB_TAG_GROUP "group"
# define XML_CIB_TAG_INCARNATION "clone"
# define XML_CIB_TAG_MASTER "master" // deprecated since 2.0.0
# define XML_CIB_TAG_CONTAINER "bundle"
# define XML_CIB_TAG_RSC_TEMPLATE "template"
# define XML_RSC_ATTR_TARGET "container-attribute-target"
# define XML_RSC_ATTR_RESTART "restart-type"
# define XML_RSC_ATTR_ORDERED "ordered"
# define XML_RSC_ATTR_INTERLEAVE "interleave"
# define XML_RSC_ATTR_INCARNATION "clone"
# define XML_RSC_ATTR_INCARNATION_MAX "clone-max"
# define XML_RSC_ATTR_INCARNATION_MIN "clone-min"
# define XML_RSC_ATTR_INCARNATION_NODEMAX "clone-node-max"
# define XML_RSC_ATTR_PROMOTABLE "promotable"
# define XML_RSC_ATTR_PROMOTED_MAX "promoted-max"
# define XML_RSC_ATTR_PROMOTED_NODEMAX "promoted-node-max"
# define XML_RSC_ATTR_MASTER_MAX "master-max" // deprecated since 2.0.0
# define XML_RSC_ATTR_MASTER_NODEMAX "master-node-max" // deprecated since 2.0.0
# define XML_RSC_ATTR_MANAGED "is-managed"
# define XML_RSC_ATTR_TARGET_ROLE "target-role"
# define XML_RSC_ATTR_UNIQUE "globally-unique"
# define XML_RSC_ATTR_NOTIFY "notify"
# define XML_RSC_ATTR_STICKINESS "resource-stickiness"
# define XML_RSC_ATTR_FAIL_STICKINESS "migration-threshold"
# define XML_RSC_ATTR_FAIL_TIMEOUT "failure-timeout"
# define XML_RSC_ATTR_MULTIPLE "multiple-active"
# define XML_RSC_ATTR_REQUIRES "requires"
# define XML_RSC_ATTR_PROVIDES "provides"
# define XML_RSC_ATTR_CONTAINER "container"
# define XML_RSC_ATTR_INTERNAL_RSC "internal_rsc"
# define XML_RSC_ATTR_MAINTENANCE "maintenance"
# define XML_RSC_ATTR_REMOTE_NODE "remote-node"
# define XML_RSC_ATTR_CLEAR_OP "clear_failure_op"
# define XML_RSC_ATTR_CLEAR_INTERVAL "clear_failure_interval"
# define XML_RSC_ATTR_REMOTE_RA_ADDR "addr"
# define XML_RSC_ATTR_REMOTE_RA_SERVER "server"
# define XML_RSC_ATTR_REMOTE_RA_PORT "port"
# define XML_REMOTE_ATTR_RECONNECT_INTERVAL "reconnect_interval"
# define XML_OP_ATTR_ON_FAIL "on-fail"
# define XML_OP_ATTR_START_DELAY "start-delay"
# define XML_OP_ATTR_ALLOW_MIGRATE "allow-migrate"
# define XML_OP_ATTR_ORIGIN "interval-origin"
# define XML_OP_ATTR_PENDING "record-pending"
# define XML_OP_ATTR_DIGESTS_ALL "digests-all"
# define XML_OP_ATTR_DIGESTS_SECURE "digests-secure"
# define XML_CIB_TAG_LRM "lrm"
# define XML_LRM_TAG_RESOURCES "lrm_resources"
# define XML_LRM_TAG_RESOURCE "lrm_resource"
# define XML_LRM_TAG_RSC_OP "lrm_rsc_op"
# define XML_AGENT_ATTR_CLASS "class"
# define XML_AGENT_ATTR_PROVIDER "provider"
# define XML_CIB_ATTR_REPLACE "replace"
# define XML_CIB_ATTR_SOURCE "source"
# define XML_CIB_ATTR_PRIORITY "priority"
# define XML_CIB_ATTR_SOURCE "source"
# define XML_NODE_JOIN_STATE "join"
# define XML_NODE_EXPECTED "expected"
# define XML_NODE_IN_CLUSTER "in_ccm"
# define XML_NODE_IS_PEER "crmd"
# define XML_NODE_IS_REMOTE "remote_node"
# define XML_NODE_IS_FENCED "node_fenced"
# define XML_NODE_IS_MAINTENANCE "node_in_maintenance"
# define XML_CIB_ATTR_SHUTDOWN "shutdown"
/* Aside from being an old name for the executor, LRM is a misnomer here because
* the controller and scheduler use these to track actions, which are not always
* executor operations.
*/
// XML attribute that takes interval specification (user-facing configuration)
# define XML_LRM_ATTR_INTERVAL "interval"
// XML attribute that takes interval in milliseconds (daemon APIs)
// (identical value as above, but different constant allows clearer code intent)
# define XML_LRM_ATTR_INTERVAL_MS XML_LRM_ATTR_INTERVAL
# define XML_LRM_ATTR_TASK "operation"
# define XML_LRM_ATTR_TASK_KEY "operation_key"
# define XML_LRM_ATTR_TARGET "on_node"
# define XML_LRM_ATTR_TARGET_UUID "on_node_uuid"
/*! Actions to be executed on Pacemaker Remote nodes are routed through the
* controller on the cluster node hosting the remote connection. That cluster
* node is considered the router node for the action.
*/
# define XML_LRM_ATTR_ROUTER_NODE "router_node"
# define XML_LRM_ATTR_RSCID "rsc-id"
# define XML_LRM_ATTR_OPSTATUS "op-status"
# define XML_LRM_ATTR_RC "rc-code"
# define XML_LRM_ATTR_CALLID "call-id"
# define XML_LRM_ATTR_OP_DIGEST "op-digest"
# define XML_LRM_ATTR_OP_RESTART "op-force-restart"
# define XML_LRM_ATTR_OP_SECURE "op-secure-params"
# define XML_LRM_ATTR_RESTART_DIGEST "op-restart-digest"
# define XML_LRM_ATTR_SECURE_DIGEST "op-secure-digest"
# define XML_LRM_ATTR_EXIT_REASON "exit-reason"
# define XML_RSC_OP_LAST_CHANGE "last-rc-change"
# define XML_RSC_OP_LAST_RUN "last-run" // deprecated since 2.0.3
# define XML_RSC_OP_T_EXEC "exec-time"
# define XML_RSC_OP_T_QUEUE "queue-time"
# define XML_LRM_ATTR_MIGRATE_SOURCE "migrate_source"
# define XML_LRM_ATTR_MIGRATE_TARGET "migrate_target"
# define XML_TAG_GRAPH "transition_graph"
# define XML_GRAPH_TAG_RSC_OP "rsc_op"
# define XML_GRAPH_TAG_PSEUDO_EVENT "pseudo_event"
# define XML_GRAPH_TAG_CRM_EVENT "crm_event"
# define XML_GRAPH_TAG_DOWNED "downed"
# define XML_GRAPH_TAG_MAINTENANCE "maintenance"
# define XML_TAG_RULE "rule"
# define XML_RULE_ATTR_SCORE "score"
# define XML_RULE_ATTR_SCORE_ATTRIBUTE "score-attribute"
# define XML_RULE_ATTR_ROLE "role"
# define XML_RULE_ATTR_BOOLEAN_OP "boolean-op"
# define XML_TAG_EXPRESSION "expression"
# define XML_EXPR_ATTR_ATTRIBUTE "attribute"
# define XML_EXPR_ATTR_OPERATION "operation"
# define XML_EXPR_ATTR_VALUE "value"
# define XML_EXPR_ATTR_TYPE "type"
# define XML_EXPR_ATTR_VALUE_SOURCE "value-source"
# define XML_CONS_TAG_RSC_DEPEND "rsc_colocation"
# define XML_CONS_TAG_RSC_ORDER "rsc_order"
# define XML_CONS_TAG_RSC_LOCATION "rsc_location"
# define XML_CONS_TAG_RSC_TICKET "rsc_ticket"
# define XML_CONS_TAG_RSC_SET "resource_set"
# define XML_CONS_ATTR_SYMMETRICAL "symmetrical"
# define XML_LOCATION_ATTR_DISCOVERY "resource-discovery"
# define XML_COLOC_ATTR_SOURCE "rsc"
# define XML_COLOC_ATTR_SOURCE_ROLE "rsc-role"
# define XML_COLOC_ATTR_TARGET "with-rsc"
# define XML_COLOC_ATTR_TARGET_ROLE "with-rsc-role"
# define XML_COLOC_ATTR_NODE_ATTR "node-attribute"
# define XML_COLOC_ATTR_SOURCE_INSTANCE "rsc-instance"
# define XML_COLOC_ATTR_TARGET_INSTANCE "with-rsc-instance"
# define XML_LOC_ATTR_SOURCE "rsc"
# define XML_LOC_ATTR_SOURCE_PATTERN "rsc-pattern"
# define XML_ORDER_ATTR_FIRST "first"
# define XML_ORDER_ATTR_THEN "then"
# define XML_ORDER_ATTR_FIRST_ACTION "first-action"
# define XML_ORDER_ATTR_THEN_ACTION "then-action"
# define XML_ORDER_ATTR_FIRST_INSTANCE "first-instance"
# define XML_ORDER_ATTR_THEN_INSTANCE "then-instance"
# define XML_ORDER_ATTR_KIND "kind"
# define XML_TICKET_ATTR_TICKET "ticket"
# define XML_TICKET_ATTR_LOSS_POLICY "loss-policy"
# define XML_NVPAIR_ATTR_NAME "name"
# define XML_NVPAIR_ATTR_VALUE "value"
# define XML_NODE_ATTR_RSC_DISCOVERY "resource-discovery-enabled"
# define XML_CONFIG_ATTR_DC_DEADTIME "dc-deadtime"
# define XML_CONFIG_ATTR_ELECTION_FAIL "election-timeout"
# define XML_CONFIG_ATTR_FORCE_QUIT "shutdown-escalation"
# define XML_CONFIG_ATTR_RECHECK "cluster-recheck-interval"
# define XML_CONFIG_ATTR_FENCE_REACTION "fence-reaction"
# define XML_CONFIG_ATTR_SHUTDOWN_LOCK "shutdown-lock"
# define XML_CONFIG_ATTR_SHUTDOWN_LOCK_LIMIT "shutdown-lock-limit"
+# define XML_CONFIG_ATTR_PRIORITY_FENCING_DELAY "priority-fencing-delay"
# define XML_ALERT_ATTR_PATH "path"
# define XML_ALERT_ATTR_TIMEOUT "timeout"
# define XML_ALERT_ATTR_TSTAMP_FORMAT "timestamp-format"
# define XML_ALERT_ATTR_REC_VALUE "value"
# define XML_CIB_TAG_GENERATION_TUPPLE "generation_tuple"
# define XML_ATTR_TRANSITION_MAGIC "transition-magic"
# define XML_ATTR_TRANSITION_KEY "transition-key"
# define XML_ATTR_TE_NOWAIT "op_no_wait"
# define XML_ATTR_TE_TARGET_RC "op_target_rc"
# define XML_TAG_TRANSIENT_NODEATTRS "transient_attributes"
# define XML_TAG_DIFF_ADDED "diff-added"
# define XML_TAG_DIFF_REMOVED "diff-removed"
# define XML_ACL_TAG_USER "acl_target"
# define XML_ACL_TAG_USERv1 "acl_user"
# define XML_ACL_TAG_GROUP "acl_group"
# define XML_ACL_TAG_ROLE "acl_role"
# define XML_ACL_TAG_PERMISSION "acl_permission"
# define XML_ACL_TAG_ROLE_REF "role"
# define XML_ACL_TAG_ROLE_REFv1 "role_ref"
# define XML_ACL_ATTR_KIND "kind"
# define XML_ACL_TAG_READ "read"
# define XML_ACL_TAG_WRITE "write"
# define XML_ACL_TAG_DENY "deny"
# define XML_ACL_ATTR_REF "reference"
# define XML_ACL_ATTR_REFv1 "ref"
# define XML_ACL_ATTR_TAG "object-type"
# define XML_ACL_ATTR_TAGv1 "tag"
# define XML_ACL_ATTR_XPATH "xpath"
# define XML_ACL_ATTR_ATTRIBUTE "attribute"
# define XML_CIB_TAG_TICKETS "tickets"
# define XML_CIB_TAG_TICKET_STATE "ticket_state"
# define XML_CIB_TAG_TAGS "tags"
# define XML_CIB_TAG_TAG "tag"
# define XML_CIB_TAG_OBJ_REF "obj_ref"
# define XML_TAG_FENCING_TOPOLOGY "fencing-topology"
# define XML_TAG_FENCING_LEVEL "fencing-level"
# define XML_ATTR_STONITH_INDEX "index"
# define XML_ATTR_STONITH_TARGET "target"
# define XML_ATTR_STONITH_TARGET_VALUE "target-value"
# define XML_ATTR_STONITH_TARGET_PATTERN "target-pattern"
# define XML_ATTR_STONITH_TARGET_ATTRIBUTE "target-attribute"
# define XML_ATTR_STONITH_DEVICES "devices"
# define XML_TAG_DIFF "diff"
# define XML_DIFF_VERSION "version"
# define XML_DIFF_VSOURCE "source"
# define XML_DIFF_VTARGET "target"
# define XML_DIFF_CHANGE "change"
# define XML_DIFF_LIST "change-list"
# define XML_DIFF_ATTR "change-attr"
# define XML_DIFF_RESULT "change-result"
# define XML_DIFF_OP "operation"
# define XML_DIFF_PATH "path"
# define XML_DIFF_POSITION "position"
# include
# define ID(x) crm_element_value(x, XML_ATTR_ID)
# define TYPE(x) crm_element_name(x)
#ifdef __cplusplus
}
#endif
#endif
diff --git a/include/crm/pengine/internal.h b/include/crm/pengine/internal.h
index 6fcb63ef56..b4ce1574cd 100644
--- a/include/crm/pengine/internal.h
+++ b/include/crm/pengine/internal.h
@@ -1,444 +1,444 @@
/*
* Copyright 2004-2020 the Pacemaker project contributors
*
* The version control history for this file may have further details.
*
* This source code is licensed under the GNU Lesser General Public License
* version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY.
*/
#ifndef PE_INTERNAL__H
# define PE_INTERNAL__H
# include
# include
# include
# include
# define pe_rsc_info(rsc, fmt, args...) crm_log_tag(LOG_INFO, rsc ? rsc->id : "", fmt, ##args)
# define pe_rsc_debug(rsc, fmt, args...) crm_log_tag(LOG_DEBUG, rsc ? rsc->id : "", fmt, ##args)
# define pe_rsc_trace(rsc, fmt, args...) crm_log_tag(LOG_TRACE, rsc ? rsc->id : "", fmt, ##args)
# define pe_err(fmt...) { was_processing_error = TRUE; crm_config_error = TRUE; crm_err(fmt); }
# define pe_warn(fmt...) { was_processing_warning = TRUE; crm_config_warning = TRUE; crm_warn(fmt); }
# define pe_proc_err(fmt...) { was_processing_error = TRUE; crm_err(fmt); }
# define pe_proc_warn(fmt...) { was_processing_warning = TRUE; crm_warn(fmt); }
# define pe_set_action_bit(action, bit) action->flags = crm_set_bit(__FUNCTION__, __LINE__, action->uuid, action->flags, bit)
# define pe_clear_action_bit(action, bit) action->flags = crm_clear_bit(__FUNCTION__, __LINE__, action->uuid, action->flags, bit)
typedef struct pe__location_constraint_s {
char *id; // Constraint XML ID
pe_resource_t *rsc_lh; // Resource being located
enum rsc_role_e role_filter; // Role to locate
enum pe_discover_e discover_mode; // Resource discovery
GListPtr node_list_rh; // List of pe_node_t*
} pe__location_t;
typedef struct pe__order_constraint_s {
int id;
enum pe_ordering type;
void *lh_opaque;
pe_resource_t *lh_rsc;
pe_action_t *lh_action;
char *lh_action_task;
void *rh_opaque;
pe_resource_t *rh_rsc;
pe_action_t *rh_action;
char *rh_action_task;
} pe__ordering_t;
typedef struct notify_data_s {
GSList *keys; // Environment variable name/value pairs
const char *action;
pe_action_t *pre;
pe_action_t *post;
pe_action_t *pre_done;
pe_action_t *post_done;
GListPtr active; /* notify_entry_t* */
GListPtr inactive; /* notify_entry_t* */
GListPtr start; /* notify_entry_t* */
GListPtr stop; /* notify_entry_t* */
GListPtr demote; /* notify_entry_t* */
GListPtr promote; /* notify_entry_t* */
GListPtr master; /* notify_entry_t* */
GListPtr slave; /* notify_entry_t* */
GHashTable *allowed_nodes;
} notify_data_t;
bool pe_can_fence(pe_working_set_t *data_set, pe_node_t *node);
int merge_weights(int w1, int w2);
void add_hash_param(GHashTable * hash, const char *name, const char *value);
char *native_parameter(pe_resource_t * rsc, pe_node_t * node, gboolean create, const char *name,
pe_working_set_t * data_set);
pe_node_t *native_location(const pe_resource_t *rsc, GList **list, int current);
void pe_metadata(void);
void verify_pe_options(GHashTable * options);
void common_update_score(pe_resource_t * rsc, const char *id, int score);
void native_add_running(pe_resource_t * rsc, pe_node_t * node, pe_working_set_t * data_set);
gboolean native_unpack(pe_resource_t * rsc, pe_working_set_t * data_set);
gboolean group_unpack(pe_resource_t * rsc, pe_working_set_t * data_set);
gboolean clone_unpack(pe_resource_t * rsc, pe_working_set_t * data_set);
gboolean pe__unpack_bundle(pe_resource_t *rsc, pe_working_set_t *data_set);
pe_resource_t *native_find_rsc(pe_resource_t *rsc, const char *id, const pe_node_t *node,
int flags);
gboolean native_active(pe_resource_t * rsc, gboolean all);
gboolean group_active(pe_resource_t * rsc, gboolean all);
gboolean clone_active(pe_resource_t * rsc, gboolean all);
gboolean pe__bundle_active(pe_resource_t *rsc, gboolean all);
void native_print(pe_resource_t * rsc, const char *pre_text, long options, void *print_data);
void group_print(pe_resource_t * rsc, const char *pre_text, long options, void *print_data);
void clone_print(pe_resource_t * rsc, const char *pre_text, long options, void *print_data);
void pe__print_bundle(pe_resource_t *rsc, const char *pre_text, long options,
void *print_data);
int pe__name_and_nvpairs_xml(pcmk__output_t *out, bool is_list, const char *tag_name
, size_t pairs_count, ...);
char *pe__node_display_name(pe_node_t *node, bool print_detail);
int pe__ban_html(pcmk__output_t *out, va_list args);
int pe__ban_text(pcmk__output_t *out, va_list args);
int pe__ban_xml(pcmk__output_t *out, va_list args);
int pe__clone_xml(pcmk__output_t *out, va_list args);
int pe__clone_html(pcmk__output_t *out, va_list args);
int pe__clone_text(pcmk__output_t *out, va_list args);
int pe__cluster_counts_html(pcmk__output_t *out, va_list args);
int pe__cluster_counts_text(pcmk__output_t *out, va_list args);
int pe__cluster_counts_xml(pcmk__output_t *out, va_list args);
int pe__cluster_dc_html(pcmk__output_t *out, va_list args);
int pe__cluster_dc_text(pcmk__output_t *out, va_list args);
int pe__cluster_dc_xml(pcmk__output_t *out, va_list args);
int pe__cluster_maint_mode_html(pcmk__output_t *out, va_list args);
int pe__cluster_maint_mode_text(pcmk__output_t *out, va_list args);
int pe__cluster_options_html(pcmk__output_t *out, va_list args);
int pe__cluster_options_log(pcmk__output_t *out, va_list args);
int pe__cluster_options_text(pcmk__output_t *out, va_list args);
int pe__cluster_options_xml(pcmk__output_t *out, va_list args);
int pe__cluster_stack_html(pcmk__output_t *out, va_list args);
int pe__cluster_stack_text(pcmk__output_t *out, va_list args);
int pe__cluster_stack_xml(pcmk__output_t *out, va_list args);
int pe__cluster_summary(pcmk__output_t *out, va_list args);
int pe__cluster_summary_html(pcmk__output_t *out, va_list args);
int pe__cluster_times_html(pcmk__output_t *out, va_list args);
int pe__cluster_times_xml(pcmk__output_t *out, va_list args);
int pe__cluster_times_text(pcmk__output_t *out, va_list args);
int pe__failed_action_text(pcmk__output_t *out, va_list args);
int pe__failed_action_xml(pcmk__output_t *out, va_list args);
int pe__group_xml(pcmk__output_t *out, va_list args);
int pe__group_html(pcmk__output_t *out, va_list args);
int pe__group_text(pcmk__output_t *out, va_list args);
int pe__bundle_xml(pcmk__output_t *out, va_list args);
int pe__bundle_html(pcmk__output_t *out, va_list args);
int pe__bundle_text(pcmk__output_t *out, va_list args);
int pe__node_html(pcmk__output_t *out, va_list args);
int pe__node_text(pcmk__output_t *out, va_list args);
int pe__node_xml(pcmk__output_t *out, va_list args);
int pe__node_attribute_html(pcmk__output_t *out, va_list args);
int pe__node_attribute_text(pcmk__output_t *out, va_list args);
int pe__node_attribute_xml(pcmk__output_t *out, va_list args);
int pe__op_history_text(pcmk__output_t *out, va_list args);
int pe__op_history_xml(pcmk__output_t *out, va_list args);
int pe__resource_history_text(pcmk__output_t *out, va_list args);
int pe__resource_history_xml(pcmk__output_t *out, va_list args);
int pe__resource_xml(pcmk__output_t *out, va_list args);
int pe__resource_html(pcmk__output_t *out, va_list args);
int pe__resource_text(pcmk__output_t *out, va_list args);
int pe__ticket_html(pcmk__output_t *out, va_list args);
int pe__ticket_text(pcmk__output_t *out, va_list args);
int pe__ticket_xml(pcmk__output_t *out, va_list args);
void native_free(pe_resource_t * rsc);
void group_free(pe_resource_t * rsc);
void clone_free(pe_resource_t * rsc);
void pe__free_bundle(pe_resource_t *rsc);
enum rsc_role_e native_resource_state(const pe_resource_t * rsc, gboolean current);
enum rsc_role_e group_resource_state(const pe_resource_t * rsc, gboolean current);
enum rsc_role_e clone_resource_state(const pe_resource_t * rsc, gboolean current);
enum rsc_role_e pe__bundle_resource_state(const pe_resource_t *rsc,
gboolean current);
void pe__count_common(pe_resource_t *rsc);
void pe__count_bundle(pe_resource_t *rsc);
gboolean common_unpack(xmlNode * xml_obj, pe_resource_t ** rsc, pe_resource_t * parent,
pe_working_set_t * data_set);
void common_free(pe_resource_t * rsc);
extern pe_node_t *node_copy(const pe_node_t *this_node);
extern time_t get_effective_time(pe_working_set_t * data_set);
/* Failure handling utilities (from failcounts.c) */
// bit flags for fail count handling options
enum pe_fc_flags_e {
pe_fc_default = 0x00,
pe_fc_effective = 0x01, // don't count expired failures
pe_fc_fillers = 0x02, // if container, include filler failures in count
};
int pe_get_failcount(pe_node_t *node, pe_resource_t *rsc, time_t *last_failure,
uint32_t flags, xmlNode *xml_op,
pe_working_set_t *data_set);
pe_action_t *pe__clear_failcount(pe_resource_t *rsc, pe_node_t *node,
const char *reason,
pe_working_set_t *data_set);
/* Functions for finding/counting a resource's active nodes */
pe_node_t *pe__find_active_on(const pe_resource_t *rsc,
unsigned int *count_all,
unsigned int *count_clean);
pe_node_t *pe__find_active_requires(const pe_resource_t *rsc,
unsigned int *count);
static inline pe_node_t *
pe__current_node(const pe_resource_t *rsc)
{
return pe__find_active_on(rsc, NULL, NULL);
}
/* Binary like operators for lists of nodes */
extern void node_list_exclude(GHashTable * list, GListPtr list2, gboolean merge_scores);
extern GListPtr node_list_dup(GListPtr list, gboolean reset, gboolean filter);
extern GHashTable *node_hash_from_list(GListPtr list);
static inline gpointer
pe_hash_table_lookup(GHashTable * hash, gconstpointer key)
{
if (hash) {
return g_hash_table_lookup(hash, key);
}
return NULL;
}
extern pe_action_t *get_pseudo_op(const char *name, pe_working_set_t * data_set);
extern gboolean order_actions(pe_action_t * lh_action, pe_action_t * rh_action, enum pe_ordering order);
GHashTable *node_hash_dup(GHashTable * hash);
/* Printing functions for debug */
extern void print_node(const char *pre_text, pe_node_t * node, gboolean details);
extern void print_str_str(gpointer key, gpointer value, gpointer user_data);
extern void pe__output_node(pe_node_t * node, gboolean details, pcmk__output_t *out);
extern void dump_node_capacity(int level, const char *comment, pe_node_t * node);
extern void dump_rsc_utilization(int level, const char *comment, pe_resource_t * rsc, pe_node_t * node);
void pe__show_node_weights_as(const char *file, const char *function,
int line, bool to_log, pe_resource_t *rsc,
const char *comment, GHashTable *nodes);
#define pe__show_node_weights(level, rsc, text, nodes) \
pe__show_node_weights_as(__FILE__, __FUNCTION__, __LINE__, \
(level), (rsc), (text), (nodes))
/* Sorting functions */
extern gint sort_rsc_priority(gconstpointer a, gconstpointer b);
extern gint sort_rsc_index(gconstpointer a, gconstpointer b);
extern xmlNode *find_rsc_op_entry(pe_resource_t * rsc, const char *key);
extern pe_action_t *custom_action(pe_resource_t * rsc, char *key, const char *task, pe_node_t * on_node,
gboolean optional, gboolean foo, pe_working_set_t * data_set);
# define delete_key(rsc) pcmk__op_key(rsc->id, CRMD_ACTION_DELETE, 0)
# define delete_action(rsc, node, optional) custom_action( \
rsc, delete_key(rsc), CRMD_ACTION_DELETE, node, \
optional, TRUE, data_set);
# define stopped_key(rsc) pcmk__op_key(rsc->id, CRMD_ACTION_STOPPED, 0)
# define stopped_action(rsc, node, optional) custom_action( \
rsc, stopped_key(rsc), CRMD_ACTION_STOPPED, node, \
optional, TRUE, data_set);
# define stop_key(rsc) pcmk__op_key(rsc->id, CRMD_ACTION_STOP, 0)
# define stop_action(rsc, node, optional) custom_action( \
rsc, stop_key(rsc), CRMD_ACTION_STOP, node, \
optional, TRUE, data_set);
# define reload_key(rsc) pcmk__op_key(rsc->id, CRMD_ACTION_RELOAD, 0)
# define start_key(rsc) pcmk__op_key(rsc->id, CRMD_ACTION_START, 0)
# define start_action(rsc, node, optional) custom_action( \
rsc, start_key(rsc), CRMD_ACTION_START, node, \
optional, TRUE, data_set)
# define started_key(rsc) pcmk__op_key(rsc->id, CRMD_ACTION_STARTED, 0)
# define started_action(rsc, node, optional) custom_action( \
rsc, started_key(rsc), CRMD_ACTION_STARTED, node, \
optional, TRUE, data_set)
# define promote_key(rsc) pcmk__op_key(rsc->id, CRMD_ACTION_PROMOTE, 0)
# define promote_action(rsc, node, optional) custom_action( \
rsc, promote_key(rsc), CRMD_ACTION_PROMOTE, node, \
optional, TRUE, data_set)
# define promoted_key(rsc) pcmk__op_key(rsc->id, CRMD_ACTION_PROMOTED, 0)
# define promoted_action(rsc, node, optional) custom_action( \
rsc, promoted_key(rsc), CRMD_ACTION_PROMOTED, node, \
optional, TRUE, data_set)
# define demote_key(rsc) pcmk__op_key(rsc->id, CRMD_ACTION_DEMOTE, 0)
# define demote_action(rsc, node, optional) custom_action( \
rsc, demote_key(rsc), CRMD_ACTION_DEMOTE, node, \
optional, TRUE, data_set)
# define demoted_key(rsc) pcmk__op_key(rsc->id, CRMD_ACTION_DEMOTED, 0)
# define demoted_action(rsc, node, optional) custom_action( \
rsc, demoted_key(rsc), CRMD_ACTION_DEMOTED, node, \
optional, TRUE, data_set)
extern int pe_get_configured_timeout(pe_resource_t *rsc, const char *action,
pe_working_set_t *data_set);
extern pe_action_t *find_first_action(GListPtr input, const char *uuid, const char *task,
pe_node_t * on_node);
extern enum action_tasks get_complex_task(pe_resource_t * rsc, const char *name,
gboolean allow_non_atomic);
extern GListPtr find_actions(GListPtr input, const char *key, const pe_node_t *on_node);
GList *find_actions_exact(GList *input, const char *key,
const pe_node_t *on_node);
extern GListPtr find_recurring_actions(GListPtr input, pe_node_t * not_on_node);
GList *pe__resource_actions(const pe_resource_t *rsc, const pe_node_t *node,
const char *task, bool require_node);
extern void pe_free_action(pe_action_t * action);
extern void resource_location(pe_resource_t * rsc, pe_node_t * node, int score, const char *tag,
pe_working_set_t * data_set);
extern gint sort_op_by_callid(gconstpointer a, gconstpointer b);
extern gboolean get_target_role(pe_resource_t * rsc, enum rsc_role_e *role);
extern pe_resource_t *find_clone_instance(pe_resource_t * rsc, const char *sub_id,
pe_working_set_t * data_set);
extern void destroy_ticket(gpointer data);
extern pe_ticket_t *ticket_new(const char *ticket_id, pe_working_set_t * data_set);
// Resources for manipulating resource names
const char *pe_base_name_end(const char *id);
char *clone_strip(const char *last_rsc_id);
char *clone_zero(const char *last_rsc_id);
static inline bool
pe_base_name_eq(pe_resource_t *rsc, const char *id)
{
if (id && rsc && rsc->id) {
// Number of characters in rsc->id before any clone suffix
size_t base_len = pe_base_name_end(rsc->id) - rsc->id + 1;
return (strlen(id) == base_len) && !strncmp(id, rsc->id, base_len);
}
return FALSE;
}
int pe__target_rc_from_xml(xmlNode *xml_op);
gint sort_node_uname(gconstpointer a, gconstpointer b);
bool is_set_recursive(pe_resource_t * rsc, long long flag, bool any);
enum rsc_digest_cmp_val {
/*! Digests are the same */
RSC_DIGEST_MATCH = 0,
/*! Params that require a restart changed */
RSC_DIGEST_RESTART,
/*! Some parameter changed. */
RSC_DIGEST_ALL,
/*! rsc op didn't have a digest associated with it, so
* it is unknown if parameters changed or not. */
RSC_DIGEST_UNKNOWN,
};
typedef struct op_digest_cache_s {
enum rsc_digest_cmp_val rc;
xmlNode *params_all;
xmlNode *params_secure;
xmlNode *params_restart;
char *digest_all_calc;
char *digest_secure_calc;
char *digest_restart_calc;
} op_digest_cache_t;
op_digest_cache_t *rsc_action_digest_cmp(pe_resource_t * rsc, xmlNode * xml_op, pe_node_t * node,
pe_working_set_t * data_set);
-pe_action_t *pe_fence_op(pe_node_t * node, const char *op, bool optional, const char *reason, pe_working_set_t * data_set);
+pe_action_t *pe_fence_op(pe_node_t * node, const char *op, bool optional, const char *reason, bool priority_delay, pe_working_set_t * data_set);
void trigger_unfencing(
pe_resource_t * rsc, pe_node_t *node, const char *reason, pe_action_t *dependency, pe_working_set_t * data_set);
void pe_action_set_reason(pe_action_t *action, const char *reason, bool overwrite);
void pe_action_set_flag_reason(const char *function, long line, pe_action_t *action, pe_action_t *reason, const char *text, enum pe_action_flags flags, bool overwrite);
#define pe_action_required(action, reason, text) pe_action_set_flag_reason(__FUNCTION__, __LINE__, action, reason, text, pe_action_optional, FALSE)
#define pe_action_implies(action, reason, flag) pe_action_set_flag_reason(__FUNCTION__, __LINE__, action, reason, NULL, flag, FALSE)
void set_bit_recursive(pe_resource_t * rsc, unsigned long long flag);
void clear_bit_recursive(pe_resource_t * rsc, unsigned long long flag);
gboolean add_tag_ref(GHashTable * tags, const char * tag_name, const char * obj_ref);
void print_rscs_brief(GListPtr rsc_list, const char * pre_text, long options,
void * print_data, gboolean print_all);
int pe__rscs_brief_output(pcmk__output_t *out, GListPtr rsc_list, long options, gboolean print_all);
-void pe_fence_node(pe_working_set_t * data_set, pe_node_t * node, const char *reason);
+void pe_fence_node(pe_working_set_t * data_set, pe_node_t * node, const char *reason, bool priority_delay);
pe_node_t *pe_create_node(const char *id, const char *uname, const char *type,
const char *score, pe_working_set_t * data_set);
void common_print(pe_resource_t * rsc, const char *pre_text, const char *name, pe_node_t *node, long options, void *print_data);
int pe__common_output_text(pcmk__output_t *out, pe_resource_t * rsc, const char *name, pe_node_t *node, long options);
int pe__common_output_html(pcmk__output_t *out, pe_resource_t * rsc, const char *name, pe_node_t *node, long options);
pe_resource_t *pe__find_bundle_replica(const pe_resource_t *bundle,
const pe_node_t *node);
bool pe__bundle_needs_remote_name(pe_resource_t *rsc);
const char *pe__add_bundle_remote_name(pe_resource_t *rsc, xmlNode *xml,
const char *field);
const char *pe_node_attribute_calculated(const pe_node_t *node,
const char *name,
const pe_resource_t *rsc);
const char *pe_node_attribute_raw(pe_node_t *node, const char *name);
bool pe__is_universal_clone(pe_resource_t *rsc,
pe_working_set_t *data_set);
void pe__add_param_check(xmlNode *rsc_op, pe_resource_t *rsc, pe_node_t *node,
enum pe_check_parameters, pe_working_set_t *data_set);
void pe__foreach_param_check(pe_working_set_t *data_set,
void (*cb)(pe_resource_t*, pe_node_t*, xmlNode*,
enum pe_check_parameters,
pe_working_set_t*));
void pe__free_param_checks(pe_working_set_t *data_set);
bool pe__shutdown_requested(pe_node_t *node);
void pe__update_recheck_time(time_t recheck, pe_working_set_t *data_set);
#define BOOL2STR(x) ((x) ? "true" : "false")
/*!
* \internal
* \brief Register xml formatting message functions.
*/
void pe__register_messages(pcmk__output_t *out);
void pe__unpack_dataset_nvpairs(xmlNode *xml_obj, const char *set_name,
GHashTable *node_hash, GHashTable *hash,
const char *always_first, gboolean overwrite,
pe_working_set_t *data_set);
bool pe__resource_is_disabled(pe_resource_t *rsc);
pe_action_t *pe__clear_resource_history(pe_resource_t *rsc, pe_node_t *node,
pe_working_set_t *data_set);
#endif
diff --git a/include/crm/pengine/pe_types.h b/include/crm/pengine/pe_types.h
index a6060e59cf..a8974f8fd0 100644
--- a/include/crm/pengine/pe_types.h
+++ b/include/crm/pengine/pe_types.h
@@ -1,534 +1,536 @@
/*
* Copyright 2004-2020 the Pacemaker project contributors
*
* The version control history for this file may have further details.
*
* This source code is licensed under the GNU Lesser General Public License
* version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY.
*/
#ifndef PE_TYPES__H
# define PE_TYPES__H
#ifdef __cplusplus
extern "C" {
#endif
/*!
* \file
* \brief Data types for cluster status
* \ingroup pengine
*/
# include // bool
# include // time_t
# include // gboolean, guint, GList, GHashTable
# include // GListPtr
# include
# include
typedef struct pe_node_s pe_node_t;
typedef struct pe_action_s pe_action_t;
typedef struct pe_resource_s pe_resource_t;
typedef struct pe_working_set_s pe_working_set_t;
enum pe_obj_types {
pe_unknown = -1,
pe_native = 0,
pe_group = 1,
pe_clone = 2,
pe_container = 3,
};
typedef struct resource_object_functions_s {
gboolean (*unpack) (pe_resource_t*, pe_working_set_t*);
pe_resource_t *(*find_rsc) (pe_resource_t *parent, const char *search,
const pe_node_t *node, int flags);
/* parameter result must be free'd */
char *(*parameter) (pe_resource_t*, pe_node_t*, gboolean, const char*,
pe_working_set_t*);
void (*print) (pe_resource_t*, const char*, long, void*);
gboolean (*active) (pe_resource_t*, gboolean);
enum rsc_role_e (*state) (const pe_resource_t*, gboolean);
pe_node_t *(*location) (const pe_resource_t*, GList**, int);
void (*free) (pe_resource_t*);
void (*count) (pe_resource_t*);
} resource_object_functions_t;
typedef struct resource_alloc_functions_s resource_alloc_functions_t;
enum pe_quorum_policy {
no_quorum_freeze,
no_quorum_stop,
no_quorum_ignore,
no_quorum_suicide
};
enum node_type {
node_ping,
node_member,
node_remote
};
//! \deprecated will be removed in a future release
enum pe_restart {
pe_restart_restart,
pe_restart_ignore
};
//! Determine behavior of pe_find_resource_with_flags()
enum pe_find {
pe_find_renamed = 0x001, //!< match resource ID or LRM history ID
pe_find_anon = 0x002, //!< match base name of anonymous clone instances
pe_find_clone = 0x004, //!< match only clone instances
pe_find_current = 0x008, //!< match resource active on specified node
pe_find_inactive = 0x010, //!< match resource not running anywhere
pe_find_any = 0x020, //!< match base name of any clone instance
};
// @TODO Make these an enum
# define pe_flag_have_quorum 0x00000001ULL
# define pe_flag_symmetric_cluster 0x00000002ULL
# define pe_flag_maintenance_mode 0x00000008ULL
# define pe_flag_stonith_enabled 0x00000010ULL
# define pe_flag_have_stonith_resource 0x00000020ULL
# define pe_flag_enable_unfencing 0x00000040ULL
# define pe_flag_concurrent_fencing 0x00000080ULL
# define pe_flag_stop_rsc_orphans 0x00000100ULL
# define pe_flag_stop_action_orphans 0x00000200ULL
# define pe_flag_stop_everything 0x00000400ULL
# define pe_flag_start_failure_fatal 0x00001000ULL
# define pe_flag_remove_after_stop 0x00002000ULL
# define pe_flag_startup_fencing 0x00004000ULL
# define pe_flag_shutdown_lock 0x00008000ULL
# define pe_flag_startup_probes 0x00010000ULL
# define pe_flag_have_status 0x00020000ULL
# define pe_flag_have_remote_nodes 0x00040000ULL
# define pe_flag_quick_location 0x00100000ULL
# define pe_flag_sanitized 0x00200000ULL
# define pe_flag_stdout 0x00400000ULL
//! Don't count total, disabled and blocked resource instances
# define pe_flag_no_counts 0x00800000ULL
/*! Skip deprecated code that is kept solely for backward API compatibility.
* (Internal code should always set this.)
*/
# define pe_flag_no_compat 0x01000000ULL
struct pe_working_set_s {
xmlNode *input;
crm_time_t *now;
/* options extracted from the input */
char *dc_uuid;
pe_node_t *dc_node;
const char *stonith_action;
const char *placement_strategy;
unsigned long long flags;
int stonith_timeout;
enum pe_quorum_policy no_quorum_policy;
GHashTable *config_hash;
GHashTable *tickets;
// Actions for which there can be only one (e.g. fence nodeX)
GHashTable *singletons;
GListPtr nodes;
GListPtr resources;
GListPtr placement_constraints;
GListPtr ordering_constraints;
GListPtr colocation_constraints;
GListPtr ticket_constraints;
GListPtr actions;
xmlNode *failed;
xmlNode *op_defaults;
xmlNode *rsc_defaults;
/* stats */
int num_synapse;
int max_valid_nodes; //! Deprecated (will be removed in a future release)
int order_id;
int action_id;
/* final output */
xmlNode *graph;
GHashTable *template_rsc_sets;
const char *localhost;
GHashTable *tags;
int blocked_resources;
int disabled_resources;
GList *param_check; // History entries that need to be checked
GList *stop_needed; // Containers that need stop actions
time_t recheck_by; // Hint to controller to re-run scheduler by this time
int ninstances; // Total number of resource instances
guint shutdown_lock;// How long (seconds) to lock resources to shutdown node
+ int priority_fencing_delay; // Enforced priority fencing delay
};
enum pe_check_parameters {
/* Clear fail count if parameters changed for un-expired start or monitor
* last_failure.
*/
pe_check_last_failure,
/* Clear fail count if parameters changed for start, monitor, promote, or
* migrate_from actions for active resources.
*/
pe_check_active,
};
struct pe_node_shared_s {
const char *id;
const char *uname;
enum node_type type;
/* @TODO convert these flags into a bitfield */
gboolean online;
gboolean standby;
gboolean standby_onfail;
gboolean pending;
gboolean unclean;
gboolean unseen;
gboolean shutdown;
gboolean expected_up;
gboolean is_dc;
gboolean maintenance;
gboolean rsc_discovery_enabled;
gboolean remote_requires_reset;
gboolean remote_was_fenced;
gboolean remote_maintenance; /* what the remote-rsc is thinking */
gboolean unpacked;
int num_resources;
pe_resource_t *remote_rsc;
GListPtr running_rsc; /* pe_resource_t* */
GListPtr allocated_rsc; /* pe_resource_t* */
GHashTable *attrs; /* char* => char* */
GHashTable *utilization;
GHashTable *digest_cache; //!< cache of calculated resource digests
+ int priority; // calculated based on the priority of resources running on the node
};
struct pe_node_s {
int weight;
gboolean fixed;
int count;
struct pe_node_shared_s *details;
int rsc_discover_mode;
};
# define pe_rsc_orphan 0x00000001ULL
# define pe_rsc_managed 0x00000002ULL
# define pe_rsc_block 0x00000004ULL
# define pe_rsc_orphan_container_filler 0x00000008ULL
# define pe_rsc_notify 0x00000010ULL
# define pe_rsc_unique 0x00000020ULL
# define pe_rsc_fence_device 0x00000040ULL
# define pe_rsc_promotable 0x00000080ULL
# define pe_rsc_provisional 0x00000100ULL
# define pe_rsc_allocating 0x00000200ULL
# define pe_rsc_merging 0x00000400ULL
# define pe_rsc_reload 0x00002000ULL
# define pe_rsc_allow_remote_remotes 0x00004000ULL
# define pe_rsc_failed 0x00010000ULL
# define pe_rsc_runnable 0x00040000ULL
# define pe_rsc_start_pending 0x00080000ULL
# define pe_rsc_starting 0x00100000ULL
# define pe_rsc_stopping 0x00200000ULL
# define pe_rsc_allow_migrate 0x00800000ULL
# define pe_rsc_failure_ignored 0x01000000ULL
# define pe_rsc_maintenance 0x04000000ULL
# define pe_rsc_is_container 0x08000000ULL
# define pe_rsc_needs_quorum 0x10000000ULL
# define pe_rsc_needs_fencing 0x20000000ULL
# define pe_rsc_needs_unfencing 0x40000000ULL
enum pe_graph_flags {
pe_graph_none = 0x00000,
pe_graph_updated_first = 0x00001,
pe_graph_updated_then = 0x00002,
pe_graph_disable = 0x00004,
};
/* *INDENT-OFF* */
enum pe_action_flags {
pe_action_pseudo = 0x00001,
pe_action_runnable = 0x00002,
pe_action_optional = 0x00004,
pe_action_print_always = 0x00008,
pe_action_have_node_attrs = 0x00010,
pe_action_implied_by_stonith = 0x00040,
pe_action_migrate_runnable = 0x00080,
pe_action_dumped = 0x00100,
pe_action_processed = 0x00200,
pe_action_clear = 0x00400,
pe_action_dangle = 0x00800,
/* This action requires one or more of its dependencies to be runnable.
* We use this to clear the runnable flag before checking dependencies.
*/
pe_action_requires_any = 0x01000,
pe_action_reschedule = 0x02000,
pe_action_tracking = 0x04000,
pe_action_dedup = 0x08000, //! Internal state tracking when creating graph
pe_action_dc = 0x10000, //! Action may run on DC instead of target
};
/* *INDENT-ON* */
struct pe_resource_s {
char *id;
char *clone_name;
xmlNode *xml;
xmlNode *orig_xml;
xmlNode *ops_xml;
pe_working_set_t *cluster;
pe_resource_t *parent;
enum pe_obj_types variant;
void *variant_opaque;
resource_object_functions_t *fns;
resource_alloc_functions_t *cmds;
enum rsc_recovery_type recovery_type;
// @TODO only pe_restart_restart is of interest, so merge into flags
enum pe_restart restart_type; //!< \deprecated will be removed in future release
int priority;
int stickiness;
int sort_index;
int failure_timeout;
int migration_threshold;
guint remote_reconnect_ms;
char *pending_task;
unsigned long long flags;
// @TODO merge these into flags
gboolean is_remote_node;
gboolean exclusive_discover;
//!@{
//! This field should be treated as internal to Pacemaker
GListPtr rsc_cons_lhs; // List of rsc_colocation_t*
GListPtr rsc_cons; // List of rsc_colocation_t*
GListPtr rsc_location; // List of pe__location_t*
GListPtr actions; // List of pe_action_t*
GListPtr rsc_tickets; // List of rsc_ticket*
//!@}
pe_node_t *allocated_to;
pe_node_t *partial_migration_target;
pe_node_t *partial_migration_source;
GListPtr running_on; /* pe_node_t* */
GHashTable *known_on; /* pe_node_t* */
GHashTable *allowed_nodes; /* pe_node_t* */
enum rsc_role_e role;
enum rsc_role_e next_role;
GHashTable *meta;
GHashTable *parameters;
GHashTable *utilization;
GListPtr children; /* pe_resource_t* */
GListPtr dangling_migrations; /* pe_node_t* */
pe_resource_t *container;
GListPtr fillers;
pe_node_t *pending_node; // Node on which pending_task is happening
pe_node_t *lock_node; // Resource is shutdown-locked to this node
time_t lock_time; // When shutdown lock started
#if ENABLE_VERSIONED_ATTRS
xmlNode *versioned_parameters;
#endif
};
#if ENABLE_VERSIONED_ATTRS
// Used as action->action_details if action->rsc is not NULL
typedef struct pe_rsc_action_details_s {
xmlNode *versioned_parameters;
xmlNode *versioned_meta;
} pe_rsc_action_details_t;
#endif
struct pe_action_s {
int id;
int priority;
pe_resource_t *rsc;
pe_node_t *node;
xmlNode *op_entry;
char *task;
char *uuid;
char *cancel_task;
char *reason;
enum pe_action_flags flags;
enum rsc_start_requirement needs;
enum action_fail_response on_fail;
enum rsc_role_e fail_role;
GHashTable *meta;
GHashTable *extra;
/*
* These two varables are associated with the constraint logic
* that involves first having one or more actions runnable before
* then allowing this action to execute.
*
* These varables are used with features such as 'clone-min' which
* requires at minimum X number of cloned instances to be running
* before an order dependency can run. Another option that uses
* this is 'require-all=false' in ordering constrants. This option
* says "only require one instance of a resource to start before
* allowing dependencies to start" -- basically, require-all=false is
* the same as clone-min=1.
*/
/* current number of known runnable actions in the before list. */
int runnable_before;
/* the number of "before" runnable actions required for this action
* to be considered runnable */
int required_runnable_before;
GListPtr actions_before; /* pe_action_wrapper_t* */
GListPtr actions_after; /* pe_action_wrapper_t* */
/* Some of the above fields could be moved to the details,
* except for API backward compatibility.
*/
void *action_details; // varies by type of action
};
typedef struct pe_ticket_s {
char *id;
gboolean granted;
time_t last_granted;
gboolean standby;
GHashTable *state;
} pe_ticket_t;
typedef struct pe_tag_s {
char *id;
GListPtr refs;
} pe_tag_t;
//! Internal tracking for transition graph creation
enum pe_link_state {
pe_link_not_dumped, //! Internal tracking for transition graph creation
pe_link_dumped, //! Internal tracking for transition graph creation
pe_link_dup, //! \deprecated No longer used by Pacemaker
};
enum pe_discover_e {
pe_discover_always = 0,
pe_discover_never,
pe_discover_exclusive,
};
/* *INDENT-OFF* */
enum pe_ordering {
pe_order_none = 0x0, /* deleted */
pe_order_optional = 0x1, /* pure ordering, nothing implied */
pe_order_apply_first_non_migratable = 0x2, /* Only apply this constraint's ordering if first is not migratable. */
pe_order_implies_first = 0x10, /* If 'then' is required, ensure 'first' is too */
pe_order_implies_then = 0x20, /* If 'first' is required, ensure 'then' is too */
pe_order_implies_first_master = 0x40, /* Imply 'first' is required when 'then' is required and then's rsc holds Master role. */
/* first requires then to be both runnable and migrate runnable. */
pe_order_implies_first_migratable = 0x80,
pe_order_runnable_left = 0x100, /* 'then' requires 'first' to be runnable */
pe_order_pseudo_left = 0x200, /* 'then' can only be pseudo if 'first' is runnable */
pe_order_implies_then_on_node = 0x400, /* If 'first' is required on 'nodeX',
* ensure instances of 'then' on 'nodeX' are too.
* Only really useful if 'then' is a clone and 'first' is not
*/
pe_order_probe = 0x800, /* If 'first->rsc' is
* - running but about to stop, ignore the constraint
* - otherwise, behave as runnable_left
*/
pe_order_restart = 0x1000, /* 'then' is runnable if 'first' is optional or runnable */
pe_order_stonith_stop = 0x2000, /* only applies if the action is non-pseudo */
pe_order_serialize_only = 0x4000, /* serialize */
pe_order_same_node = 0x8000, /* applies only if 'first' and 'then' are on same node */
pe_order_implies_first_printed = 0x10000, /* Like ..implies_first but only ensures 'first' is printed, not mandatory */
pe_order_implies_then_printed = 0x20000, /* Like ..implies_then but only ensures 'then' is printed, not mandatory */
pe_order_asymmetrical = 0x100000, /* Indicates asymmetrical one way ordering constraint. */
pe_order_load = 0x200000, /* Only relevant if... */
pe_order_one_or_more = 0x400000, /* 'then' is runnable only if one or more of its dependencies are too */
pe_order_anti_colocation = 0x800000,
pe_order_preserve = 0x1000000, /* Hack for breaking user ordering constraints with container resources */
pe_order_then_cancels_first = 0x2000000, // if 'then' becomes required, 'first' becomes optional
pe_order_trace = 0x4000000, /* test marker */
};
/* *INDENT-ON* */
typedef struct pe_action_wrapper_s {
enum pe_ordering type;
enum pe_link_state state;
pe_action_t *action;
} pe_action_wrapper_t;
#ifndef PCMK__NO_COMPAT
/* Everything here is deprecated and kept only for public API backward
* compatibility. It will be moved to compatibility.h when 2.1.0 is released.
*/
//!< \deprecated Use pe_action_t instead
typedef struct pe_action_s action_t;
//!< \deprecated Use pe_action_wrapper_t instead
typedef struct pe_action_wrapper_s action_wrapper_t;
//!< \deprecated Use pe_node_t instead
typedef struct pe_node_s node_t;
//!< \deprecated Use enum pe_quorum_policy instead
typedef enum pe_quorum_policy no_quorum_policy_t;
//!< \deprecated use pe_resource_t instead
typedef struct pe_resource_s resource_t;
//!< \deprecated Use pe_tag_t instead
typedef struct pe_tag_s tag_t;
//!< \deprecated Use pe_ticket_t instead
typedef struct pe_ticket_s ticket_t;
#endif
#ifdef __cplusplus
}
#endif
#endif // PE_TYPES__H
diff --git a/include/crm/stonith-ng.h b/include/crm/stonith-ng.h
index 3637bc7aa0..06812e7675 100644
--- a/include/crm/stonith-ng.h
+++ b/include/crm/stonith-ng.h
@@ -1,562 +1,581 @@
/*
* Copyright 2004-2019 the Pacemaker project contributors
*
* The version control history for this file may have further details.
*
* This source code is licensed under the GNU Lesser General Public License
* version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY.
*/
#ifdef __cplusplus
extern "C" {
#endif
/**
* \file
* \brief Fencing aka. STONITH
* \ingroup fencing
*/
#ifndef STONITH_NG__H
# define STONITH_NG__H
# include
# include
# include // bool
# include // uint32_t
# include // time_t
# define T_STONITH_NOTIFY_DISCONNECT "st_notify_disconnect"
# define T_STONITH_NOTIFY_FENCE "st_notify_fence"
# define T_STONITH_NOTIFY_HISTORY "st_notify_history"
# define T_STONITH_NOTIFY_HISTORY_SYNCED "st_notify_history_synced"
/* *INDENT-OFF* */
enum stonith_state {
stonith_connected_command,
stonith_connected_query,
stonith_disconnected,
};
enum stonith_call_options {
st_opt_none = 0x00000000,
st_opt_verbose = 0x00000001,
st_opt_allow_suicide = 0x00000002,
st_opt_manual_ack = 0x00000008,
st_opt_discard_reply = 0x00000010,
/* st_opt_all_replies = 0x00000020, */
st_opt_topology = 0x00000040,
st_opt_scope_local = 0x00000100,
st_opt_cs_nodeid = 0x00000200,
st_opt_sync_call = 0x00001000,
/*! Allow the timeout period for a callback to be adjusted
* based on the time the server reports the operation will take. */
st_opt_timeout_updates = 0x00002000,
/*! Only report back if operation is a success in callback */
st_opt_report_only_success = 0x00004000,
/* used where ever apropriate - e.g. cleanup of history */
st_opt_cleanup = 0x000080000,
/* used where ever apropriate - e.g. send out a history query to all nodes */
st_opt_broadcast = 0x000100000,
};
/*! Order matters here, do not change values */
enum op_state
{
st_query,
st_exec,
st_done,
st_duplicate,
st_failed,
};
// Supported fence agent interface standards
enum stonith_namespace {
st_namespace_invalid,
st_namespace_any,
st_namespace_internal, // Implemented internally by Pacemaker
/* Neither of these projects are active any longer, but the fence agent
* interfaces they created are still in use and supported by Pacemaker.
*/
st_namespace_rhcs, // Red Hat Cluster Suite compatible
st_namespace_lha, // Linux-HA compatible
};
enum stonith_namespace stonith_text2namespace(const char *namespace_s);
const char *stonith_namespace2text(enum stonith_namespace st_namespace);
enum stonith_namespace stonith_get_namespace(const char *agent,
const char *namespace_s);
typedef struct stonith_key_value_s {
char *key;
char *value;
struct stonith_key_value_s *next;
} stonith_key_value_t;
typedef struct stonith_history_s {
char *target;
char *action;
char *origin;
char *delegate;
char *client;
int state;
time_t completed;
struct stonith_history_s *next;
} stonith_history_t;
typedef struct stonith_s stonith_t;
typedef struct stonith_event_s
{
char *id;
char *type;
char *message;
char *operation;
int result;
char *origin;
char *target;
char *action;
char *executioner;
char *device;
/*! The name of the client that initiated the action. */
char *client_origin;
} stonith_event_t;
typedef struct stonith_callback_data_s
{
int rc;
int call_id;
void *userdata;
} stonith_callback_data_t;
typedef struct stonith_api_operations_s
{
/*!
* \brief Destroy the stonith api structure.
*/
int (*free) (stonith_t *st);
/*!
* \brief Connect to the local stonith daemon.
*
* \retval 0, success
* \retval negative error code on failure
*/
int (*connect) (stonith_t *st, const char *name, int *stonith_fd);
/*!
* \brief Disconnect from the local stonith daemon.
*
* \retval 0, success
* \retval negative error code on failure
*/
int (*disconnect)(stonith_t *st);
/*!
* \brief Remove a registered stonith device with the local stonith daemon.
*
* \note Synchronous, guaranteed to occur in daemon before function returns.
*
* \retval 0, success
* \retval negative error code on failure
*/
int (*remove_device)(
stonith_t *st, int options, const char *name);
/*!
* \brief Register a stonith device with the local stonith daemon.
*
* \note Synchronous, guaranteed to occur in daemon before function returns.
*
* \retval 0, success
* \retval negative error code on failure
*/
int (*register_device)(
stonith_t *st, int options, const char *id,
const char *provider, const char *agent, stonith_key_value_t *params);
/*!
* \brief Remove a fencing level for a specific node.
*
* \note This feature is not available when stonith is in standalone mode.
*
* \retval 0, success
* \retval negative error code on failure
*/
int (*remove_level)(
stonith_t *st, int options, const char *node, int level);
/*!
* \brief Register a fencing level containing the fencing devices to be used
* at that level for a specific node.
*
* \note This feature is not available when stonith is in standalone mode.
*
* \retval 0, success
* \retval negative error code on failure
*/
int (*register_level)(
stonith_t *st, int options, const char *node, int level, stonith_key_value_t *device_list);
/*!
* \brief Get the metadata documentation for a resource.
*
* \note Value is returned in output. Output must be freed when set.
*
* \retval 0 success
* \retval negative error code on failure
*/
int (*metadata)(stonith_t *st, int options,
const char *device, const char *provider, char **output, int timeout);
/*!
* \brief Retrieve a list of installed stonith agents
*
* \note if provider is not provided, all known agents will be returned
* \note list must be freed using stonith_key_value_freeall()
* \note call_options parameter is not used, it is reserved for future use.
*
* \retval num items in list on success
* \retval negative error code on failure
*/
int (*list_agents)(stonith_t *stonith, int call_options, const char *provider,
stonith_key_value_t **devices, int timeout);
/*!
* \brief Retrieve string listing hosts and port assignments from a local stonith device.
*
* \retval 0 on success
* \retval negative error code on failure
*/
int (*list)(stonith_t *st, int options, const char *id, char **list_output, int timeout);
/*!
* \brief Check to see if a local stonith device is reachable
*
* \retval 0 on success
* \retval negative error code on failure
*/
int (*monitor)(stonith_t *st, int options, const char *id, int timeout);
/*!
* \brief Check to see if a local stonith device's port is reachable
*
* \retval 0 on success
* \retval negative error code on failure
*/
int (*status)(stonith_t *st, int options, const char *id, const char *port, int timeout);
/*!
* \brief Retrieve a list of registered stonith devices.
*
* \note If node is provided, only devices that can fence the node id
* will be returned.
*
* \retval num items in list on success
* \retval negative error code on failure
*/
int (*query)(stonith_t *st, int options, const char *node,
stonith_key_value_t **devices, int timeout);
/*!
* \brief Issue a fencing action against a node.
*
* \note Possible actions are, 'on', 'off', and 'reboot'.
*
* \param st, stonith connection
* \param options, call options
* \param node, The target node to fence
* \param action, The fencing action to take
* \param timeout, The default per device timeout to use with each device
* capable of fencing the target.
*
* \retval 0 success
* \retval negative error code on failure.
*/
int (*fence)(stonith_t *st, int options, const char *node, const char *action,
int timeout, int tolerance);
/*!
* \brief Manually confirm that a node is down.
*
* \retval 0 success
* \retval negative error code on failure.
*/
int (*confirm)(stonith_t *st, int options, const char *node);
/*!
* \brief Retrieve a list of fencing operations that have occurred for a specific node.
*
* \note History is not available in standalone mode.
*
* \retval 0 success
* \retval negative error code on failure.
*/
int (*history)(stonith_t *st, int options, const char *node, stonith_history_t **output, int timeout);
int (*register_notification)(
stonith_t *st, const char *event,
void (*notify)(stonith_t *st, stonith_event_t *e));
int (*remove_notification)(stonith_t *st, const char *event);
/*!
* \brief Register a callback to receive the result of an asynchronous call
*
* \param[in] call_id The call ID to register callback for
* \param[in] timeout Default time to wait until callback expires
* \param[in] options Bitmask of \c stonith_call_options (respects
* \c st_opt_timeout_updates and
* \c st_opt_report_only_success)
* \param[in] userdata Pointer that will be given to callback
* \param[in] callback_name Unique name to identify callback
* \param[in] callback The callback function to register
*
* \return \c TRUE on success, \c FALSE if call_id is negative, -errno otherwise
*
* \todo This function should return \c pcmk_ok on success, and \c call_id
* when negative, but that would break backward compatibility.
*/
int (*register_callback)(stonith_t *st,
int call_id,
int timeout,
int options,
void *userdata,
const char *callback_name,
void (*callback)(stonith_t *st, stonith_callback_data_t *data));
/*!
* \brief Remove a registered callback for a given call id.
*/
int (*remove_callback)(stonith_t *st, int call_id, bool all_callbacks);
/*!
* \brief Remove fencing level for specific node, node regex or attribute
*
* \param[in] st Fencer connection to use
* \param[in] options Bitmask of stonith_call_options to pass to the fencer
* \param[in] node If not NULL, target level by this node name
* \param[in] pattern If not NULL, target by node name using this regex
* \param[in] attr If not NULL, target by this node attribute
* \param[in] value If not NULL, target by this node attribute value
* \param[in] level Index number of level to remove
*
* \return 0 on success, negative error code otherwise
*
* \note This feature is not available when stonith is in standalone mode.
* The caller should set only one of node, pattern or attr/value.
*/
int (*remove_level_full)(stonith_t *st, int options,
const char *node, const char *pattern,
const char *attr, const char *value, int level);
/*!
* \brief Register fencing level for specific node, node regex or attribute
*
* \param[in] st Fencer connection to use
* \param[in] options Bitmask of stonith_call_options to pass to fencer
* \param[in] node If not NULL, target level by this node name
* \param[in] pattern If not NULL, target by node name using this regex
* \param[in] attr If not NULL, target by this node attribute
* \param[in] value If not NULL, target by this node attribute value
* \param[in] level Index number of level to add
* \param[in] device_list Devices to use in level
*
* \return 0 on success, negative error code otherwise
*
* \note This feature is not available when stonith is in standalone mode.
* The caller should set only one of node, pattern or attr/value.
*/
int (*register_level_full)(stonith_t *st, int options,
const char *node, const char *pattern,
const char *attr, const char *value,
int level, stonith_key_value_t *device_list);
/*!
* \brief Validate an arbitrary stonith device configuration
*
* \param[in] st Stonithd connection to use
* \param[in] call_options Bitmask of stonith_call_options to use with fencer
* \param[in] rsc_id ID used to replace CIB secrets in params
* \param[in] namespace_s Namespace of fence agent to validate (optional)
* \param[in] agent Fence agent to validate
* \param[in] params Configuration parameters to pass to fence agent
* \param[in] timeout Fail if no response within this many seconds
* \param[out] output If non-NULL, where to store any agent output
* \param[out] error_output If non-NULL, where to store agent error output
*
* \return pcmk_ok if validation succeeds, -errno otherwise
*
* \note If pcmk_ok is returned, the caller is responsible for freeing
* the output (if requested).
*/
int (*validate)(stonith_t *st, int call_options, const char *rsc_id,
const char *namespace_s, const char *agent,
stonith_key_value_t *params, int timeout, char **output,
char **error_output);
+ /*!
+ * \brief Issue a fencing action against a node with enforced fencing delay.
+ *
+ * \note Possible actions are, 'on', 'off', and 'reboot'.
+ *
+ * \param st, stonith connection
+ * \param options, call options
+ * \param node, The target node to fence
+ * \param action, The fencing action to take
+ * \param timeout, The default per device timeout to use with each device
+ * capable of fencing the target.
+ * \param delay, Any enforced fencing delay. -1 to disable
+ *
+ * \retval 0 success
+ * \retval negative error code on failure.
+ */
+ int (*fence_with_delay)(stonith_t *st, int options, const char *node, const char *action,
+ int timeout, int tolerance, int delay);
+
} stonith_api_operations_t;
struct stonith_s
{
enum stonith_state state;
int call_id;
int call_timeout;
void *st_private;
stonith_api_operations_t *cmds;
};
/* *INDENT-ON* */
/* Core functions */
stonith_t *stonith_api_new(void);
void stonith_api_delete(stonith_t * st);
void stonith_dump_pending_callbacks(stonith_t * st);
bool stonith_dispatch(stonith_t * st);
stonith_key_value_t *stonith_key_value_add(stonith_key_value_t * kvp, const char *key,
const char *value);
void stonith_key_value_freeall(stonith_key_value_t * kvp, int keys, int values);
void stonith_history_free(stonith_history_t *history);
// Convenience functions
int stonith_api_connect_retry(stonith_t *st, const char *name,
int max_attempts);
/* Basic helpers that allows nodes to be fenced and the history to be
* queried without mainloop or the caller understanding the full API
*
* At least one of nodeid and uname are required
*/
int stonith_api_kick(uint32_t nodeid, const char *uname, int timeout, bool off);
time_t stonith_api_time(uint32_t nodeid, const char *uname, bool in_progress);
/*
* Helpers for using the above functions without install-time dependencies
*
* Usage:
* #include
*
* To turn a node off by corosync nodeid:
* stonith_api_kick_helper(nodeid, 120, 1);
*
* To check the last fence date/time (also by nodeid):
* last = stonith_api_time_helper(nodeid, 0);
*
* To check if fencing is in progress:
* if(stonith_api_time_helper(nodeid, 1) > 0) { ... }
*
* eg.
#include
#include
#include
int
main(int argc, char ** argv)
{
int rc = 0;
int nodeid = 102;
rc = stonith_api_time_helper(nodeid, 0);
printf("%d last fenced at %s\n", nodeid, ctime(rc));
rc = stonith_api_kick_helper(nodeid, 120, 1);
printf("%d fence result: %d\n", nodeid, rc);
rc = stonith_api_time_helper(nodeid, 0);
printf("%d last fenced at %s\n", nodeid, ctime(rc));
return 0;
}
*/
# define STONITH_LIBRARY "libstonithd.so.26"
typedef int (*st_api_kick_fn) (int nodeid, const char *uname, int timeout, bool off);
typedef time_t (*st_api_time_fn) (int nodeid, const char *uname, bool in_progress);
static inline int
stonith_api_kick_helper(uint32_t nodeid, int timeout, bool off)
{
static void *st_library = NULL;
static st_api_kick_fn st_kick_fn;
if (st_library == NULL) {
st_library = dlopen(STONITH_LIBRARY, RTLD_LAZY);
}
if (st_library && st_kick_fn == NULL) {
st_kick_fn = (st_api_kick_fn) dlsym(st_library, "stonith_api_kick");
}
if (st_kick_fn == NULL) {
#ifdef ELIBACC
return -ELIBACC;
#else
return -ENOSYS;
#endif
}
return (*st_kick_fn) (nodeid, NULL, timeout, off);
}
static inline time_t
stonith_api_time_helper(uint32_t nodeid, bool in_progress)
{
static void *st_library = NULL;
static st_api_time_fn st_time_fn;
if (st_library == NULL) {
st_library = dlopen(STONITH_LIBRARY, RTLD_LAZY);
}
if (st_library && st_time_fn == NULL) {
st_time_fn = (st_api_time_fn) dlsym(st_library, "stonith_api_time");
}
if (st_time_fn == NULL) {
return 0;
}
return (*st_time_fn) (nodeid, NULL, in_progress);
}
/**
* Does the given agent describe a stonith resource that can exist?
*
* \param[in] agent What is the name of the agent?
* \param[in] timeout Timeout to use when querying. If 0 is given,
* use a default of 120.
*
* \return A boolean
*/
bool stonith_agent_exists(const char *agent, int timeout);
/*!
* \brief Turn stonith action into a more readable string.
*
* \param action Stonith action
*/
const char *stonith_action_str(const char *action);
#ifndef PCMK__NO_COMPAT
/* Everything here is deprecated and kept only for public API backward
* compatibility. It will be moved to compatibility.h when 2.1.0 is released.
*/
//! \deprecated Use stonith_get_namespace() instead
const char *get_stonith_provider(const char *agent, const char *provider);
#endif
#ifdef __cplusplus
}
#endif
#endif
diff --git a/include/pcmki/pcmki_fence.h b/include/pcmki/pcmki_fence.h
index 6f699e27ec..e091a17104 100644
--- a/include/pcmki/pcmki_fence.h
+++ b/include/pcmki/pcmki_fence.h
@@ -1,220 +1,222 @@
/*
* Copyright 2019 the Pacemaker project contributors
*
* The version control history for this file may have further details.
*
* This source code is licensed under the GNU Lesser General Public License
* version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY.
*/
#ifndef PCMKI_STONITH_H
# define PCMKI_STONITH_H
# include
# include
/*!
* \brief Perform a STONITH action.
*
* \note This is the internal version of pcmk_fence_action(). External users
* of the pacemaker API should use that function instead.
*
* \param[in] st A connection to the STONITH API.
* \param[in] target The node receiving the action.
* \param[in] action The action to perform.
* \param[in] name Who requested the fence action?
* \param[in] timeout How long to wait for the operation to complete (in ms).
* \param[in] tolerance If a successful action for \p target happened within
* this many ms, return 0 without performing the action
* again.
+ * \param[in] delay Enforce a fencing delay. Value -1 means disabled.
*
* \return Standard Pacemaker return code
*/
int pcmk__fence_action(stonith_t *st, const char *target, const char *action,
- const char *name, unsigned int timeout, unsigned int tolerance);
+ const char *name, unsigned int timeout, unsigned int tolerance,
+ int delay);
/*!
* \brief List the fencing operations that have occurred for a specific node.
*
* \note This is the internal version of pcmk_fence_history(). External users
* of the pacemaker API should use that function instead.
*
* \note \p out should be initialized with pcmk__output_new() before calling this
* function and destroyed with out->finish and pcmk__output_free() before
* reusing it with any other functions in this library.
*
* \param[in,out] out The output functions structure.
* \param[in] st A connection to the STONITH API.
* \param[in] target The node to get history for.
* \param[in] timeout How long to wait for the operation to complete (in ms).
* \param[in] quiet Suppress most output.
* \param[in] verbose Include additional output.
* \param[in] broadcast Gather fencing history from all nodes.
* \param[in] cleanup Clean up fencing history after listing.
*
* \return Standard Pacemaker return code
*/
int pcmk__fence_history(pcmk__output_t *out, stonith_t *st, char *target,
unsigned int timeout, bool quiet, int verbose,
bool broadcast, bool cleanup);
/**
* \brief List all installed STONITH agents.
*
* \note This is the internal version of pcmk_fence_installed(). External users
* of the pacemaker API should use that function instead.
*
* \note \p out should be initialized with pcmk__output_new() before calling this
* function and destroyed with out->finish and pcmk__output_free() before
* reusing it with any other functions in this library.
*
* \param[in,out] out The output functions structure.
* \param[in] st A connection to the STONITH API.
* \param[in] timeout How long to wait for the operation to complete (in ms).
*
* \return Standard Pacemaker return code
*/
int pcmk__fence_installed(pcmk__output_t *out, stonith_t *st, unsigned int timeout);
/*!
* \brief When was a device last fenced?
*
* \note This is the internal version of pcmk_fence_last(). External users
* of the pacemaker API should use that function instead.
*
* \note \p out should be initialized with pcmk__output_new() before calling this
* function and destroyed with out->finish and pcmk__output_free() before
* reusing it with any other functions in this library.
*
* \param[in,out] out The output functions structure.
* \param[in] target The node that was fenced.
* \param[in] as_nodeid
*
* \return Standard Pacemaker return code
*/
int pcmk__fence_last(pcmk__output_t *out, const char *target, bool as_nodeid);
/*!
* \brief List nodes that can be fenced.
*
* \note This is the internal version of pcmk_fence_list_targets(). External users
* of the pacemaker API should use that function instead.
*
* \note \p out should be initialized with pcmk__output_new() before calling this
* function and destroyed with out->finish and pcmk__output_free() before
* reusing it with any other functions in this library.
*
* \param[in,out] out The output functions structure.
* \param[in] st A connection to the STONITH API.
* \param[in] agent The agent that can do the fencing.
* \param[in] timeout How long to wait for the operation to complete (in ms).
*
* \return Standard Pacemaker return code
*/
int pcmk__fence_list_targets(pcmk__output_t *out, stonith_t *st, char *agent,
unsigned int timeout);
/*!
* \brief Get metadata for a resource.
*
* \note This is the internal version of pcmk_fence_metadata(). External users
* of the pacemaker API should use that function instead.
*
* \note \p out should be initialized with pcmk__output_new() before calling this
* function and destroyed with out->finish and pcmk__output_free() before
* reusing it with any other functions in this library.
*
* \param[in,out] out The output functions structure.
* \param[in] st A connection to the STONITH API.
* \param[in] agent The fence agent to get metadata for.
* \param[in] timeout How long to wait for the operation to complete (in ms).
*
* \return Standard Pacemaker return code
*/
int pcmk__fence_metadata(pcmk__output_t *out, stonith_t *st, char *agent,
unsigned int timeout);
/*!
* \brief List registered fence devices.
*
* \note This is the internal version of pcmk_fence_metadata(). External users
* of the pacemaker API should use that function instead.
*
* \note \p out should be initialized with pcmk__output_new() before calling this
* function and destroyed with out->finish and pcmk__output_free() before
* reusing it with any other functions in this library.
*
* \param[in,out] out The output functions structure.
* \param[in] st A connection to the STONITH API.
* \param[in] target If not NULL, only return devices that can fence
* this node.
* \param[in] timeout How long to wait for the operation to complete (in ms).
*
* \return Standard Pacemaker return code
*/
int pcmk__fence_registered(pcmk__output_t *out, stonith_t *st, char *target,
unsigned int timeout);
/*!
* \brief Register a fencing level for a specific node, node regex, or attribute.
*
* \note This is the internal version of pcmk_fence_register_level(). External users
* of the pacemaker API should use that function instead.
*
* \p target can take three different forms:
* - name=value, in which case \p target is an attribute.
* - @pattern, in which case \p target is a node regex.
* - Otherwise, \p target is a node name.
*
* \param[in] st A connection to the STONITH API.
* \param[in] target The object to register a fencing level for.
* \param[in] fence_level Index number of level to add.
* \param[in] devices Devices to use in level.
*
* \return Standard Pacemaker return code
*/
int pcmk__fence_register_level(stonith_t *st, char *target, int fence_level,
stonith_key_value_t *devices);
/*!
* \brief Unregister a fencing level for a specific node, node regex, or attribute.
*
* \note This is the internal version of pcmk_fence_unregister_level(). External users
* of the pacemaker API should use that function instead.
*
* \p target can take three different forms:
* - name=value, in which case \p target is an attribute.
* - @pattern, in which case \p target is a node regex.
* - Otherwise, \p target is a node name.
*
* \param[in] st A connection to the STONITH API.
* \param[in] target The object to unregister a fencing level for.
* \param[in] fence_level Index number of level to remove.
*
* \return Standard Pacemaker return code
*/
int pcmk__fence_unregister_level(stonith_t *st, char *target, int fence_level);
/**
* \brief Validate a STONITH device configuration.
*
* \note This is the internal version of pcmk_stonith_validate(). External users
* of the pacemaker API should use that function instead.
*
* \note \p out should be initialized with pcmk__output_new() before calling this
* function and destroyed with out->finish and pcmk__output_free() before
* reusing it with any other functions in this library.
*
* \param[in,out] out The output functions structure.
* \param[in] st A connection to the STONITH API.
* \param[in] agent The agent to validate (for example, "fence_xvm").
* \param[in] id STONITH device ID (may be NULL).
* \param[in] params STONITH device configuration parameters.
* \param[in] timeout How long to wait for the operation to complete (in ms).
*
* \return Standard Pacemaker return code
*/
int pcmk__fence_validate(pcmk__output_t *out, stonith_t *st, const char *agent,
const char *id, stonith_key_value_t *params,
unsigned int timeout);
#endif
diff --git a/lib/fencing/st_client.c b/lib/fencing/st_client.c
index 025944136c..d76b956bf4 100644
--- a/lib/fencing/st_client.c
+++ b/lib/fencing/st_client.c
@@ -1,2598 +1,2619 @@
/*
* Copyright 2004-2020 the Pacemaker project contributors
*
* The version control history for this file may have further details.
*
* This source code is licensed under the GNU Lesser General Public License
* version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY.
*/
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
CRM_TRACE_INIT_DATA(stonith);
struct stonith_action_s {
/*! user defined data */
char *agent;
char *action;
char *victim;
GHashTable *args;
int timeout;
int async;
void *userdata;
void (*done_cb) (GPid pid, gint status, const char *output, gpointer user_data);
void (*fork_cb) (GPid pid, gpointer user_data);
svc_action_t *svc_action;
/*! internal timing information */
time_t initial_start_time;
int tries;
int remaining_timeout;
int max_retries;
/* device output data */
GPid pid;
int rc;
char *output;
char *error;
};
typedef struct stonith_private_s {
char *token;
crm_ipc_t *ipc;
mainloop_io_t *source;
GHashTable *stonith_op_callback_table;
GList *notify_list;
int notify_refcnt;
bool notify_deletes;
void (*op_callback) (stonith_t * st, stonith_callback_data_t * data);
} stonith_private_t;
typedef struct stonith_notify_client_s {
const char *event;
const char *obj_id; /* implement one day */
const char *obj_type; /* implement one day */
void (*notify) (stonith_t * st, stonith_event_t * e);
bool delete;
} stonith_notify_client_t;
typedef struct stonith_callback_client_s {
void (*callback) (stonith_t * st, stonith_callback_data_t * data);
const char *id;
void *user_data;
gboolean only_success;
gboolean allow_timeout_updates;
struct timer_rec_s *timer;
} stonith_callback_client_t;
struct notify_blob_s {
stonith_t *stonith;
xmlNode *xml;
};
struct timer_rec_s {
int call_id;
int timeout;
guint ref;
stonith_t *stonith;
};
typedef int (*stonith_op_t) (const char *, int, const char *, xmlNode *,
xmlNode *, xmlNode *, xmlNode **, xmlNode **);
bool stonith_dispatch(stonith_t * st);
xmlNode *stonith_create_op(int call_id, const char *token, const char *op, xmlNode * data,
int call_options);
static int stonith_send_command(stonith_t *stonith, const char *op,
xmlNode *data, xmlNode **output_data,
int call_options, int timeout);
static void stonith_connection_destroy(gpointer user_data);
static void stonith_send_notification(gpointer data, gpointer user_data);
static int internal_stonith_action_execute(stonith_action_t * action);
static void log_action(stonith_action_t *action, pid_t pid);
/*!
* \brief Get agent namespace by name
*
* \param[in] namespace_s Name of namespace as string
*
* \return Namespace as enum value
*/
enum stonith_namespace
stonith_text2namespace(const char *namespace_s)
{
if ((namespace_s == NULL) || !strcmp(namespace_s, "any")) {
return st_namespace_any;
} else if (!strcmp(namespace_s, "redhat")
|| !strcmp(namespace_s, "stonith-ng")) {
return st_namespace_rhcs;
} else if (!strcmp(namespace_s, "internal")) {
return st_namespace_internal;
} else if (!strcmp(namespace_s, "heartbeat")) {
return st_namespace_lha;
}
return st_namespace_invalid;
}
/*!
* \brief Get agent namespace name
*
* \param[in] namespace Namespace as enum value
*
* \return Namespace name as string
*/
const char *
stonith_namespace2text(enum stonith_namespace st_namespace)
{
switch (st_namespace) {
case st_namespace_any: return "any";
case st_namespace_rhcs: return "stonith-ng";
case st_namespace_internal: return "internal";
case st_namespace_lha: return "heartbeat";
default: break;
}
return "unsupported";
}
/*!
* \brief Determine namespace of a fence agent
*
* \param[in] agent Fence agent type
* \param[in] namespace_s Name of agent namespace as string, if known
*
* \return Namespace of specified agent, as enum value
*/
enum stonith_namespace
stonith_get_namespace(const char *agent, const char *namespace_s)
{
if (safe_str_eq(namespace_s, "internal")) {
return st_namespace_internal;
}
if (stonith__agent_is_rhcs(agent)) {
return st_namespace_rhcs;
}
#if HAVE_STONITH_STONITH_H
if (stonith__agent_is_lha(agent)) {
return st_namespace_lha;
}
#endif
crm_err("Unknown fence agent: %s", agent);
return st_namespace_invalid;
}
static void
log_action(stonith_action_t *action, pid_t pid)
{
if (action->output) {
/* Logging the whole string confuses syslog when the string is xml */
char *prefix = crm_strdup_printf("%s[%d] stdout:", action->agent, pid);
crm_log_output(LOG_TRACE, prefix, action->output);
free(prefix);
}
if (action->error) {
/* Logging the whole string confuses syslog when the string is xml */
char *prefix = crm_strdup_printf("%s[%d] stderr:", action->agent, pid);
crm_log_output(LOG_WARNING, prefix, action->error);
free(prefix);
}
}
/* when cycling through the list we don't want to delete items
so just mark them and when we know nobody is using the list
loop over it to remove the marked items
*/
static void
foreach_notify_entry (stonith_private_t *private,
GFunc func,
gpointer user_data)
{
private->notify_refcnt++;
g_list_foreach(private->notify_list, func, user_data);
private->notify_refcnt--;
if ((private->notify_refcnt == 0) &&
private->notify_deletes) {
GList *list_item = private->notify_list;
private->notify_deletes = FALSE;
while (list_item != NULL)
{
stonith_notify_client_t *list_client = list_item->data;
GList *next = g_list_next(list_item);
if (list_client->delete) {
free(list_client);
private->notify_list =
g_list_delete_link(private->notify_list, list_item);
}
list_item = next;
}
}
}
static void
stonith_connection_destroy(gpointer user_data)
{
stonith_t *stonith = user_data;
stonith_private_t *native = NULL;
struct notify_blob_s blob;
crm_trace("Sending destroyed notification");
blob.stonith = stonith;
blob.xml = create_xml_node(NULL, "notify");
native = stonith->st_private;
native->ipc = NULL;
native->source = NULL;
free(native->token); native->token = NULL;
stonith->state = stonith_disconnected;
crm_xml_add(blob.xml, F_TYPE, T_STONITH_NOTIFY);
crm_xml_add(blob.xml, F_SUBTYPE, T_STONITH_NOTIFY_DISCONNECT);
foreach_notify_entry(native, stonith_send_notification, &blob);
free_xml(blob.xml);
}
xmlNode *
create_device_registration_xml(const char *id, enum stonith_namespace namespace,
const char *agent, stonith_key_value_t *params,
const char *rsc_provides)
{
xmlNode *data = create_xml_node(NULL, F_STONITH_DEVICE);
xmlNode *args = create_xml_node(data, XML_TAG_ATTRS);
#if HAVE_STONITH_STONITH_H
if (namespace == st_namespace_any) {
namespace = stonith_get_namespace(agent, NULL);
}
if (namespace == st_namespace_lha) {
hash2field((gpointer) "plugin", (gpointer) agent, args);
agent = "fence_legacy";
}
#endif
crm_xml_add(data, XML_ATTR_ID, id);
crm_xml_add(data, F_STONITH_ORIGIN, __FUNCTION__);
crm_xml_add(data, "agent", agent);
if ((namespace != st_namespace_any) && (namespace != st_namespace_invalid)) {
crm_xml_add(data, "namespace", stonith_namespace2text(namespace));
}
if (rsc_provides) {
crm_xml_add(data, "rsc_provides", rsc_provides);
}
for (; params; params = params->next) {
hash2field((gpointer) params->key, (gpointer) params->value, args);
}
return data;
}
static int
stonith_api_register_device(stonith_t * st, int call_options,
const char *id, const char *namespace, const char *agent,
stonith_key_value_t * params)
{
int rc = 0;
xmlNode *data = NULL;
data = create_device_registration_xml(id, stonith_text2namespace(namespace),
agent, params, NULL);
rc = stonith_send_command(st, STONITH_OP_DEVICE_ADD, data, NULL, call_options, 0);
free_xml(data);
return rc;
}
static int
stonith_api_remove_device(stonith_t * st, int call_options, const char *name)
{
int rc = 0;
xmlNode *data = NULL;
data = create_xml_node(NULL, F_STONITH_DEVICE);
crm_xml_add(data, F_STONITH_ORIGIN, __FUNCTION__);
crm_xml_add(data, XML_ATTR_ID, name);
rc = stonith_send_command(st, STONITH_OP_DEVICE_DEL, data, NULL, call_options, 0);
free_xml(data);
return rc;
}
static int
stonith_api_remove_level_full(stonith_t *st, int options,
const char *node, const char *pattern,
const char *attr, const char *value, int level)
{
int rc = 0;
xmlNode *data = NULL;
CRM_CHECK(node || pattern || (attr && value), return -EINVAL);
data = create_xml_node(NULL, XML_TAG_FENCING_LEVEL);
crm_xml_add(data, F_STONITH_ORIGIN, __FUNCTION__);
if (node) {
crm_xml_add(data, XML_ATTR_STONITH_TARGET, node);
} else if (pattern) {
crm_xml_add(data, XML_ATTR_STONITH_TARGET_PATTERN, pattern);
} else {
crm_xml_add(data, XML_ATTR_STONITH_TARGET_ATTRIBUTE, attr);
crm_xml_add(data, XML_ATTR_STONITH_TARGET_VALUE, value);
}
crm_xml_add_int(data, XML_ATTR_STONITH_INDEX, level);
rc = stonith_send_command(st, STONITH_OP_LEVEL_DEL, data, NULL, options, 0);
free_xml(data);
return rc;
}
static int
stonith_api_remove_level(stonith_t * st, int options, const char *node, int level)
{
return stonith_api_remove_level_full(st, options, node,
NULL, NULL, NULL, level);
}
/*!
* \internal
* \brief Create XML for fence topology level registration request
*
* \param[in] node If not NULL, target level by this node name
* \param[in] pattern If not NULL, target by node name using this regex
* \param[in] attr If not NULL, target by this node attribute
* \param[in] value If not NULL, target by this node attribute value
* \param[in] level Index number of level to register
* \param[in] device_list List of devices in level
*
* \return Newly allocated XML tree on success, NULL otherwise
*
* \note The caller should set only one of node, pattern or attr/value.
*/
xmlNode *
create_level_registration_xml(const char *node, const char *pattern,
const char *attr, const char *value,
int level, stonith_key_value_t *device_list)
{
int len = 0;
char *list = NULL;
xmlNode *data;
CRM_CHECK(node || pattern || (attr && value), return NULL);
data = create_xml_node(NULL, XML_TAG_FENCING_LEVEL);
CRM_CHECK(data, return NULL);
crm_xml_add(data, F_STONITH_ORIGIN, __FUNCTION__);
crm_xml_add_int(data, XML_ATTR_ID, level);
crm_xml_add_int(data, XML_ATTR_STONITH_INDEX, level);
if (node) {
crm_xml_add(data, XML_ATTR_STONITH_TARGET, node);
} else if (pattern) {
crm_xml_add(data, XML_ATTR_STONITH_TARGET_PATTERN, pattern);
} else {
crm_xml_add(data, XML_ATTR_STONITH_TARGET_ATTRIBUTE, attr);
crm_xml_add(data, XML_ATTR_STONITH_TARGET_VALUE, value);
}
for (; device_list; device_list = device_list->next) {
int adding = strlen(device_list->value);
if(list) {
adding++; /* +1 space */
}
crm_trace("Adding %s (%dc) at offset %d", device_list->value, adding, len);
list = realloc_safe(list, len + adding + 1); /* +1 EOS */
if (list == NULL) {
crm_perror(LOG_CRIT, "Could not create device list");
free_xml(data);
return NULL;
}
sprintf(list + len, "%s%s", len?",":"", device_list->value);
len += adding;
}
crm_xml_add(data, XML_ATTR_STONITH_DEVICES, list);
free(list);
return data;
}
static int
stonith_api_register_level_full(stonith_t * st, int options, const char *node,
const char *pattern,
const char *attr, const char *value,
int level, stonith_key_value_t *device_list)
{
int rc = 0;
xmlNode *data = create_level_registration_xml(node, pattern, attr, value,
level, device_list);
CRM_CHECK(data != NULL, return -EINVAL);
rc = stonith_send_command(st, STONITH_OP_LEVEL_ADD, data, NULL, options, 0);
free_xml(data);
return rc;
}
static int
stonith_api_register_level(stonith_t * st, int options, const char *node, int level,
stonith_key_value_t * device_list)
{
return stonith_api_register_level_full(st, options, node, NULL, NULL, NULL,
level, device_list);
}
static void
append_arg(const char *key, const char *value, GHashTable **args)
{
CRM_CHECK(key != NULL, return);
CRM_CHECK(value != NULL, return);
CRM_CHECK(args != NULL, return);
if (strstr(key, "pcmk_")) {
return;
} else if (strstr(key, CRM_META)) {
return;
} else if (safe_str_eq(key, "crm_feature_set")) {
return;
}
if (!*args) {
*args = crm_str_table_new();
}
CRM_CHECK(*args != NULL, return);
crm_trace("Appending: %s=%s", key, value);
g_hash_table_replace(*args, strdup(key), strdup(value));
}
static void
append_config_arg(gpointer key, gpointer value, gpointer user_data)
{
/* The fencer will filter action out when it registers the device,
* but ignore it here just in case any other library callers
* fail to do so.
*/
if (safe_str_neq(key, STONITH_ATTR_ACTION_OP)) {
append_arg(key, value, user_data);
return;
}
}
static GHashTable *
make_args(const char *agent, const char *action, const char *victim, uint32_t victim_nodeid, GHashTable * device_args,
GHashTable * port_map)
{
char buffer[512];
GHashTable *arg_list = NULL;
const char *value = NULL;
CRM_CHECK(action != NULL, return NULL);
snprintf(buffer, sizeof(buffer), "pcmk_%s_action", action);
if (device_args) {
value = g_hash_table_lookup(device_args, buffer);
}
if (value) {
crm_debug("Substituting action '%s' for requested operation '%s'", value, action);
action = value;
}
append_arg(STONITH_ATTR_ACTION_OP, action, &arg_list);
if (victim && device_args) {
const char *alias = victim;
const char *param = g_hash_table_lookup(device_args, STONITH_ATTR_HOSTARG);
if (port_map && g_hash_table_lookup(port_map, victim)) {
alias = g_hash_table_lookup(port_map, victim);
}
/* Always supply the node's name, too:
* https://github.com/ClusterLabs/fence-agents/blob/master/doc/FenceAgentAPI.md
*/
append_arg("nodename", victim, &arg_list);
if (victim_nodeid) {
char nodeid_str[33] = { 0, };
if (snprintf(nodeid_str, 33, "%u", (unsigned int)victim_nodeid)) {
crm_info("For stonith action (%s) for victim %s, adding nodeid (%s) to parameters",
action, victim, nodeid_str);
append_arg("nodeid", nodeid_str, &arg_list);
}
}
/* Check if we need to supply the victim in any other form */
if(safe_str_eq(agent, "fence_legacy")) {
value = agent;
} else if (param == NULL) {
param = "port";
value = g_hash_table_lookup(device_args, param);
} else if (safe_str_eq(param, "none")) {
value = param; /* Nothing more to do */
} else {
value = g_hash_table_lookup(device_args, param);
}
/* Don't overwrite explictly set values for $param */
if (value == NULL || safe_str_eq(value, "dynamic")) {
crm_debug("Performing '%s' action targeting '%s' as '%s=%s'", action, victim, param,
alias);
append_arg(param, alias, &arg_list);
}
}
if (device_args) {
g_hash_table_foreach(device_args, append_config_arg, &arg_list);
}
return arg_list;
}
/*!
* \internal
* \brief Free all memory used by a stonith action
*
* \param[in,out] action Action to free
*/
void
stonith__destroy_action(stonith_action_t *action)
{
if (action) {
free(action->agent);
if (action->args) {
g_hash_table_destroy(action->args);
}
free(action->action);
free(action->victim);
if (action->svc_action) {
services_action_free(action->svc_action);
}
free(action->output);
free(action->error);
free(action);
}
}
/*!
* \internal
* \brief Get the result of an executed stonith action
*
* \param[in,out] action Executed action
* \param[out] rc Where to store result code (or NULL)
* \param[out] output Where to store standard output (or NULL)
* \param[out] error_output Where to store standard error output (or NULL)
*
* \note If output or error_output is not NULL, the caller is responsible for
* freeing the memory.
*/
void
stonith__action_result(stonith_action_t *action, int *rc, char **output,
char **error_output)
{
if (rc) {
*rc = pcmk_ok;
}
if (output) {
*output = NULL;
}
if (error_output) {
*error_output = NULL;
}
if (action != NULL) {
if (rc) {
*rc = action->rc;
}
if (output && action->output) {
*output = action->output;
action->output = NULL; // hand off memory management to caller
}
if (error_output && action->error) {
*error_output = action->error;
action->error = NULL; // hand off memory management to caller
}
}
}
#define FAILURE_MAX_RETRIES 2
stonith_action_t *
stonith_action_create(const char *agent,
const char *_action,
const char *victim,
uint32_t victim_nodeid,
int timeout, GHashTable * device_args, GHashTable * port_map)
{
stonith_action_t *action;
action = calloc(1, sizeof(stonith_action_t));
action->args = make_args(agent, _action, victim, victim_nodeid, device_args, port_map);
crm_debug("Preparing '%s' action for %s using agent %s",
_action, (victim? victim : "no target"), agent);
action->agent = strdup(agent);
action->action = strdup(_action);
if (victim) {
action->victim = strdup(victim);
}
action->timeout = action->remaining_timeout = timeout;
action->max_retries = FAILURE_MAX_RETRIES;
if (device_args) {
char buffer[512];
const char *value = NULL;
snprintf(buffer, sizeof(buffer), "pcmk_%s_retries", _action);
value = g_hash_table_lookup(device_args, buffer);
if (value) {
action->max_retries = atoi(value);
}
}
return action;
}
static gboolean
update_remaining_timeout(stonith_action_t * action)
{
int diff = time(NULL) - action->initial_start_time;
if (action->tries >= action->max_retries) {
crm_info("Attempted to execute agent %s (%s) the maximum number of times (%d) allowed",
action->agent, action->action, action->max_retries);
action->remaining_timeout = 0;
} else if ((action->rc != -ETIME) && diff < (action->timeout * 0.7)) {
/* only set remaining timeout period if there is 30%
* or greater of the original timeout period left */
action->remaining_timeout = action->timeout - diff;
} else {
action->remaining_timeout = 0;
}
return action->remaining_timeout ? TRUE : FALSE;
}
static int
svc_action_to_errno(svc_action_t *svc_action) {
int rv = pcmk_ok;
if (svc_action->rc > 0) {
/* Try to provide a useful error code based on the fence agent's
* error output.
*/
if (svc_action->rc == PCMK_OCF_TIMEOUT) {
rv = -ETIME;
} else if (svc_action->stderr_data == NULL) {
rv = -ENODATA;
} else if (strstr(svc_action->stderr_data, "imed out")) {
/* Some agents have their own internal timeouts */
rv = -ETIME;
} else if (strstr(svc_action->stderr_data, "Unrecognised action")) {
rv = -EOPNOTSUPP;
} else {
rv = -pcmk_err_generic;
}
}
return rv;
}
static void
stonith_action_async_done(svc_action_t *svc_action)
{
stonith_action_t *action = (stonith_action_t *) svc_action->cb_data;
action->rc = svc_action_to_errno(svc_action);
action->output = svc_action->stdout_data;
svc_action->stdout_data = NULL;
action->error = svc_action->stderr_data;
svc_action->stderr_data = NULL;
svc_action->params = NULL;
crm_debug("Child process %d performing action '%s' exited with rc %d",
action->pid, action->action, svc_action->rc);
log_action(action, action->pid);
if (action->rc != pcmk_ok && update_remaining_timeout(action)) {
int rc = internal_stonith_action_execute(action);
if (rc == pcmk_ok) {
return;
}
}
if (action->done_cb) {
action->done_cb(action->pid, action->rc, action->output, action->userdata);
}
action->svc_action = NULL; // don't remove our caller
stonith__destroy_action(action);
}
static void
stonith_action_async_forked(svc_action_t *svc_action)
{
stonith_action_t *action = (stonith_action_t *) svc_action->cb_data;
action->pid = svc_action->pid;
action->svc_action = svc_action;
if (action->fork_cb) {
(action->fork_cb) (svc_action->pid, action->userdata);
}
crm_trace("Child process %d performing action '%s' successfully forked",
action->pid, action->action);
}
static int
internal_stonith_action_execute(stonith_action_t * action)
{
int rc = -EPROTO;
int is_retry = 0;
svc_action_t *svc_action = NULL;
static int stonith_sequence = 0;
char *buffer = NULL;
if (!action->tries) {
action->initial_start_time = time(NULL);
}
action->tries++;
if (action->tries > 1) {
crm_info("Attempt %d to execute %s (%s). remaining timeout is %d",
action->tries, action->agent, action->action, action->remaining_timeout);
is_retry = 1;
}
if (action->args == NULL || action->agent == NULL)
goto fail;
buffer = crm_strdup_printf(RH_STONITH_DIR "/%s", basename(action->agent));
svc_action = services_action_create_generic(buffer, NULL);
free(buffer);
svc_action->timeout = 1000 * action->remaining_timeout;
svc_action->standard = strdup(PCMK_RESOURCE_CLASS_STONITH);
svc_action->id = crm_strdup_printf("%s_%s_%d", basename(action->agent),
action->action, action->tries);
svc_action->agent = strdup(action->agent);
svc_action->sequence = stonith_sequence++;
svc_action->params = action->args;
svc_action->cb_data = (void *) action;
set_bit(svc_action->flags, SVC_ACTION_NON_BLOCKED);
/* keep retries from executing out of control and free previous results */
if (is_retry) {
free(action->output);
action->output = NULL;
free(action->error);
action->error = NULL;
sleep(1);
}
if (action->async) {
/* async */
if(services_action_async_fork_notify(svc_action,
&stonith_action_async_done,
&stonith_action_async_forked) == FALSE) {
services_action_free(svc_action);
svc_action = NULL;
} else {
rc = 0;
}
} else {
/* sync */
if (services_action_sync(svc_action)) {
rc = 0;
action->rc = svc_action_to_errno(svc_action);
action->output = svc_action->stdout_data;
svc_action->stdout_data = NULL;
action->error = svc_action->stderr_data;
svc_action->stderr_data = NULL;
} else {
action->rc = -ECONNABORTED;
rc = action->rc;
}
svc_action->params = NULL;
services_action_free(svc_action);
}
fail:
return rc;
}
/*!
* \internal
* \brief Kick off execution of an async stonith action
*
* \param[in,out] action Action to be executed
* \param[in,out] userdata Datapointer to be passed to callbacks
* \param[in] done Callback to notify action has failed/succeeded
* \param[in] fork_callback Callback to notify successful fork of child
*
* \return pcmk_ok if ownership of action has been taken, -errno otherwise
*/
int
stonith_action_execute_async(stonith_action_t * action,
void *userdata,
void (*done) (GPid pid, int rc, const char *output,
gpointer user_data),
void (*fork_cb) (GPid pid, gpointer user_data))
{
if (!action) {
return -EINVAL;
}
action->userdata = userdata;
action->done_cb = done;
action->fork_cb = fork_cb;
action->async = 1;
return internal_stonith_action_execute(action);
}
/*!
* \internal
* \brief Execute a stonith action
*
* \param[in,out] action Action to execute
*
* \return pcmk_ok on success, -errno otherwise
*/
int
stonith__execute(stonith_action_t *action)
{
int rc = pcmk_ok;
CRM_CHECK(action != NULL, return -EINVAL);
// Keep trying until success, max retries, or timeout
do {
rc = internal_stonith_action_execute(action);
} while ((rc != pcmk_ok) && update_remaining_timeout(action));
return rc;
}
static int
stonith_api_device_list(stonith_t * stonith, int call_options, const char *namespace,
stonith_key_value_t ** devices, int timeout)
{
int count = 0;
enum stonith_namespace ns = stonith_text2namespace(namespace);
if (devices == NULL) {
crm_err("Parameter error: stonith_api_device_list");
return -EFAULT;
}
#if HAVE_STONITH_STONITH_H
// Include Linux-HA agents if requested
if ((ns == st_namespace_any) || (ns == st_namespace_lha)) {
count += stonith__list_lha_agents(devices);
}
#endif
// Include Red Hat agents if requested
if ((ns == st_namespace_any) || (ns == st_namespace_rhcs)) {
count += stonith__list_rhcs_agents(devices);
}
return count;
}
static int
stonith_api_device_metadata(stonith_t * stonith, int call_options, const char *agent,
const char *namespace, char **output, int timeout)
{
/* By executing meta-data directly, we can get it from stonith_admin when
* the cluster is not running, which is important for higher-level tools.
*/
enum stonith_namespace ns = stonith_get_namespace(agent, namespace);
crm_trace("Looking up metadata for %s agent %s",
stonith_namespace2text(ns), agent);
switch (ns) {
case st_namespace_rhcs:
return stonith__rhcs_metadata(agent, timeout, output);
#if HAVE_STONITH_STONITH_H
case st_namespace_lha:
return stonith__lha_metadata(agent, timeout, output);
#endif
default:
crm_err("Can't get fence agent '%s' meta-data: No such agent",
agent);
break;
}
return -ENODEV;
}
static int
stonith_api_query(stonith_t * stonith, int call_options, const char *target,
stonith_key_value_t ** devices, int timeout)
{
int rc = 0, lpc = 0, max = 0;
xmlNode *data = NULL;
xmlNode *output = NULL;
xmlXPathObjectPtr xpathObj = NULL;
CRM_CHECK(devices != NULL, return -EINVAL);
data = create_xml_node(NULL, F_STONITH_DEVICE);
crm_xml_add(data, F_STONITH_ORIGIN, __FUNCTION__);
crm_xml_add(data, F_STONITH_TARGET, target);
crm_xml_add(data, F_STONITH_ACTION, "off");
rc = stonith_send_command(stonith, STONITH_OP_QUERY, data, &output, call_options, timeout);
if (rc < 0) {
return rc;
}
xpathObj = xpath_search(output, "//@agent");
if (xpathObj) {
max = numXpathResults(xpathObj);
for (lpc = 0; lpc < max; lpc++) {
xmlNode *match = getXpathResult(xpathObj, lpc);
CRM_LOG_ASSERT(match != NULL);
if(match != NULL) {
xmlChar *match_path = xmlGetNodePath(match);
crm_info("%s[%d] = %s", "//@agent", lpc, match_path);
free(match_path);
*devices = stonith_key_value_add(*devices, NULL, crm_element_value(match, XML_ATTR_ID));
}
}
freeXpathObject(xpathObj);
}
free_xml(output);
free_xml(data);
return max;
}
static int
stonith_api_call(stonith_t * stonith,
int call_options,
const char *id,
const char *action, const char *victim, int timeout, xmlNode ** output)
{
int rc = 0;
xmlNode *data = NULL;
data = create_xml_node(NULL, F_STONITH_DEVICE);
crm_xml_add(data, F_STONITH_ORIGIN, __FUNCTION__);
crm_xml_add(data, F_STONITH_DEVICE, id);
crm_xml_add(data, F_STONITH_ACTION, action);
crm_xml_add(data, F_STONITH_TARGET, victim);
rc = stonith_send_command(stonith, STONITH_OP_EXEC, data, output, call_options, timeout);
free_xml(data);
return rc;
}
static int
stonith_api_list(stonith_t * stonith, int call_options, const char *id, char **list_info,
int timeout)
{
int rc;
xmlNode *output = NULL;
rc = stonith_api_call(stonith, call_options, id, "list", NULL, timeout, &output);
if (output && list_info) {
const char *list_str;
list_str = crm_element_value(output, "st_output");
if (list_str) {
*list_info = strdup(list_str);
}
}
if (output) {
free_xml(output);
}
return rc;
}
static int
stonith_api_monitor(stonith_t * stonith, int call_options, const char *id, int timeout)
{
return stonith_api_call(stonith, call_options, id, "monitor", NULL, timeout, NULL);
}
static int
stonith_api_status(stonith_t * stonith, int call_options, const char *id, const char *port,
int timeout)
{
return stonith_api_call(stonith, call_options, id, "status", port, timeout, NULL);
}
static int
-stonith_api_fence(stonith_t * stonith, int call_options, const char *node, const char *action,
- int timeout, int tolerance)
+stonith_api_fence_with_delay(stonith_t * stonith, int call_options, const char *node,
+ const char *action, int timeout, int tolerance, int delay)
{
int rc = 0;
xmlNode *data = NULL;
data = create_xml_node(NULL, __FUNCTION__);
crm_xml_add(data, F_STONITH_TARGET, node);
crm_xml_add(data, F_STONITH_ACTION, action);
crm_xml_add_int(data, F_STONITH_TIMEOUT, timeout);
crm_xml_add_int(data, F_STONITH_TOLERANCE, tolerance);
+ if (delay >= 0) {
+ crm_xml_add_int(data, F_STONITH_DELAY, delay);
+ }
+
rc = stonith_send_command(stonith, STONITH_OP_FENCE, data, NULL, call_options, timeout);
free_xml(data);
return rc;
}
+static int
+stonith_api_fence(stonith_t * stonith, int call_options, const char *node, const char *action,
+ int timeout, int tolerance)
+{
+ return stonith_api_fence_with_delay(stonith, call_options, node, action,
+ timeout, tolerance, -1);
+}
+
static int
stonith_api_confirm(stonith_t * stonith, int call_options, const char *target)
{
return stonith_api_fence(stonith, call_options | st_opt_manual_ack, target, "off", 0, 0);
}
static int
stonith_api_history(stonith_t * stonith, int call_options, const char *node,
stonith_history_t ** history, int timeout)
{
int rc = 0;
xmlNode *data = NULL;
xmlNode *output = NULL;
stonith_history_t *last = NULL;
*history = NULL;
if (node) {
data = create_xml_node(NULL, __FUNCTION__);
crm_xml_add(data, F_STONITH_TARGET, node);
}
rc = stonith_send_command(stonith, STONITH_OP_FENCE_HISTORY, data, &output,
call_options | st_opt_sync_call, timeout);
free_xml(data);
if (rc == 0) {
xmlNode *op = NULL;
xmlNode *reply = get_xpath_object("//" F_STONITH_HISTORY_LIST, output,
LOG_NEVER);
for (op = __xml_first_child(reply); op != NULL; op = __xml_next(op)) {
stonith_history_t *kvp;
int completed;
kvp = calloc(1, sizeof(stonith_history_t));
kvp->target = crm_element_value_copy(op, F_STONITH_TARGET);
kvp->action = crm_element_value_copy(op, F_STONITH_ACTION);
kvp->origin = crm_element_value_copy(op, F_STONITH_ORIGIN);
kvp->delegate = crm_element_value_copy(op, F_STONITH_DELEGATE);
kvp->client = crm_element_value_copy(op, F_STONITH_CLIENTNAME);
crm_element_value_int(op, F_STONITH_DATE, &completed);
kvp->completed = (time_t) completed;
crm_element_value_int(op, F_STONITH_STATE, &kvp->state);
if (last) {
last->next = kvp;
} else {
*history = kvp;
}
last = kvp;
}
}
free_xml(output);
return rc;
}
void stonith_history_free(stonith_history_t *history)
{
stonith_history_t *hp, *hp_old;
for (hp = history; hp; hp_old = hp, hp = hp->next, free(hp_old)) {
free(hp->target);
free(hp->action);
free(hp->origin);
free(hp->delegate);
free(hp->client);
}
}
static gint
stonithlib_GCompareFunc(gconstpointer a, gconstpointer b)
{
int rc = 0;
const stonith_notify_client_t *a_client = a;
const stonith_notify_client_t *b_client = b;
if (a_client->delete || b_client->delete) {
/* make entries marked for deletion not findable */
return -1;
}
CRM_CHECK(a_client->event != NULL && b_client->event != NULL, return 0);
rc = strcmp(a_client->event, b_client->event);
if (rc == 0) {
if (a_client->notify == NULL || b_client->notify == NULL) {
return 0;
} else if (a_client->notify == b_client->notify) {
return 0;
} else if (((long)a_client->notify) < ((long)b_client->notify)) {
crm_err("callbacks for %s are not equal: %p vs. %p",
a_client->event, a_client->notify, b_client->notify);
return -1;
}
crm_err("callbacks for %s are not equal: %p vs. %p",
a_client->event, a_client->notify, b_client->notify);
return 1;
}
return rc;
}
xmlNode *
stonith_create_op(int call_id, const char *token, const char *op, xmlNode * data, int call_options)
{
xmlNode *op_msg = create_xml_node(NULL, "stonith_command");
CRM_CHECK(op_msg != NULL, return NULL);
CRM_CHECK(token != NULL, return NULL);
crm_xml_add(op_msg, F_XML_TAGNAME, "stonith_command");
crm_xml_add(op_msg, F_TYPE, T_STONITH_NG);
crm_xml_add(op_msg, F_STONITH_CALLBACK_TOKEN, token);
crm_xml_add(op_msg, F_STONITH_OPERATION, op);
crm_xml_add_int(op_msg, F_STONITH_CALLID, call_id);
crm_trace("Sending call options: %.8lx, %d", (long)call_options, call_options);
crm_xml_add_int(op_msg, F_STONITH_CALLOPTS, call_options);
if (data != NULL) {
add_message_xml(op_msg, F_STONITH_CALLDATA, data);
}
return op_msg;
}
static void
stonith_destroy_op_callback(gpointer data)
{
stonith_callback_client_t *blob = data;
if (blob->timer && blob->timer->ref > 0) {
g_source_remove(blob->timer->ref);
}
free(blob->timer);
free(blob);
}
static int
stonith_api_signoff(stonith_t * stonith)
{
stonith_private_t *native = stonith->st_private;
crm_debug("Disconnecting from the fencer");
if (native->source != NULL) {
/* Attached to mainloop */
mainloop_del_ipc_client(native->source);
native->source = NULL;
native->ipc = NULL;
} else if (native->ipc) {
/* Not attached to mainloop */
crm_ipc_t *ipc = native->ipc;
native->ipc = NULL;
crm_ipc_close(ipc);
crm_ipc_destroy(ipc);
}
free(native->token); native->token = NULL;
stonith->state = stonith_disconnected;
return pcmk_ok;
}
static int
stonith_api_del_callback(stonith_t * stonith, int call_id, bool all_callbacks)
{
stonith_private_t *private = stonith->st_private;
if (all_callbacks) {
private->op_callback = NULL;
g_hash_table_destroy(private->stonith_op_callback_table);
private->stonith_op_callback_table = g_hash_table_new_full(g_direct_hash, g_direct_equal,
NULL,
stonith_destroy_op_callback);
} else if (call_id == 0) {
private->op_callback = NULL;
} else {
g_hash_table_remove(private->stonith_op_callback_table, GINT_TO_POINTER(call_id));
}
return pcmk_ok;
}
static void
invoke_callback(stonith_t * st, int call_id, int rc, void *userdata,
void (*callback) (stonith_t * st, stonith_callback_data_t * data))
{
stonith_callback_data_t data = { 0, };
data.call_id = call_id;
data.rc = rc;
data.userdata = userdata;
callback(st, &data);
}
static void
stonith_perform_callback(stonith_t * stonith, xmlNode * msg, int call_id, int rc)
{
stonith_private_t *private = NULL;
stonith_callback_client_t *blob = NULL;
stonith_callback_client_t local_blob;
CRM_CHECK(stonith != NULL, return);
CRM_CHECK(stonith->st_private != NULL, return);
private = stonith->st_private;
local_blob.id = NULL;
local_blob.callback = NULL;
local_blob.user_data = NULL;
local_blob.only_success = FALSE;
if (msg != NULL) {
crm_element_value_int(msg, F_STONITH_RC, &rc);
crm_element_value_int(msg, F_STONITH_CALLID, &call_id);
}
CRM_CHECK(call_id > 0, crm_log_xml_err(msg, "Bad result"));
blob = g_hash_table_lookup(private->stonith_op_callback_table, GINT_TO_POINTER(call_id));
if (blob != NULL) {
local_blob = *blob;
blob = NULL;
stonith_api_del_callback(stonith, call_id, FALSE);
} else {
crm_trace("No callback found for call %d", call_id);
local_blob.callback = NULL;
}
if (local_blob.callback != NULL && (rc == pcmk_ok || local_blob.only_success == FALSE)) {
crm_trace("Invoking callback %s for call %d", crm_str(local_blob.id), call_id);
invoke_callback(stonith, call_id, rc, local_blob.user_data, local_blob.callback);
} else if (private->op_callback == NULL && rc != pcmk_ok) {
crm_warn("Fencing command failed: %s", pcmk_strerror(rc));
crm_log_xml_debug(msg, "Failed fence update");
}
if (private->op_callback != NULL) {
crm_trace("Invoking global callback for call %d", call_id);
invoke_callback(stonith, call_id, rc, NULL, private->op_callback);
}
crm_trace("OP callback activated.");
}
static gboolean
stonith_async_timeout_handler(gpointer data)
{
struct timer_rec_s *timer = data;
crm_err("Async call %d timed out after %dms", timer->call_id, timer->timeout);
stonith_perform_callback(timer->stonith, NULL, timer->call_id, -ETIME);
/* Always return TRUE, never remove the handler
* We do that in stonith_del_callback()
*/
return TRUE;
}
static void
set_callback_timeout(stonith_callback_client_t * callback, stonith_t * stonith, int call_id,
int timeout)
{
struct timer_rec_s *async_timer = callback->timer;
if (timeout <= 0) {
return;
}
if (!async_timer) {
async_timer = calloc(1, sizeof(struct timer_rec_s));
callback->timer = async_timer;
}
async_timer->stonith = stonith;
async_timer->call_id = call_id;
/* Allow a fair bit of grace to allow the server to tell us of a timeout
* This is only a fallback
*/
async_timer->timeout = (timeout + 60) * 1000;
if (async_timer->ref) {
g_source_remove(async_timer->ref);
}
async_timer->ref =
g_timeout_add(async_timer->timeout, stonith_async_timeout_handler, async_timer);
}
static void
update_callback_timeout(int call_id, int timeout, stonith_t * st)
{
stonith_callback_client_t *callback = NULL;
stonith_private_t *private = st->st_private;
callback = g_hash_table_lookup(private->stonith_op_callback_table, GINT_TO_POINTER(call_id));
if (!callback || !callback->allow_timeout_updates) {
return;
}
set_callback_timeout(callback, st, call_id, timeout);
}
static int
stonith_dispatch_internal(const char *buffer, ssize_t length, gpointer userdata)
{
const char *type = NULL;
struct notify_blob_s blob;
stonith_t *st = userdata;
stonith_private_t *private = NULL;
CRM_ASSERT(st != NULL);
private = st->st_private;
blob.stonith = st;
blob.xml = string2xml(buffer);
if (blob.xml == NULL) {
crm_warn("Received malformed message from fencer: %s", buffer);
return 0;
}
/* do callbacks */
type = crm_element_value(blob.xml, F_TYPE);
crm_trace("Activating %s callbacks...", type);
if (safe_str_eq(type, T_STONITH_NG)) {
stonith_perform_callback(st, blob.xml, 0, 0);
} else if (safe_str_eq(type, T_STONITH_NOTIFY)) {
foreach_notify_entry(private, stonith_send_notification, &blob);
} else if (safe_str_eq(type, T_STONITH_TIMEOUT_VALUE)) {
int call_id = 0;
int timeout = 0;
crm_element_value_int(blob.xml, F_STONITH_TIMEOUT, &timeout);
crm_element_value_int(blob.xml, F_STONITH_CALLID, &call_id);
update_callback_timeout(call_id, timeout, st);
} else {
crm_err("Unknown message type: %s", type);
crm_log_xml_warn(blob.xml, "BadReply");
}
free_xml(blob.xml);
return 1;
}
static int
stonith_api_signon(stonith_t * stonith, const char *name, int *stonith_fd)
{
int rc = pcmk_ok;
stonith_private_t *native = NULL;
const char *display_name = name? name : "client";
struct ipc_client_callbacks st_callbacks = {
.dispatch = stonith_dispatch_internal,
.destroy = stonith_connection_destroy
};
CRM_CHECK(stonith != NULL, return -EINVAL);
native = stonith->st_private;
CRM_ASSERT(native != NULL);
crm_debug("Attempting fencer connection by %s with%s mainloop",
display_name, (stonith_fd? "out" : ""));
stonith->state = stonith_connected_command;
if (stonith_fd) {
/* No mainloop */
native->ipc = crm_ipc_new("stonith-ng", 0);
if (native->ipc && crm_ipc_connect(native->ipc)) {
*stonith_fd = crm_ipc_get_fd(native->ipc);
} else if (native->ipc) {
crm_ipc_close(native->ipc);
crm_ipc_destroy(native->ipc);
native->ipc = NULL;
}
} else {
/* With mainloop */
native->source =
mainloop_add_ipc_client("stonith-ng", G_PRIORITY_MEDIUM, 0, stonith, &st_callbacks);
native->ipc = mainloop_get_ipc_client(native->source);
}
if (native->ipc == NULL) {
rc = -ENOTCONN;
} else {
xmlNode *reply = NULL;
xmlNode *hello = create_xml_node(NULL, "stonith_command");
crm_xml_add(hello, F_TYPE, T_STONITH_NG);
crm_xml_add(hello, F_STONITH_OPERATION, CRM_OP_REGISTER);
crm_xml_add(hello, F_STONITH_CLIENTNAME, name);
rc = crm_ipc_send(native->ipc, hello, crm_ipc_client_response, -1, &reply);
if (rc < 0) {
crm_debug("Couldn't register with the fencer: %s "
CRM_XS " rc=%d", pcmk_strerror(rc), rc);
rc = -ECOMM;
} else if (reply == NULL) {
crm_debug("Couldn't register with the fencer: no reply");
rc = -EPROTO;
} else {
const char *msg_type = crm_element_value(reply, F_STONITH_OPERATION);
native->token = crm_element_value_copy(reply, F_STONITH_CLIENTID);
if (safe_str_neq(msg_type, CRM_OP_REGISTER)) {
crm_debug("Couldn't register with the fencer: invalid reply type '%s'",
(msg_type? msg_type : "(missing)"));
crm_log_xml_debug(reply, "Invalid fencer reply");
rc = -EPROTO;
} else if (native->token == NULL) {
crm_debug("Couldn't register with the fencer: no token in reply");
crm_log_xml_debug(reply, "Invalid fencer reply");
rc = -EPROTO;
} else {
#if HAVE_MSGFROMIPC_TIMEOUT
stonith->call_timeout = MAX_IPC_DELAY;
#endif
crm_debug("Connection to fencer by %s succeeded (registration token: %s)",
display_name, native->token);
rc = pcmk_ok;
}
}
free_xml(reply);
free_xml(hello);
}
if (rc != pcmk_ok) {
crm_debug("Connection attempt to fencer by %s failed: %s "
CRM_XS " rc=%d", display_name, pcmk_strerror(rc), rc);
stonith->cmds->disconnect(stonith);
}
return rc;
}
static int
stonith_set_notification(stonith_t * stonith, const char *callback, int enabled)
{
int rc = pcmk_ok;
xmlNode *notify_msg = create_xml_node(NULL, __FUNCTION__);
stonith_private_t *native = stonith->st_private;
if (stonith->state != stonith_disconnected) {
crm_xml_add(notify_msg, F_STONITH_OPERATION, T_STONITH_NOTIFY);
if (enabled) {
crm_xml_add(notify_msg, F_STONITH_NOTIFY_ACTIVATE, callback);
} else {
crm_xml_add(notify_msg, F_STONITH_NOTIFY_DEACTIVATE, callback);
}
rc = crm_ipc_send(native->ipc, notify_msg, crm_ipc_client_response, -1, NULL);
if (rc < 0) {
crm_perror(LOG_DEBUG, "Couldn't register for fencing notifications: %d", rc);
rc = -ECOMM;
} else {
rc = pcmk_ok;
}
}
free_xml(notify_msg);
return rc;
}
static int
stonith_api_add_notification(stonith_t * stonith, const char *event,
void (*callback) (stonith_t * stonith, stonith_event_t * e))
{
GList *list_item = NULL;
stonith_notify_client_t *new_client = NULL;
stonith_private_t *private = NULL;
private = stonith->st_private;
crm_trace("Adding callback for %s events (%d)", event, g_list_length(private->notify_list));
new_client = calloc(1, sizeof(stonith_notify_client_t));
new_client->event = event;
new_client->notify = callback;
list_item = g_list_find_custom(private->notify_list, new_client, stonithlib_GCompareFunc);
if (list_item != NULL) {
crm_warn("Callback already present");
free(new_client);
return -ENOTUNIQ;
} else {
private->notify_list = g_list_append(private->notify_list, new_client);
stonith_set_notification(stonith, event, 1);
crm_trace("Callback added (%d)", g_list_length(private->notify_list));
}
return pcmk_ok;
}
static int
stonith_api_del_notification(stonith_t * stonith, const char *event)
{
GList *list_item = NULL;
stonith_notify_client_t *new_client = NULL;
stonith_private_t *private = NULL;
crm_debug("Removing callback for %s events", event);
private = stonith->st_private;
new_client = calloc(1, sizeof(stonith_notify_client_t));
new_client->event = event;
new_client->notify = NULL;
list_item = g_list_find_custom(private->notify_list, new_client, stonithlib_GCompareFunc);
stonith_set_notification(stonith, event, 0);
if (list_item != NULL) {
stonith_notify_client_t *list_client = list_item->data;
if (private->notify_refcnt) {
list_client->delete = TRUE;
private->notify_deletes = TRUE;
} else {
private->notify_list = g_list_remove(private->notify_list, list_client);
free(list_client);
}
crm_trace("Removed callback");
} else {
crm_trace("Callback not present");
}
free(new_client);
return pcmk_ok;
}
static int
stonith_api_add_callback(stonith_t * stonith, int call_id, int timeout, int options,
void *user_data, const char *callback_name,
void (*callback) (stonith_t * st, stonith_callback_data_t * data))
{
stonith_callback_client_t *blob = NULL;
stonith_private_t *private = NULL;
CRM_CHECK(stonith != NULL, return -EINVAL);
CRM_CHECK(stonith->st_private != NULL, return -EINVAL);
private = stonith->st_private;
if (call_id == 0) {
private->op_callback = callback;
} else if (call_id < 0) {
if (!(options & st_opt_report_only_success)) {
crm_trace("Call failed, calling %s: %s", callback_name, pcmk_strerror(call_id));
invoke_callback(stonith, call_id, call_id, user_data, callback);
} else {
crm_warn("Fencer call failed: %s", pcmk_strerror(call_id));
}
return FALSE;
}
blob = calloc(1, sizeof(stonith_callback_client_t));
blob->id = callback_name;
blob->only_success = (options & st_opt_report_only_success) ? TRUE : FALSE;
blob->user_data = user_data;
blob->callback = callback;
blob->allow_timeout_updates = (options & st_opt_timeout_updates) ? TRUE : FALSE;
if (timeout > 0) {
set_callback_timeout(blob, stonith, call_id, timeout);
}
g_hash_table_insert(private->stonith_op_callback_table, GINT_TO_POINTER(call_id), blob);
crm_trace("Added callback to %s for call %d", callback_name, call_id);
return TRUE;
}
static void
stonith_dump_pending_op(gpointer key, gpointer value, gpointer user_data)
{
int call = GPOINTER_TO_INT(key);
stonith_callback_client_t *blob = value;
crm_debug("Call %d (%s): pending", call, crm_str(blob->id));
}
void
stonith_dump_pending_callbacks(stonith_t * stonith)
{
stonith_private_t *private = stonith->st_private;
if (private->stonith_op_callback_table == NULL) {
return;
}
return g_hash_table_foreach(private->stonith_op_callback_table, stonith_dump_pending_op, NULL);
}
/*
*/
static stonith_event_t *
xml_to_event(xmlNode * msg)
{
stonith_event_t *event = calloc(1, sizeof(stonith_event_t));
const char *ntype = crm_element_value(msg, F_SUBTYPE);
char *data_addr = crm_strdup_printf("//%s", ntype);
xmlNode *data = get_xpath_object(data_addr, msg, LOG_DEBUG);
crm_log_xml_trace(msg, "stonith_notify");
crm_element_value_int(msg, F_STONITH_RC, &(event->result));
if (safe_str_eq(ntype, T_STONITH_NOTIFY_FENCE)) {
event->operation = crm_element_value_copy(msg, F_STONITH_OPERATION);
if (data) {
event->origin = crm_element_value_copy(data, F_STONITH_ORIGIN);
event->action = crm_element_value_copy(data, F_STONITH_ACTION);
event->target = crm_element_value_copy(data, F_STONITH_TARGET);
event->executioner = crm_element_value_copy(data, F_STONITH_DELEGATE);
event->id = crm_element_value_copy(data, F_STONITH_REMOTE_OP_ID);
event->client_origin = crm_element_value_copy(data, F_STONITH_CLIENTNAME);
event->device = crm_element_value_copy(data, F_STONITH_DEVICE);
} else {
crm_err("No data for %s event", ntype);
crm_log_xml_notice(msg, "BadEvent");
}
}
free(data_addr);
return event;
}
static void
event_free(stonith_event_t * event)
{
free(event->id);
free(event->type);
free(event->message);
free(event->operation);
free(event->origin);
free(event->action);
free(event->target);
free(event->executioner);
free(event->device);
free(event->client_origin);
free(event);
}
static void
stonith_send_notification(gpointer data, gpointer user_data)
{
struct notify_blob_s *blob = user_data;
stonith_notify_client_t *entry = data;
stonith_event_t *st_event = NULL;
const char *event = NULL;
if (blob->xml == NULL) {
crm_warn("Skipping callback - NULL message");
return;
}
event = crm_element_value(blob->xml, F_SUBTYPE);
if (entry == NULL) {
crm_warn("Skipping callback - NULL callback client");
return;
} else if (entry->delete) {
crm_trace("Skipping callback - marked for deletion");
return;
} else if (entry->notify == NULL) {
crm_warn("Skipping callback - NULL callback");
return;
} else if (safe_str_neq(entry->event, event)) {
crm_trace("Skipping callback - event mismatch %p/%s vs. %s", entry, entry->event, event);
return;
}
st_event = xml_to_event(blob->xml);
crm_trace("Invoking callback for %p/%s event...", entry, event);
entry->notify(blob->stonith, st_event);
crm_trace("Callback invoked...");
event_free(st_event);
}
/*!
* \internal
* \brief Create and send an API request
*
* \param[in] stonith Stonith connection
* \param[in] op API operation to request
* \param[in] data Data to attach to request
* \param[out] output_data If not NULL, will be set to reply if synchronous
* \param[in] call_options Bitmask of stonith_call_options to use
* \param[in] timeout Error if not completed within this many seconds
*
* \return pcmk_ok (for synchronous requests) or positive call ID
* (for asynchronous requests) on success, -errno otherwise
*/
static int
stonith_send_command(stonith_t * stonith, const char *op, xmlNode * data, xmlNode ** output_data,
int call_options, int timeout)
{
int rc = 0;
int reply_id = -1;
xmlNode *op_msg = NULL;
xmlNode *op_reply = NULL;
stonith_private_t *native = NULL;
CRM_ASSERT(stonith && stonith->st_private && op);
native = stonith->st_private;
if (output_data != NULL) {
*output_data = NULL;
}
if ((stonith->state == stonith_disconnected) || (native->token == NULL)) {
return -ENOTCONN;
}
/* Increment the call ID, which must be positive to avoid conflicting with
* error codes. This shouldn't be a problem unless the client mucked with
* it or the counter wrapped around.
*/
stonith->call_id++;
if (stonith->call_id < 1) {
stonith->call_id = 1;
}
op_msg = stonith_create_op(stonith->call_id, native->token, op, data, call_options);
if (op_msg == NULL) {
return -EINVAL;
}
crm_xml_add_int(op_msg, F_STONITH_TIMEOUT, timeout);
crm_trace("Sending %s message to fencer with timeout %ds", op, timeout);
+ if (data) {
+ const char *delay_s = crm_element_value(data, F_STONITH_DELAY);
+
+ if (delay_s) {
+ crm_xml_add(op_msg, F_STONITH_DELAY, delay_s);
+ }
+ }
+
{
enum crm_ipc_flags ipc_flags = crm_ipc_flags_none;
if (call_options & st_opt_sync_call) {
ipc_flags |= crm_ipc_client_response;
}
rc = crm_ipc_send(native->ipc, op_msg, ipc_flags,
1000 * (timeout + 60), &op_reply);
}
free_xml(op_msg);
if (rc < 0) {
crm_perror(LOG_ERR, "Couldn't perform %s operation (timeout=%ds): %d", op, timeout, rc);
rc = -ECOMM;
goto done;
}
crm_log_xml_trace(op_reply, "Reply");
if (!(call_options & st_opt_sync_call)) {
crm_trace("Async call %d, returning", stonith->call_id);
free_xml(op_reply);
return stonith->call_id;
}
rc = pcmk_ok;
crm_element_value_int(op_reply, F_STONITH_CALLID, &reply_id);
if (reply_id == stonith->call_id) {
crm_trace("Synchronous reply %d received", reply_id);
if (crm_element_value_int(op_reply, F_STONITH_RC, &rc) != 0) {
rc = -ENOMSG;
}
if ((call_options & st_opt_discard_reply) || output_data == NULL) {
crm_trace("Discarding reply");
} else {
*output_data = op_reply;
op_reply = NULL; /* Prevent subsequent free */
}
} else if (reply_id <= 0) {
crm_err("Received bad reply: No id set");
crm_log_xml_err(op_reply, "Bad reply");
free_xml(op_reply);
rc = -ENOMSG;
} else {
crm_err("Received bad reply: %d (wanted %d)", reply_id, stonith->call_id);
crm_log_xml_err(op_reply, "Old reply");
free_xml(op_reply);
rc = -ENOMSG;
}
done:
if (crm_ipc_connected(native->ipc) == FALSE) {
crm_err("Fencer disconnected");
free(native->token); native->token = NULL;
stonith->state = stonith_disconnected;
}
free_xml(op_reply);
return rc;
}
/* Not used with mainloop */
bool
stonith_dispatch(stonith_t * st)
{
gboolean stay_connected = TRUE;
stonith_private_t *private = NULL;
CRM_ASSERT(st != NULL);
private = st->st_private;
while (crm_ipc_ready(private->ipc)) {
if (crm_ipc_read(private->ipc) > 0) {
const char *msg = crm_ipc_buffer(private->ipc);
stonith_dispatch_internal(msg, strlen(msg), st);
}
if (crm_ipc_connected(private->ipc) == FALSE) {
crm_err("Connection closed");
stay_connected = FALSE;
}
}
return stay_connected;
}
static int
stonith_api_free(stonith_t * stonith)
{
int rc = pcmk_ok;
crm_trace("Destroying %p", stonith);
if (stonith->state != stonith_disconnected) {
crm_trace("Disconnecting %p first", stonith);
rc = stonith->cmds->disconnect(stonith);
}
if (stonith->state == stonith_disconnected) {
stonith_private_t *private = stonith->st_private;
crm_trace("Removing %d callbacks", g_hash_table_size(private->stonith_op_callback_table));
g_hash_table_destroy(private->stonith_op_callback_table);
crm_trace("Destroying %d notification clients", g_list_length(private->notify_list));
g_list_free_full(private->notify_list, free);
free(stonith->st_private);
free(stonith->cmds);
free(stonith);
} else {
crm_err("Not free'ing active connection: %s (%d)", pcmk_strerror(rc), rc);
}
return rc;
}
void
stonith_api_delete(stonith_t * stonith)
{
crm_trace("Destroying %p", stonith);
if(stonith) {
stonith->cmds->free(stonith);
}
}
static int
stonith_api_validate(stonith_t *st, int call_options, const char *rsc_id,
const char *namespace_s, const char *agent,
stonith_key_value_t *params, int timeout, char **output,
char **error_output)
{
/* Validation should be done directly via the agent, so we can get it from
* stonith_admin when the cluster is not running, which is important for
* higher-level tools.
*/
int rc = pcmk_ok;
/* Use a dummy node name in case the agent requires a target. We assume the
* actual target doesn't matter for validation purposes (if in practice,
* that is incorrect, we will need to allow the caller to pass the target).
*/
const char *target = "node1";
GHashTable *params_table = crm_str_table_new();
// Convert parameter list to a hash table
for (; params; params = params->next) {
// Strip out Pacemaker-implemented parameters
if (!pcmk__starts_with(params->key, "pcmk_")
&& strcmp(params->key, "provides")
&& strcmp(params->key, "stonith-timeout")) {
g_hash_table_insert(params_table, strdup(params->key),
strdup(params->value));
}
}
#if SUPPORT_CIBSECRETS
rc = pcmk__substitute_secrets(rsc_id, params_table);
if (rc != pcmk_rc_ok) {
crm_warn("Could not replace secret parameters for validation of %s: %s",
agent, pcmk_rc_str(rc));
rc = pcmk_rc2legacy(rc);
}
#endif
if (output) {
*output = NULL;
}
if (error_output) {
*error_output = NULL;
}
switch (stonith_get_namespace(agent, namespace_s)) {
case st_namespace_rhcs:
rc = stonith__rhcs_validate(st, call_options, target, agent,
params_table, timeout, output,
error_output);
break;
#if HAVE_STONITH_STONITH_H
case st_namespace_lha:
rc = stonith__lha_validate(st, call_options, target, agent,
params_table, timeout, output,
error_output);
break;
#endif
default:
rc = -EINVAL;
errno = EINVAL;
crm_perror(LOG_ERR,
"Agent %s not found or does not support validation",
agent);
break;
}
g_hash_table_destroy(params_table);
return rc;
}
stonith_t *
stonith_api_new(void)
{
stonith_t *new_stonith = NULL;
stonith_private_t *private = NULL;
new_stonith = calloc(1, sizeof(stonith_t));
if (new_stonith == NULL) {
return NULL;
}
private = calloc(1, sizeof(stonith_private_t));
if (private == NULL) {
free(new_stonith);
return NULL;
}
new_stonith->st_private = private;
private->stonith_op_callback_table = g_hash_table_new_full(g_direct_hash, g_direct_equal,
NULL, stonith_destroy_op_callback);
private->notify_list = NULL;
private->notify_refcnt = 0;
private->notify_deletes = FALSE;
new_stonith->call_id = 1;
new_stonith->state = stonith_disconnected;
new_stonith->cmds = calloc(1, sizeof(stonith_api_operations_t));
if (new_stonith->cmds == NULL) {
free(new_stonith->st_private);
free(new_stonith);
return NULL;
}
/* *INDENT-OFF* */
new_stonith->cmds->free = stonith_api_free;
new_stonith->cmds->connect = stonith_api_signon;
new_stonith->cmds->disconnect = stonith_api_signoff;
new_stonith->cmds->list = stonith_api_list;
new_stonith->cmds->monitor = stonith_api_monitor;
new_stonith->cmds->status = stonith_api_status;
new_stonith->cmds->fence = stonith_api_fence;
+ new_stonith->cmds->fence_with_delay = stonith_api_fence_with_delay;
new_stonith->cmds->confirm = stonith_api_confirm;
new_stonith->cmds->history = stonith_api_history;
new_stonith->cmds->list_agents = stonith_api_device_list;
new_stonith->cmds->metadata = stonith_api_device_metadata;
new_stonith->cmds->query = stonith_api_query;
new_stonith->cmds->remove_device = stonith_api_remove_device;
new_stonith->cmds->register_device = stonith_api_register_device;
new_stonith->cmds->remove_level = stonith_api_remove_level;
new_stonith->cmds->remove_level_full = stonith_api_remove_level_full;
new_stonith->cmds->register_level = stonith_api_register_level;
new_stonith->cmds->register_level_full = stonith_api_register_level_full;
new_stonith->cmds->remove_callback = stonith_api_del_callback;
new_stonith->cmds->register_callback = stonith_api_add_callback;
new_stonith->cmds->remove_notification = stonith_api_del_notification;
new_stonith->cmds->register_notification = stonith_api_add_notification;
new_stonith->cmds->validate = stonith_api_validate;
/* *INDENT-ON* */
return new_stonith;
}
/*!
* \brief Make a blocking connection attempt to the fencer
*
* \param[in,out] st Fencer API object
* \param[in] name Client name to use with fencer
* \param[in] max_attempts Return error if this many attempts fail
*
* \return pcmk_ok on success, result of last attempt otherwise
*/
int
stonith_api_connect_retry(stonith_t *st, const char *name, int max_attempts)
{
int rc = -EINVAL; // if max_attempts is not positive
for (int attempt = 1; attempt <= max_attempts; attempt++) {
rc = st->cmds->connect(st, name, NULL);
if (rc == pcmk_ok) {
return pcmk_ok;
} else if (attempt < max_attempts) {
crm_notice("Fencer connection attempt %d of %d failed (retrying in 2s): %s "
CRM_XS " rc=%d",
attempt, max_attempts, pcmk_strerror(rc), rc);
sleep(2);
}
}
crm_notice("Could not connect to fencer: %s " CRM_XS " rc=%d",
pcmk_strerror(rc), rc);
return rc;
}
stonith_key_value_t *
stonith_key_value_add(stonith_key_value_t * head, const char *key, const char *value)
{
stonith_key_value_t *p, *end;
p = calloc(1, sizeof(stonith_key_value_t));
if (key) {
p->key = strdup(key);
}
if (value) {
p->value = strdup(value);
}
end = head;
while (end && end->next) {
end = end->next;
}
if (end) {
end->next = p;
} else {
head = p;
}
return head;
}
void
stonith_key_value_freeall(stonith_key_value_t * head, int keys, int values)
{
stonith_key_value_t *p;
while (head) {
p = head->next;
if (keys) {
free(head->key);
}
if (values) {
free(head->value);
}
free(head);
head = p;
}
}
#define api_log_open() openlog("stonith-api", LOG_CONS | LOG_NDELAY | LOG_PID, LOG_DAEMON)
#define api_log(level, fmt, args...) syslog(level, "%s: "fmt, __FUNCTION__, args)
int
stonith_api_kick(uint32_t nodeid, const char *uname, int timeout, bool off)
{
int rc = pcmk_ok;
stonith_t *st = stonith_api_new();
const char *action = off? "off" : "reboot";
api_log_open();
if (st == NULL) {
api_log(LOG_ERR, "API initialization failed, could not kick (%s) node %u/%s",
action, nodeid, uname);
return -EPROTO;
}
rc = st->cmds->connect(st, "stonith-api", NULL);
if (rc != pcmk_ok) {
api_log(LOG_ERR, "Connection failed, could not kick (%s) node %u/%s : %s (%d)",
action, nodeid, uname, pcmk_strerror(rc), rc);
} else {
char *name = NULL;
enum stonith_call_options opts = st_opt_sync_call | st_opt_allow_suicide;
if (uname != NULL) {
name = strdup(uname);
} else if (nodeid > 0) {
opts |= st_opt_cs_nodeid;
name = crm_itoa(nodeid);
}
rc = st->cmds->fence(st, opts, name, action, timeout, 0);
free(name);
if (rc != pcmk_ok) {
api_log(LOG_ERR, "Could not kick (%s) node %u/%s : %s (%d)",
action, nodeid, uname, pcmk_strerror(rc), rc);
} else {
api_log(LOG_NOTICE, "Node %u/%s kicked: %s", nodeid, uname, action);
}
}
stonith_api_delete(st);
return rc;
}
time_t
stonith_api_time(uint32_t nodeid, const char *uname, bool in_progress)
{
int rc = pcmk_ok;
time_t when = 0;
stonith_t *st = stonith_api_new();
stonith_history_t *history = NULL, *hp = NULL;
if (st == NULL) {
api_log(LOG_ERR, "Could not retrieve fence history for %u/%s: "
"API initialization failed", nodeid, uname);
return when;
}
rc = st->cmds->connect(st, "stonith-api", NULL);
if (rc != pcmk_ok) {
api_log(LOG_NOTICE, "Connection failed: %s (%d)", pcmk_strerror(rc), rc);
} else {
int entries = 0;
int progress = 0;
int completed = 0;
char *name = NULL;
enum stonith_call_options opts = st_opt_sync_call;
if (uname != NULL) {
name = strdup(uname);
} else if (nodeid > 0) {
opts |= st_opt_cs_nodeid;
name = crm_itoa(nodeid);
}
rc = st->cmds->history(st, opts, name, &history, 120);
free(name);
for (hp = history; hp; hp = hp->next) {
entries++;
if (in_progress) {
progress++;
if (hp->state != st_done && hp->state != st_failed) {
when = time(NULL);
}
} else if (hp->state == st_done) {
completed++;
if (hp->completed > when) {
when = hp->completed;
}
}
}
stonith_history_free(history);
if(rc == pcmk_ok) {
api_log(LOG_INFO, "Found %d entries for %u/%s: %d in progress, %d completed", entries, nodeid, uname, progress, completed);
} else {
api_log(LOG_ERR, "Could not retrieve fence history for %u/%s: %s (%d)", nodeid, uname, pcmk_strerror(rc), rc);
}
}
stonith_api_delete(st);
if(when) {
api_log(LOG_INFO, "Node %u/%s last kicked at: %ld", nodeid, uname, (long int)when);
}
return when;
}
bool
stonith_agent_exists(const char *agent, int timeout)
{
stonith_t *st = NULL;
stonith_key_value_t *devices = NULL;
stonith_key_value_t *dIter = NULL;
bool rc = FALSE;
if (agent == NULL) {
return rc;
}
st = stonith_api_new();
if (st == NULL) {
crm_err("Could not list fence agents: API memory allocation failed");
return FALSE;
}
st->cmds->list_agents(st, st_opt_sync_call, NULL, &devices, timeout == 0 ? 120 : timeout);
for (dIter = devices; dIter != NULL; dIter = dIter->next) {
if (crm_str_eq(dIter->value, agent, TRUE)) {
rc = TRUE;
break;
}
}
stonith_key_value_freeall(devices, 1, 1);
stonith_api_delete(st);
return rc;
}
const char *
stonith_action_str(const char *action)
{
if (action == NULL) {
return "fencing";
} else if (!strcmp(action, "on")) {
return "unfencing";
} else if (!strcmp(action, "off")) {
return "turning off";
} else {
return action;
}
}
/*!
* \internal
* \brief Parse a target name from one line of a target list string
*
* \param[in] line One line of a target list string
* \parma[in] len String length of line
* \param[in,out] output List to add newly allocated target name to
*/
static void
parse_list_line(const char *line, int len, GList **output)
{
size_t i = 0;
size_t entry_start = 0;
/* Skip complaints about additional parameters device doesn't understand
*
* @TODO Document or eliminate the implied restriction of target names
*/
if (strstr(line, "invalid") || strstr(line, "variable")) {
crm_debug("Skipping list output line: %s", line);
return;
}
// Process line content, character by character
for (i = 0; i <= len; i++) {
if (isspace(line[i]) || (line[i] == ',') || (line[i] == ';')
|| (line[i] == '\0')) {
// We've found a separator (i.e. the end of an entry)
int rc = 0;
char *entry = NULL;
if (i == entry_start) {
// Skip leading and sequential separators
entry_start = i + 1;
continue;
}
entry = calloc(i - entry_start + 1, sizeof(char));
CRM_ASSERT(entry != NULL);
/* Read entry, stopping at first separator
*
* @TODO Document or eliminate these character restrictions
*/
rc = sscanf(line + entry_start, "%[a-zA-Z0-9_-.]", entry);
if (rc != 1) {
crm_warn("Could not parse list output entry: %s "
CRM_XS " entry_start=%d position=%d",
line + entry_start, entry_start, i);
free(entry);
} else if (safe_str_eq(entry, "on") || safe_str_eq(entry, "off")) {
/* Some agents print the target status in the list output,
* though none are known now (the separate list-status command
* is used for this, but it can also print "UNKNOWN"). To handle
* this possibility, skip such entries.
*
* @TODO Document or eliminate the implied restriction of target
* names.
*/
free(entry);
} else {
// We have a valid entry
*output = g_list_append(*output, entry);
}
entry_start = i + 1;
}
}
}
/*!
* \internal
* \brief Parse a list of targets from a string
*
* \param[in] list_output Target list as a string
*
* \return List of target names
* \note The target list string format is flexible, to allow for user-specified
* lists such pcmk_host_list and the output of an agent's list action
* (whether direct or via the API, which escapes newlines). There may be
* multiple lines, separated by either a newline or an escaped newline
* (backslash n). Each line may have one or more target names, separated
* by any combination of whitespace, commas, and semi-colons. Lines
* containing "invalid" or "variable" will be ignored entirely. Target
* names "on" or "off" (case-insensitive) will be ignored. Target names
* may contain only alphanumeric characters, underbars (_), dashes (-),
* and dots (.) (if any other character occurs in the name, it and all
* subsequent characters in the name will be ignored).
* \note The caller is responsible for freeing the result with
* g_list_free_full(result, free).
*/
GList *
stonith__parse_targets(const char *target_spec)
{
GList *targets = NULL;
if (target_spec != NULL) {
size_t out_len = strlen(target_spec);
size_t line_start = 0; // Starting index of line being processed
for (size_t i = 0; i <= out_len; ++i) {
if ((target_spec[i] == '\n') || (target_spec[i] == '\0')
|| ((target_spec[i] == '\\') && (target_spec[i + 1] == 'n'))) {
// We've reached the end of one line of output
int len = i - line_start;
if (len > 0) {
char *line = strndup(target_spec + line_start, len);
line[len] = '\0'; // Because it might be a newline
parse_list_line(line, len, &targets);
free(line);
}
if (target_spec[i] == '\\') {
++i; // backslash-n takes up two positions
}
line_start = i + 1;
}
}
}
return targets;
}
/*!
* \internal
* \brief Determine if a later stonith event succeeded.
*
* \note Before calling this function, use stonith__sort_history() to sort the
* top_history argument.
*/
gboolean
stonith__later_succeeded(stonith_history_t *event, stonith_history_t *top_history)
{
gboolean ret = FALSE;
for (stonith_history_t *prev_hp = top_history; prev_hp; prev_hp = prev_hp->next) {
if (prev_hp == event) {
break;
}
if ((prev_hp->state == st_done) &&
safe_str_eq(event->target, prev_hp->target) &&
safe_str_eq(event->action, prev_hp->action) &&
safe_str_eq(event->delegate, prev_hp->delegate) &&
(event->completed < prev_hp->completed)) {
ret = TRUE;
break;
}
}
return ret;
}
/*!
* \internal
* \brief Sort the stonith-history
* sort by competed most current on the top
* pending actions lacking a completed-stamp are gathered at the top
*
* \param[in] history List of stonith actions
*
*/
stonith_history_t *
stonith__sort_history(stonith_history_t *history)
{
stonith_history_t *new = NULL, *pending = NULL, *hp, *np, *tmp;
for (hp = history; hp; ) {
tmp = hp->next;
if ((hp->state == st_done) || (hp->state == st_failed)) {
/* sort into new */
if ((!new) || (hp->completed > new->completed)) {
hp->next = new;
new = hp;
} else {
np = new;
do {
if ((!np->next) || (hp->completed > np->next->completed)) {
hp->next = np->next;
np->next = hp;
break;
}
np = np->next;
} while (1);
}
} else {
/* put into pending */
hp->next = pending;
pending = hp;
}
hp = tmp;
}
/* pending actions don't have a completed-stamp so make them go front */
if (pending) {
stonith_history_t *last_pending = pending;
while (last_pending->next) {
last_pending = last_pending->next;
}
last_pending->next = new;
new = pending;
}
return new;
}
// Deprecated functions kept only for backward API compatibility
const char *get_stonith_provider(const char *agent, const char *provider);
/*!
* \brief Deprecated (use stonith_get_namespace() instead)
*/
const char *
get_stonith_provider(const char *agent, const char *provider)
{
return stonith_namespace2text(stonith_get_namespace(agent, provider));
}
diff --git a/lib/pacemaker/pcmk_fence.c b/lib/pacemaker/pcmk_fence.c
index 86e5895ee6..5e122bc73c 100644
--- a/lib/pacemaker/pcmk_fence.c
+++ b/lib/pacemaker/pcmk_fence.c
@@ -1,487 +1,493 @@
/*
* Copyright 2009-2020 the Pacemaker project contributors
*
* The version control history for this file may have further details.
*
* This source code is licensed under the GNU General Public License version 2
* or later (GPLv2+) WITHOUT ANY WARRANTY.
*/
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
static int st_opts = st_opt_sync_call | st_opt_allow_suicide;
static GMainLoop *mainloop = NULL;
static struct {
stonith_t *st;
const char *target;
const char *action;
char *name;
unsigned int timeout;
unsigned int tolerance;
+ int delay;
int rc;
} async_fence_data;
static int
handle_level(stonith_t *st, char *target, int fence_level,
stonith_key_value_t *devices, bool added) {
char *node = NULL;
char *pattern = NULL;
char *name = NULL;
char *value = NULL;
int rc = pcmk_rc_ok;
if (target == NULL) {
// Not really possible, but makes static analysis happy
return EINVAL;
}
/* Determine if targeting by attribute, node name pattern or node name */
value = strchr(target, '=');
if (value != NULL) {
name = target;
*value++ = '\0';
} else if (*target == '@') {
pattern = target + 1;
} else {
node = target;
}
/* Register or unregister level as appropriate */
if (added) {
rc = st->cmds->register_level_full(st, st_opts, node, pattern,
name, value, fence_level,
devices);
} else {
rc = st->cmds->remove_level_full(st, st_opts, node, pattern,
name, value, fence_level);
}
return pcmk_legacy2rc(rc);
}
static void
notify_callback(stonith_t * st, stonith_event_t * e)
{
if (e->result != pcmk_ok) {
return;
}
if (safe_str_eq(async_fence_data.target, e->target) &&
safe_str_eq(async_fence_data.action, e->action)) {
async_fence_data.rc = e->result;
g_main_loop_quit(mainloop);
}
}
static void
fence_callback(stonith_t * stonith, stonith_callback_data_t * data)
{
async_fence_data.rc = data->rc;
g_main_loop_quit(mainloop);
}
static gboolean
async_fence_helper(gpointer user_data)
{
stonith_t *st = async_fence_data.st;
int call_id = 0;
int rc = stonith_api_connect_retry(st, async_fence_data.name, 10);
if (rc != pcmk_ok) {
fprintf(stderr, "Could not connect to fencer: %s\n", pcmk_strerror(rc));
g_main_loop_quit(mainloop);
return TRUE;
}
st->cmds->register_notification(st, T_STONITH_NOTIFY_FENCE, notify_callback);
- call_id = st->cmds->fence(st,
- st_opt_allow_suicide,
- async_fence_data.target,
- async_fence_data.action,
- async_fence_data.timeout/1000,
- async_fence_data.tolerance/1000);
+ call_id = st->cmds->fence_with_delay(st,
+ st_opt_allow_suicide,
+ async_fence_data.target,
+ async_fence_data.action,
+ async_fence_data.timeout/1000,
+ async_fence_data.tolerance/1000,
+ async_fence_data.delay);
if (call_id < 0) {
g_main_loop_quit(mainloop);
return TRUE;
}
st->cmds->register_callback(st,
call_id,
async_fence_data.timeout/1000,
st_opt_timeout_updates, NULL, "callback", fence_callback);
return TRUE;
}
int
pcmk__fence_action(stonith_t *st, const char *target, const char *action,
- const char *name, unsigned int timeout, unsigned int tolerance)
+ const char *name, unsigned int timeout, unsigned int tolerance,
+ int delay)
{
crm_trigger_t *trig;
async_fence_data.st = st;
async_fence_data.name = strdup(name);
async_fence_data.target = target;
async_fence_data.action = action;
async_fence_data.timeout = timeout;
async_fence_data.tolerance = tolerance;
+ async_fence_data.delay = delay;
async_fence_data.rc = pcmk_err_generic;
trig = mainloop_add_trigger(G_PRIORITY_HIGH, async_fence_helper, NULL);
mainloop_set_trigger(trig);
mainloop = g_main_loop_new(NULL, FALSE);
g_main_loop_run(mainloop);
free(async_fence_data.name);
return pcmk_legacy2rc(async_fence_data.rc);
}
#ifdef BUILD_PUBLIC_LIBPACEMAKER
int
pcmk_fence_action(stonith_t *st, const char *target, const char *action,
- const char *name, unsigned int timeout, unsigned int tolerance) {
- return pcmk__fence_action(st, target, action, name, timeout, tolerance);
+ const char *name, unsigned int timeout, unsigned int tolerance,
+ int delay)
+{
+ return pcmk__fence_action(st, target, action, name, timeout, tolerance, delay);
}
#endif
int
pcmk__fence_history(pcmk__output_t *out, stonith_t *st, char *target,
unsigned int timeout, bool quiet, int verbose,
bool broadcast, bool cleanup) {
stonith_history_t *history = NULL, *hp, *latest = NULL;
int rc = pcmk_rc_ok;
if (!quiet) {
if (cleanup) {
out->info(out, "cleaning up fencing-history%s%s",
target ? " for node " : "", target ? target : "");
}
if (broadcast) {
out->info(out, "gather fencing-history from all nodes");
}
}
rc = st->cmds->history(st, st_opts | (cleanup ? st_opt_cleanup : 0) |
(broadcast ? st_opt_broadcast : 0),
safe_str_eq(target, "*") ? NULL : target,
&history, timeout/1000);
out->begin_list(out, "event", "events", "Fencing history");
history = stonith__sort_history(history);
for (hp = history; hp; hp = hp->next) {
if (hp->state == st_done) {
latest = hp;
}
if (quiet || !verbose) {
continue;
}
out->message(out, "stonith-event", hp, 1, stonith__later_succeeded(hp, history));
out->increment_list(out);
}
if (latest) {
if (quiet && out->supports_quiet) {
out->info(out, "%lld", (long long) latest->completed);
} else if (!verbose) { // already printed if verbose
out->message(out, "stonith-event", latest, 0, FALSE);
out->increment_list(out);
}
}
out->end_list(out);
stonith_history_free(history);
return pcmk_legacy2rc(rc);
}
#ifdef BUILD_PUBLIC_LIBPACEMAKER
int
pcmk_fence_history(xmlNodePtr *xml, stonith_t *st, char *target, unsigned int timeout,
bool quiet, int verbose, bool broadcast, bool cleanup) {
pcmk__output_t *out = NULL;
int rc = pcmk_rc_ok;
rc = pcmk__out_prologue(&out, xml);
if (rc != pcmk_rc_ok) {
return rc;
}
rc = pcmk__fence_history(out, st, target, timeout, quiet, verbose,
broadcast, cleanup);
pcmk__out_epilogue(out, xml, rc);
return rc;
}
#endif
int
pcmk__fence_installed(pcmk__output_t *out, stonith_t *st, unsigned int timeout) {
stonith_key_value_t *devices = NULL;
int rc = pcmk_rc_ok;
rc = st->cmds->list_agents(st, st_opt_sync_call, NULL, &devices, timeout/1000);
/* list_agents returns a negative error code or a positive number of agents. */
if (rc < 0) {
return pcmk_legacy2rc(rc);
}
out->begin_list(out, "fence device", "fence devices", "Installed fence devices");
for (stonith_key_value_t *dIter = devices; dIter; dIter = dIter->next) {
out->list_item(out, "device", "%s", dIter->value);
}
out->end_list(out);
stonith_key_value_freeall(devices, 1, 1);
return pcmk_rc_ok;
}
#ifdef BUILD_PUBLIC_LIBPACEMAKER
int
pcmk_fence_installed(xmlNodePtr *xml, stonith_t *st, unsigned int timeout) {
pcmk__output_t *out = NULL;
int rc = pcmk_rc_ok;
rc = pcmk__out_prologue(&out, xml);
if (rc != pcmk_rc_ok) {
return rc;
}
rc = pcmk__fence_installed(out, st, timeout);
pcmk__out_epilogue(out, xml, rc);
return rc;
}
#endif
int
pcmk__fence_last(pcmk__output_t *out, const char *target, bool as_nodeid) {
time_t when = 0;
if (target == NULL) {
return pcmk_rc_ok;
}
if (as_nodeid) {
when = stonith_api_time(atol(target), NULL, FALSE);
} else {
when = stonith_api_time(0, target, FALSE);
}
return out->message(out, "last-fenced", target, when);
}
#ifdef BUILD_PUBLIC_LIBPACEMAKER
int
pcmk_fence_last(xmlNodePtr *xml, const char *target, bool as_nodeid) {
pcmk__output_t *out = NULL;
int rc = pcmk_rc_ok;
rc = pcmk__out_prologue(&out, xml);
if (rc != pcmk_rc_ok) {
return rc;
}
rc = pcmk__fence_last(out, target, as_nodeid);
pcmk__out_epilogue(out, xml, rc);
return rc;
}
#endif
int
pcmk__fence_list_targets(pcmk__output_t *out, stonith_t *st, char *agent,
unsigned int timeout) {
GList *targets = NULL;
char *lists = NULL;
int rc = pcmk_rc_ok;
rc = st->cmds->list(st, st_opts, agent, &lists, timeout/1000);
if (rc != pcmk_rc_ok) {
return pcmk_legacy2rc(rc);
}
targets = stonith__parse_targets(lists);
out->begin_list(out, "fence target", "fence targets", "Fence Targets");
while (targets != NULL) {
out->list_item(out, NULL, "%s", (const char *) targets->data);
targets = targets->next;
}
out->end_list(out);
free(lists);
return rc;
}
#ifdef BUILD_PUBLIC_LIBPACEMAKER
int
pcmk_fence_list_targets(xmlNodePtr *xml, stonith_t *st, char *agent,
unsigned int timeout) {
pcmk__output_t *out = NULL;
int rc = pcmk_rc_ok;
rc = pcmk__out_prologue(&out, xml);
if (rc != pcmk_rc_ok) {
return rc;
}
rc = pcmk__fence_list_targets(out, st, agent, timeout);
pcmk__out_epilogue(out, xml, rc);
return rc;
}
#endif
int
pcmk__fence_metadata(pcmk__output_t *out, stonith_t *st, char *agent,
unsigned int timeout) {
char *buffer = NULL;
int rc = st->cmds->metadata(st, st_opt_sync_call, agent, NULL, &buffer,
timeout/1000);
if (rc != pcmk_rc_ok) {
return pcmk_legacy2rc(rc);
}
out->output_xml(out, "metadata", buffer);
free(buffer);
return rc;
}
#ifdef BUILD_PUBLIC_LIBPACEMAKER
int
pcmk_fence_metadata(xmlNodePtr *xml, stonith_t *st, char *agent,
unsigned int timeout) {
pcmk__output_t *out = NULL;
int rc = pcmk_rc_ok;
rc = pcmk__out_prologue(&out, xml);
if (rc != pcmk_rc_ok) {
return rc;
}
rc = pcmk__fence_metadata(out, st, agent, timeout);
pcmk__out_epilogue(out, xml, rc);
return rc;
}
#endif
int
pcmk__fence_registered(pcmk__output_t *out, stonith_t *st, char *target,
unsigned int timeout) {
stonith_key_value_t *devices = NULL;
int rc = pcmk_rc_ok;
rc = st->cmds->query(st, st_opts, target, &devices, timeout/1000);
/* query returns a negative error code or a positive number of results. */
if (rc < 0) {
return pcmk_legacy2rc(rc);
}
out->begin_list(out, "fence device", "fence devices", "Registered fence devices");
for (stonith_key_value_t *dIter = devices; dIter; dIter = dIter->next) {
out->list_item(out, "device", "%s", dIter->value);
}
out->end_list(out);
stonith_key_value_freeall(devices, 1, 1);
/* Return pcmk_rc_ok here, not the number of results. Callers probably
* don't care.
*/
return pcmk_rc_ok;
}
#ifdef BUILD_PUBLIC_LIBPACEMAKER
int
pcmk_fence_registered(xmlNodePtr *xml, stonith_t *st, char *target,
unsigned int timeout) {
pcmk__output_t *out = NULL;
int rc = pcmk_rc_ok;
rc = pcmk__out_prologue(&out, xml);
if (rc != pcmk_rc_ok) {
return rc;
}
rc = pcmk__fence_registered(out, st, target, timeout);
pcmk__out_epilogue(out, xml, rc);
return rc;
}
#endif
int
pcmk__fence_register_level(stonith_t *st, char *target, int fence_level,
stonith_key_value_t *devices) {
return handle_level(st, target, fence_level, devices, true);
}
#ifdef BUILD_PUBLIC_LIBPACEMAKER
int
pcmk_fence_register_level(stonith_t *st, char *target, int fence_level,
stonith_key_value_t *devices) {
return pcmk__fence_register_level(st, target, fence_level, devices);
}
#endif
int
pcmk__fence_unregister_level(stonith_t *st, char *target, int fence_level) {
return handle_level(st, target, fence_level, NULL, false);
}
#ifdef BUILD_PUBLIC_LIBPACEMAKER
int
pcmk_fence_unregister_level(stonith_t *st, char *target, int fence_level) {
return pcmk__fence_unregister_level(st, target, fence_level);
}
#endif
int
pcmk__fence_validate(pcmk__output_t *out, stonith_t *st, const char *agent,
const char *id, stonith_key_value_t *params,
unsigned int timeout) {
char *output = NULL;
char *error_output = NULL;
int rc;
rc = st->cmds->validate(st, st_opt_sync_call, id, NULL, agent, params,
timeout/1000, &output, &error_output);
out->message(out, "validate", agent, id, output, error_output, rc);
return pcmk_legacy2rc(rc);
}
#ifdef BUILD_PUBLIC_LIBPACEMAKER
int
pcmk_fence_validate(xmlNodePtr *xml, stonith_t *st, const char *agent,
const char *id, stonith_key_value_t *params,
unsigned int timeout) {
pcmk__output_t *out = NULL;
int rc = pcmk_rc_ok;
rc = pcmk__out_prologue(&out, xml);
if (rc != pcmk_rc_ok) {
return rc;
}
rc = pcmk__fence_validate(out, st, agent, id, params, timeout);
pcmk__out_epilogue(out, xml, rc);
return rc;
}
#endif
diff --git a/lib/pacemaker/pcmk_sched_allocate.c b/lib/pacemaker/pcmk_sched_allocate.c
index 5a39f79f8e..156f9668a5 100644
--- a/lib/pacemaker/pcmk_sched_allocate.c
+++ b/lib/pacemaker/pcmk_sched_allocate.c
@@ -1,3068 +1,3068 @@
/*
* Copyright 2004-2020 the Pacemaker project contributors
*
* The version control history for this file may have further details.
*
* This source code is licensed under the GNU General Public License version 2
* or later (GPLv2+) WITHOUT ANY WARRANTY.
*/
#include
#include
#include
#include
#include
#include
#include
#include
#include
CRM_TRACE_INIT_DATA(pacemaker);
void set_alloc_actions(pe_working_set_t * data_set);
extern void ReloadRsc(pe_resource_t * rsc, pe_node_t *node, pe_working_set_t * data_set);
extern gboolean DeleteRsc(pe_resource_t * rsc, pe_node_t * node, gboolean optional, pe_working_set_t * data_set);
static void apply_remote_node_ordering(pe_working_set_t *data_set);
static enum remote_connection_state get_remote_node_state(pe_node_t *node);
enum remote_connection_state {
remote_state_unknown = 0,
remote_state_alive = 1,
remote_state_resting = 2,
remote_state_failed = 3,
remote_state_stopped = 4
};
static const char *
state2text(enum remote_connection_state state)
{
switch (state) {
case remote_state_unknown:
return "unknown";
case remote_state_alive:
return "alive";
case remote_state_resting:
return "resting";
case remote_state_failed:
return "failed";
case remote_state_stopped:
return "stopped";
}
return "impossible";
}
resource_alloc_functions_t resource_class_alloc_functions[] = {
{
pcmk__native_merge_weights,
pcmk__native_allocate,
native_create_actions,
native_create_probe,
native_internal_constraints,
native_rsc_colocation_lh,
native_rsc_colocation_rh,
native_rsc_location,
native_action_flags,
native_update_actions,
native_expand,
native_append_meta,
},
{
pcmk__group_merge_weights,
pcmk__group_allocate,
group_create_actions,
native_create_probe,
group_internal_constraints,
group_rsc_colocation_lh,
group_rsc_colocation_rh,
group_rsc_location,
group_action_flags,
group_update_actions,
group_expand,
group_append_meta,
},
{
pcmk__native_merge_weights,
pcmk__clone_allocate,
clone_create_actions,
clone_create_probe,
clone_internal_constraints,
clone_rsc_colocation_lh,
clone_rsc_colocation_rh,
clone_rsc_location,
clone_action_flags,
pcmk__multi_update_actions,
clone_expand,
clone_append_meta,
},
{
pcmk__native_merge_weights,
pcmk__bundle_allocate,
pcmk__bundle_create_actions,
pcmk__bundle_create_probe,
pcmk__bundle_internal_constraints,
pcmk__bundle_rsc_colocation_lh,
pcmk__bundle_rsc_colocation_rh,
pcmk__bundle_rsc_location,
pcmk__bundle_action_flags,
pcmk__multi_update_actions,
pcmk__bundle_expand,
pcmk__bundle_append_meta,
}
};
gboolean
update_action_flags(pe_action_t * action, enum pe_action_flags flags, const char *source, int line)
{
static unsigned long calls = 0;
gboolean changed = FALSE;
gboolean clear = is_set(flags, pe_action_clear);
enum pe_action_flags last = action->flags;
if (clear) {
action->flags = crm_clear_bit(source, line, action->uuid, action->flags, flags);
} else {
action->flags = crm_set_bit(source, line, action->uuid, action->flags, flags);
}
if (last != action->flags) {
calls++;
changed = TRUE;
/* Useful for tracking down _who_ changed a specific flag */
/* CRM_ASSERT(calls != 534); */
clear_bit(flags, pe_action_clear);
crm_trace("%s on %s: %sset flags 0x%.6x (was 0x%.6x, now 0x%.6x, %lu, %s)",
action->uuid, action->node ? action->node->details->uname : "[none]",
clear ? "un-" : "", flags, last, action->flags, calls, source);
}
return changed;
}
static gboolean
check_rsc_parameters(pe_resource_t * rsc, pe_node_t * node, xmlNode * rsc_entry,
gboolean active_here, pe_working_set_t * data_set)
{
int attr_lpc = 0;
gboolean force_restart = FALSE;
gboolean delete_resource = FALSE;
gboolean changed = FALSE;
const char *value = NULL;
const char *old_value = NULL;
const char *attr_list[] = {
XML_ATTR_TYPE,
XML_AGENT_ATTR_CLASS,
XML_AGENT_ATTR_PROVIDER
};
for (; attr_lpc < DIMOF(attr_list); attr_lpc++) {
value = crm_element_value(rsc->xml, attr_list[attr_lpc]);
old_value = crm_element_value(rsc_entry, attr_list[attr_lpc]);
if (value == old_value /* i.e. NULL */
|| crm_str_eq(value, old_value, TRUE)) {
continue;
}
changed = TRUE;
trigger_unfencing(rsc, node, "Device definition changed", NULL, data_set);
if (active_here) {
force_restart = TRUE;
crm_notice("Forcing restart of %s on %s, %s changed: %s -> %s",
rsc->id, node->details->uname, attr_list[attr_lpc],
crm_str(old_value), crm_str(value));
}
}
if (force_restart) {
/* make sure the restart happens */
stop_action(rsc, node, FALSE);
set_bit(rsc->flags, pe_rsc_start_pending);
delete_resource = TRUE;
} else if (changed) {
delete_resource = TRUE;
}
return delete_resource;
}
static void
CancelXmlOp(pe_resource_t * rsc, xmlNode * xml_op, pe_node_t * active_node,
const char *reason, pe_working_set_t * data_set)
{
guint interval_ms = 0;
pe_action_t *cancel = NULL;
const char *task = NULL;
const char *call_id = NULL;
CRM_CHECK(xml_op != NULL, return);
CRM_CHECK(active_node != NULL, return);
task = crm_element_value(xml_op, XML_LRM_ATTR_TASK);
call_id = crm_element_value(xml_op, XML_LRM_ATTR_CALLID);
crm_element_value_ms(xml_op, XML_LRM_ATTR_INTERVAL_MS, &interval_ms);
crm_info("Action " PCMK__OP_FMT " on %s will be stopped: %s",
rsc->id, task, interval_ms,
active_node->details->uname, (reason? reason : "unknown"));
cancel = pe_cancel_op(rsc, task, interval_ms, active_node, data_set);
add_hash_param(cancel->meta, XML_LRM_ATTR_CALLID, call_id);
custom_action_order(rsc, stop_key(rsc), NULL, rsc, NULL, cancel, pe_order_optional, data_set);
}
static gboolean
check_action_definition(pe_resource_t * rsc, pe_node_t * active_node, xmlNode * xml_op,
pe_working_set_t * data_set)
{
char *key = NULL;
guint interval_ms = 0;
const op_digest_cache_t *digest_data = NULL;
gboolean did_change = FALSE;
const char *task = crm_element_value(xml_op, XML_LRM_ATTR_TASK);
const char *digest_secure = NULL;
CRM_CHECK(active_node != NULL, return FALSE);
crm_element_value_ms(xml_op, XML_LRM_ATTR_INTERVAL_MS, &interval_ms);
if (interval_ms > 0) {
xmlNode *op_match = NULL;
/* we need to reconstruct the key because of the way we used to construct resource IDs */
key = pcmk__op_key(rsc->id, task, interval_ms);
pe_rsc_trace(rsc, "Checking parameters for %s", key);
op_match = find_rsc_op_entry(rsc, key);
if (op_match == NULL && is_set(data_set->flags, pe_flag_stop_action_orphans)) {
CancelXmlOp(rsc, xml_op, active_node, "orphan", data_set);
free(key);
return TRUE;
} else if (op_match == NULL) {
pe_rsc_debug(rsc, "Orphan action detected: %s on %s", key, active_node->details->uname);
free(key);
return TRUE;
}
free(key);
key = NULL;
}
crm_trace("Testing " PCMK__OP_FMT " on %s",
rsc->id, task, interval_ms, active_node->details->uname);
if ((interval_ms == 0) && safe_str_eq(task, RSC_STATUS)) {
/* Reload based on the start action not a probe */
task = RSC_START;
} else if ((interval_ms == 0) && safe_str_eq(task, RSC_MIGRATED)) {
/* Reload based on the start action not a migrate */
task = RSC_START;
} else if ((interval_ms == 0) && safe_str_eq(task, RSC_PROMOTE)) {
/* Reload based on the start action not a promote */
task = RSC_START;
}
digest_data = rsc_action_digest_cmp(rsc, xml_op, active_node, data_set);
if(is_set(data_set->flags, pe_flag_sanitized)) {
digest_secure = crm_element_value(xml_op, XML_LRM_ATTR_SECURE_DIGEST);
}
if(digest_data->rc != RSC_DIGEST_MATCH
&& digest_secure
&& digest_data->digest_secure_calc
&& strcmp(digest_data->digest_secure_calc, digest_secure) == 0) {
if (is_set(data_set->flags, pe_flag_stdout)) {
printf("Only 'private' parameters to " PCMK__OP_FMT
" on %s changed: %s\n",
rsc->id, task, interval_ms, active_node->details->uname,
crm_element_value(xml_op, XML_ATTR_TRANSITION_MAGIC));
}
} else if (digest_data->rc == RSC_DIGEST_RESTART) {
/* Changes that force a restart */
pe_action_t *required = NULL;
did_change = TRUE;
key = pcmk__op_key(rsc->id, task, interval_ms);
crm_log_xml_info(digest_data->params_restart, "params:restart");
required = custom_action(rsc, key, task, NULL, TRUE, TRUE, data_set);
pe_action_set_flag_reason(__FUNCTION__, __LINE__, required, NULL,
"resource definition change", pe_action_optional, TRUE);
trigger_unfencing(rsc, active_node, "Device parameters changed", NULL, data_set);
} else if ((digest_data->rc == RSC_DIGEST_ALL) || (digest_data->rc == RSC_DIGEST_UNKNOWN)) {
/* Changes that can potentially be handled by a reload */
const char *digest_restart = crm_element_value(xml_op, XML_LRM_ATTR_RESTART_DIGEST);
did_change = TRUE;
trigger_unfencing(rsc, active_node, "Device parameters changed (reload)", NULL, data_set);
crm_log_xml_info(digest_data->params_all, "params:reload");
key = pcmk__op_key(rsc->id, task, interval_ms);
if (interval_ms > 0) {
pe_action_t *op = NULL;
#if 0
/* Always reload/restart the entire resource */
ReloadRsc(rsc, active_node, data_set);
#else
/* Re-sending the recurring op is sufficient - the old one will be cancelled automatically */
op = custom_action(rsc, key, task, active_node, TRUE, TRUE, data_set);
set_bit(op->flags, pe_action_reschedule);
#endif
} else if (digest_restart) {
pe_rsc_trace(rsc, "Reloading '%s' action for resource %s", task, rsc->id);
/* Reload this resource */
ReloadRsc(rsc, active_node, data_set);
free(key);
} else {
pe_action_t *required = NULL;
pe_rsc_trace(rsc, "Resource %s doesn't know how to reload", rsc->id);
/* Re-send the start/demote/promote op
* Recurring ops will be detected independently
*/
required = custom_action(rsc, key, task, NULL, TRUE, TRUE, data_set);
pe_action_set_flag_reason(__FUNCTION__, __LINE__, required, NULL,
"resource definition change", pe_action_optional, TRUE);
}
}
return did_change;
}
/*!
* \internal
* \brief Do deferred action checks after allocation
*
* \param[in] data_set Working set for cluster
*/
static void
check_params(pe_resource_t *rsc, pe_node_t *node, xmlNode *rsc_op,
enum pe_check_parameters check, pe_working_set_t *data_set)
{
const char *reason = NULL;
op_digest_cache_t *digest_data = NULL;
switch (check) {
case pe_check_active:
if (check_action_definition(rsc, node, rsc_op, data_set)
&& pe_get_failcount(node, rsc, NULL, pe_fc_effective, NULL,
data_set)) {
reason = "action definition changed";
}
break;
case pe_check_last_failure:
digest_data = rsc_action_digest_cmp(rsc, rsc_op, node, data_set);
switch (digest_data->rc) {
case RSC_DIGEST_UNKNOWN:
crm_trace("Resource %s history entry %s on %s has no digest to compare",
rsc->id, ID(rsc_op), node->details->id);
break;
case RSC_DIGEST_MATCH:
break;
default:
reason = "resource parameters have changed";
break;
}
break;
}
if (reason) {
pe__clear_failcount(rsc, node, reason, data_set);
}
}
static void
check_actions_for(xmlNode * rsc_entry, pe_resource_t * rsc, pe_node_t * node, pe_working_set_t * data_set)
{
GListPtr gIter = NULL;
int offset = -1;
int stop_index = 0;
int start_index = 0;
const char *task = NULL;
xmlNode *rsc_op = NULL;
GListPtr op_list = NULL;
GListPtr sorted_op_list = NULL;
CRM_CHECK(node != NULL, return);
if (is_set(rsc->flags, pe_rsc_orphan)) {
pe_resource_t *parent = uber_parent(rsc);
if(parent == NULL
|| pe_rsc_is_clone(parent) == FALSE
|| is_set(parent->flags, pe_rsc_unique)) {
pe_rsc_trace(rsc, "Skipping param check for %s and deleting: orphan", rsc->id);
DeleteRsc(rsc, node, FALSE, data_set);
} else {
pe_rsc_trace(rsc, "Skipping param check for %s (orphan clone)", rsc->id);
}
return;
} else if (pe_find_node_id(rsc->running_on, node->details->id) == NULL) {
if (check_rsc_parameters(rsc, node, rsc_entry, FALSE, data_set)) {
DeleteRsc(rsc, node, FALSE, data_set);
}
pe_rsc_trace(rsc, "Skipping param check for %s: no longer active on %s",
rsc->id, node->details->uname);
return;
}
pe_rsc_trace(rsc, "Processing %s on %s", rsc->id, node->details->uname);
if (check_rsc_parameters(rsc, node, rsc_entry, TRUE, data_set)) {
DeleteRsc(rsc, node, FALSE, data_set);
}
for (rsc_op = __xml_first_child_element(rsc_entry); rsc_op != NULL;
rsc_op = __xml_next_element(rsc_op)) {
if (crm_str_eq((const char *)rsc_op->name, XML_LRM_TAG_RSC_OP, TRUE)) {
op_list = g_list_prepend(op_list, rsc_op);
}
}
sorted_op_list = g_list_sort(op_list, sort_op_by_callid);
calculate_active_ops(sorted_op_list, &start_index, &stop_index);
for (gIter = sorted_op_list; gIter != NULL; gIter = gIter->next) {
xmlNode *rsc_op = (xmlNode *) gIter->data;
guint interval_ms = 0;
offset++;
if (start_index < stop_index) {
/* stopped */
continue;
} else if (offset < start_index) {
/* action occurred prior to a start */
continue;
}
task = crm_element_value(rsc_op, XML_LRM_ATTR_TASK);
crm_element_value_ms(rsc_op, XML_LRM_ATTR_INTERVAL_MS, &interval_ms);
if ((interval_ms > 0) &&
(is_set(rsc->flags, pe_rsc_maintenance) || node->details->maintenance)) {
// Maintenance mode cancels recurring operations
CancelXmlOp(rsc, rsc_op, node, "maintenance mode", data_set);
} else if ((interval_ms > 0)
|| safe_str_eq(task, RSC_STATUS)
|| safe_str_eq(task, RSC_START)
|| safe_str_eq(task, RSC_PROMOTE)
|| safe_str_eq(task, RSC_MIGRATED)) {
/* If a resource operation failed, and the operation's definition
* has changed, clear any fail count so they can be retried fresh.
*/
if (pe__bundle_needs_remote_name(rsc)) {
/* We haven't allocated resources to nodes yet, so if the
* REMOTE_CONTAINER_HACK is used, we may calculate the digest
* based on the literal "#uname" value rather than the properly
* substituted value. That would mistakenly make the action
* definition appear to have been changed. Defer the check until
* later in this case.
*/
pe__add_param_check(rsc_op, rsc, node, pe_check_active,
data_set);
} else if (check_action_definition(rsc, node, rsc_op, data_set)
&& pe_get_failcount(node, rsc, NULL, pe_fc_effective, NULL,
data_set)) {
pe__clear_failcount(rsc, node, "action definition changed",
data_set);
}
}
}
g_list_free(sorted_op_list);
}
static GListPtr
find_rsc_list(GListPtr result, pe_resource_t * rsc, const char *id, gboolean renamed_clones,
gboolean partial, pe_working_set_t * data_set)
{
GListPtr gIter = NULL;
gboolean match = FALSE;
if (id == NULL) {
return NULL;
}
if (rsc == NULL) {
if (data_set == NULL) {
return NULL;
}
for (gIter = data_set->resources; gIter != NULL; gIter = gIter->next) {
pe_resource_t *child = (pe_resource_t *) gIter->data;
result = find_rsc_list(result, child, id, renamed_clones, partial,
NULL);
}
return result;
}
if (partial) {
if (strstr(rsc->id, id)) {
match = TRUE;
} else if (renamed_clones && rsc->clone_name && strstr(rsc->clone_name, id)) {
match = TRUE;
}
} else {
if (strcmp(rsc->id, id) == 0) {
match = TRUE;
} else if (renamed_clones && rsc->clone_name && strcmp(rsc->clone_name, id) == 0) {
match = TRUE;
}
}
if (match) {
result = g_list_prepend(result, rsc);
}
if (rsc->children) {
gIter = rsc->children;
for (; gIter != NULL; gIter = gIter->next) {
pe_resource_t *child = (pe_resource_t *) gIter->data;
result = find_rsc_list(result, child, id, renamed_clones, partial, NULL);
}
}
return result;
}
static void
check_actions(pe_working_set_t * data_set)
{
const char *id = NULL;
pe_node_t *node = NULL;
xmlNode *lrm_rscs = NULL;
xmlNode *status = get_object_root(XML_CIB_TAG_STATUS, data_set->input);
xmlNode *node_state = NULL;
for (node_state = __xml_first_child_element(status); node_state != NULL;
node_state = __xml_next_element(node_state)) {
if (crm_str_eq((const char *)node_state->name, XML_CIB_TAG_STATE, TRUE)) {
id = crm_element_value(node_state, XML_ATTR_ID);
lrm_rscs = find_xml_node(node_state, XML_CIB_TAG_LRM, FALSE);
lrm_rscs = find_xml_node(lrm_rscs, XML_LRM_TAG_RESOURCES, FALSE);
node = pe_find_node_id(data_set->nodes, id);
if (node == NULL) {
continue;
/* Still need to check actions for a maintenance node to cancel existing monitor operations */
} else if (can_run_resources(node) == FALSE && node->details->maintenance == FALSE) {
crm_trace("Skipping param check for %s: can't run resources",
node->details->uname);
continue;
}
crm_trace("Processing node %s", node->details->uname);
if (node->details->online || is_set(data_set->flags, pe_flag_stonith_enabled)) {
xmlNode *rsc_entry = NULL;
for (rsc_entry = __xml_first_child_element(lrm_rscs);
rsc_entry != NULL;
rsc_entry = __xml_next_element(rsc_entry)) {
if (crm_str_eq((const char *)rsc_entry->name, XML_LRM_TAG_RESOURCE, TRUE)) {
if (xml_has_children(rsc_entry)) {
GListPtr gIter = NULL;
GListPtr result = NULL;
const char *rsc_id = ID(rsc_entry);
CRM_CHECK(rsc_id != NULL, return);
result = find_rsc_list(NULL, NULL, rsc_id, TRUE, FALSE, data_set);
for (gIter = result; gIter != NULL; gIter = gIter->next) {
pe_resource_t *rsc = (pe_resource_t *) gIter->data;
if (rsc->variant != pe_native) {
continue;
}
check_actions_for(rsc_entry, rsc, node, data_set);
}
g_list_free(result);
}
}
}
}
}
}
}
static void
apply_placement_constraints(pe_working_set_t * data_set)
{
for (GList *gIter = data_set->placement_constraints;
gIter != NULL; gIter = gIter->next) {
pe__location_t *cons = gIter->data;
cons->rsc_lh->cmds->rsc_location(cons->rsc_lh, cons);
}
}
static gboolean
failcount_clear_action_exists(pe_node_t * node, pe_resource_t * rsc)
{
gboolean rc = FALSE;
GList *list = pe__resource_actions(rsc, node, CRM_OP_CLEAR_FAILCOUNT, TRUE);
if (list) {
rc = TRUE;
}
g_list_free(list);
return rc;
}
/*!
* \internal
* \brief Force resource away if failures hit migration threshold
*
* \param[in,out] rsc Resource to check for failures
* \param[in,out] node Node to check for failures
* \param[in,out] data_set Cluster working set to update
*/
static void
check_migration_threshold(pe_resource_t *rsc, pe_node_t *node,
pe_working_set_t *data_set)
{
int fail_count, countdown;
pe_resource_t *failed;
/* Migration threshold of 0 means never force away */
if (rsc->migration_threshold == 0) {
return;
}
// If we're ignoring failures, also ignore the migration threshold
if (is_set(rsc->flags, pe_rsc_failure_ignored)) {
return;
}
/* If there are no failures, there's no need to force away */
fail_count = pe_get_failcount(node, rsc, NULL,
pe_fc_effective|pe_fc_fillers, NULL,
data_set);
if (fail_count <= 0) {
return;
}
/* How many more times recovery will be tried on this node */
countdown = QB_MAX(rsc->migration_threshold - fail_count, 0);
/* If failed resource has a parent, we'll force the parent away */
failed = rsc;
if (is_not_set(rsc->flags, pe_rsc_unique)) {
failed = uber_parent(rsc);
}
if (countdown == 0) {
resource_location(failed, node, -INFINITY, "__fail_limit__", data_set);
crm_warn("Forcing %s away from %s after %d failures (max=%d)",
failed->id, node->details->uname, fail_count,
rsc->migration_threshold);
} else {
crm_info("%s can fail %d more times on %s before being forced off",
failed->id, countdown, node->details->uname);
}
}
static void
common_apply_stickiness(pe_resource_t * rsc, pe_node_t * node, pe_working_set_t * data_set)
{
if (rsc->children) {
GListPtr gIter = rsc->children;
for (; gIter != NULL; gIter = gIter->next) {
pe_resource_t *child_rsc = (pe_resource_t *) gIter->data;
common_apply_stickiness(child_rsc, node, data_set);
}
return;
}
if (is_set(rsc->flags, pe_rsc_managed)
&& rsc->stickiness != 0 && pcmk__list_of_1(rsc->running_on)) {
pe_node_t *current = pe_find_node_id(rsc->running_on, node->details->id);
pe_node_t *match = pe_hash_table_lookup(rsc->allowed_nodes, node->details->id);
if (current == NULL) {
} else if (match != NULL || is_set(data_set->flags, pe_flag_symmetric_cluster)) {
pe_resource_t *sticky_rsc = rsc;
resource_location(sticky_rsc, node, rsc->stickiness, "stickiness", data_set);
pe_rsc_debug(sticky_rsc, "Resource %s: preferring current location"
" (node=%s, weight=%d)", sticky_rsc->id,
node->details->uname, rsc->stickiness);
} else {
GHashTableIter iter;
pe_node_t *nIter = NULL;
pe_rsc_debug(rsc, "Ignoring stickiness for %s: the cluster is asymmetric"
" and node %s is not explicitly allowed", rsc->id, node->details->uname);
g_hash_table_iter_init(&iter, rsc->allowed_nodes);
while (g_hash_table_iter_next(&iter, NULL, (void **)&nIter)) {
crm_err("%s[%s] = %d", rsc->id, nIter->details->uname, nIter->weight);
}
}
}
/* Check the migration threshold only if a failcount clear action
* has not already been placed for this resource on the node.
* There is no sense in potentially forcing the resource from this
* node if the failcount is being reset anyway.
*
* @TODO A clear_failcount operation can be scheduled in stage4() via
* check_actions_for(), or in stage5() via check_params(). This runs in
* stage2(), so it cannot detect those, meaning we might check the migration
* threshold when we shouldn't -- worst case, we stop or move the resource,
* then move it back next transition.
*/
if (failcount_clear_action_exists(node, rsc) == FALSE) {
check_migration_threshold(rsc, node, data_set);
}
}
void
complex_set_cmds(pe_resource_t * rsc)
{
GListPtr gIter = rsc->children;
rsc->cmds = &resource_class_alloc_functions[rsc->variant];
for (; gIter != NULL; gIter = gIter->next) {
pe_resource_t *child_rsc = (pe_resource_t *) gIter->data;
complex_set_cmds(child_rsc);
}
}
void
set_alloc_actions(pe_working_set_t * data_set)
{
GListPtr gIter = data_set->resources;
for (; gIter != NULL; gIter = gIter->next) {
pe_resource_t *rsc = (pe_resource_t *) gIter->data;
complex_set_cmds(rsc);
}
}
static void
calculate_system_health(gpointer gKey, gpointer gValue, gpointer user_data)
{
const char *key = (const char *)gKey;
const char *value = (const char *)gValue;
int *system_health = (int *)user_data;
if (!gKey || !gValue || !user_data) {
return;
}
if (pcmk__starts_with(key, "#health")) {
int score;
/* Convert the value into an integer */
score = char2score(value);
/* Add it to the running total */
*system_health = merge_weights(score, *system_health);
}
}
static gboolean
apply_system_health(pe_working_set_t * data_set)
{
GListPtr gIter = NULL;
const char *health_strategy = pe_pref(data_set->config_hash, "node-health-strategy");
int base_health = 0;
if (health_strategy == NULL || safe_str_eq(health_strategy, "none")) {
/* Prevent any accidental health -> score translation */
node_score_red = 0;
node_score_yellow = 0;
node_score_green = 0;
return TRUE;
} else if (safe_str_eq(health_strategy, "migrate-on-red")) {
/* Resources on nodes which have health values of red are
* weighted away from that node.
*/
node_score_red = -INFINITY;
node_score_yellow = 0;
node_score_green = 0;
} else if (safe_str_eq(health_strategy, "only-green")) {
/* Resources on nodes which have health values of red or yellow
* are forced away from that node.
*/
node_score_red = -INFINITY;
node_score_yellow = -INFINITY;
node_score_green = 0;
} else if (safe_str_eq(health_strategy, "progressive")) {
/* Same as the above, but use the r/y/g scores provided by the user
* Defaults are provided by the pe_prefs table
* Also, custom health "base score" can be used
*/
base_health = crm_parse_int(pe_pref(data_set->config_hash, "node-health-base"), "0");
} else if (safe_str_eq(health_strategy, "custom")) {
/* Requires the admin to configure the rsc_location constaints for
* processing the stored health scores
*/
/* TODO: Check for the existence of appropriate node health constraints */
return TRUE;
} else {
crm_err("Unknown node health strategy: %s", health_strategy);
return FALSE;
}
crm_info("Applying automated node health strategy: %s", health_strategy);
for (gIter = data_set->nodes; gIter != NULL; gIter = gIter->next) {
int system_health = base_health;
pe_node_t *node = (pe_node_t *) gIter->data;
/* Search through the node hash table for system health entries. */
g_hash_table_foreach(node->details->attrs, calculate_system_health, &system_health);
crm_info(" Node %s has an combined system health of %d",
node->details->uname, system_health);
/* If the health is non-zero, then create a new rsc2node so that the
* weight will be added later on.
*/
if (system_health != 0) {
GListPtr gIter2 = data_set->resources;
for (; gIter2 != NULL; gIter2 = gIter2->next) {
pe_resource_t *rsc = (pe_resource_t *) gIter2->data;
rsc2node_new(health_strategy, rsc, system_health, NULL, node, data_set);
}
}
}
return TRUE;
}
gboolean
stage0(pe_working_set_t * data_set)
{
xmlNode *cib_constraints = get_object_root(XML_CIB_TAG_CONSTRAINTS, data_set->input);
if (data_set->input == NULL) {
return FALSE;
}
if (is_set(data_set->flags, pe_flag_have_status) == FALSE) {
crm_trace("Calculating status");
cluster_status(data_set);
}
set_alloc_actions(data_set);
apply_system_health(data_set);
unpack_constraints(cib_constraints, data_set);
return TRUE;
}
/*
* Check nodes for resources started outside of the LRM
*/
gboolean
probe_resources(pe_working_set_t * data_set)
{
pe_action_t *probe_node_complete = NULL;
for (GListPtr gIter = data_set->nodes; gIter != NULL; gIter = gIter->next) {
pe_node_t *node = (pe_node_t *) gIter->data;
const char *probed = pe_node_attribute_raw(node, CRM_OP_PROBED);
if (node->details->online == FALSE) {
if (pe__is_remote_node(node) && node->details->remote_rsc
&& (get_remote_node_state(node) == remote_state_failed)) {
- pe_fence_node(data_set, node, "the connection is unrecoverable");
+ pe_fence_node(data_set, node, "the connection is unrecoverable", FALSE);
}
continue;
} else if (node->details->unclean) {
continue;
} else if (node->details->rsc_discovery_enabled == FALSE) {
/* resource discovery is disabled for this node */
continue;
}
if (probed != NULL && crm_is_true(probed) == FALSE) {
pe_action_t *probe_op = custom_action(NULL, crm_strdup_printf("%s-%s", CRM_OP_REPROBE, node->details->uname),
CRM_OP_REPROBE, node, FALSE, TRUE, data_set);
add_hash_param(probe_op->meta, XML_ATTR_TE_NOWAIT, XML_BOOLEAN_TRUE);
continue;
}
for (GListPtr gIter2 = data_set->resources; gIter2 != NULL; gIter2 = gIter2->next) {
pe_resource_t *rsc = (pe_resource_t *) gIter2->data;
rsc->cmds->create_probe(rsc, node, probe_node_complete, FALSE, data_set);
}
}
return TRUE;
}
static void
rsc_discover_filter(pe_resource_t *rsc, pe_node_t *node)
{
GListPtr gIter = rsc->children;
pe_resource_t *top = uber_parent(rsc);
pe_node_t *match;
if (rsc->exclusive_discover == FALSE && top->exclusive_discover == FALSE) {
return;
}
for (; gIter != NULL; gIter = gIter->next) {
pe_resource_t *child_rsc = (pe_resource_t *) gIter->data;
rsc_discover_filter(child_rsc, node);
}
match = g_hash_table_lookup(rsc->allowed_nodes, node->details->id);
if (match && match->rsc_discover_mode != pe_discover_exclusive) {
match->weight = -INFINITY;
}
}
static time_t
shutdown_time(pe_node_t *node, pe_working_set_t *data_set)
{
const char *shutdown = pe_node_attribute_raw(node, XML_CIB_ATTR_SHUTDOWN);
time_t result = 0;
if (shutdown) {
errno = 0;
result = (time_t) crm_parse_ll(shutdown, NULL);
if (errno != 0) {
result = 0;
}
}
return result? result : get_effective_time(data_set);
}
static void
apply_shutdown_lock(pe_resource_t *rsc, pe_working_set_t *data_set)
{
const char *class;
// Only primitives and (uncloned) groups may be locked
if (rsc->variant == pe_group) {
for (GList *item = rsc->children; item != NULL;
item = item->next) {
apply_shutdown_lock((pe_resource_t *) item->data, data_set);
}
} else if (rsc->variant != pe_native) {
return;
}
// Fence devices and remote connections can't be locked
class = crm_element_value(rsc->xml, XML_AGENT_ATTR_CLASS);
if ((class == NULL) || !strcmp(class, PCMK_RESOURCE_CLASS_STONITH)
|| pe__resource_is_remote_conn(rsc, data_set)) {
return;
}
if (rsc->lock_node != NULL) {
// The lock was obtained from resource history
if (rsc->running_on != NULL) {
/* The resource was started elsewhere even though it is now
* considered locked. This shouldn't be possible, but as a
* failsafe, we don't want to disturb the resource now.
*/
pe_rsc_info(rsc,
"Cancelling shutdown lock because %s is already active",
rsc->id);
pe__clear_resource_history(rsc, rsc->lock_node, data_set);
rsc->lock_node = NULL;
rsc->lock_time = 0;
}
// Only a resource active on exactly one node can be locked
} else if (pcmk__list_of_1(rsc->running_on)) {
pe_node_t *node = rsc->running_on->data;
if (node->details->shutdown) {
if (node->details->unclean) {
pe_rsc_debug(rsc, "Not locking %s to unclean %s for shutdown",
rsc->id, node->details->uname);
} else {
rsc->lock_node = node;
rsc->lock_time = shutdown_time(node, data_set);
}
}
}
if (rsc->lock_node == NULL) {
// No lock needed
return;
}
if (data_set->shutdown_lock > 0) {
time_t lock_expiration = rsc->lock_time + data_set->shutdown_lock;
pe_rsc_info(rsc, "Locking %s to %s due to shutdown (expires @%lld)",
rsc->id, rsc->lock_node->details->uname,
(long long) lock_expiration);
pe__update_recheck_time(++lock_expiration, data_set);
} else {
pe_rsc_info(rsc, "Locking %s to %s due to shutdown",
rsc->id, rsc->lock_node->details->uname);
}
// If resource is locked to one node, ban it from all other nodes
for (GList *item = data_set->nodes; item != NULL; item = item->next) {
pe_node_t *node = item->data;
if (strcmp(node->details->uname, rsc->lock_node->details->uname)) {
resource_location(rsc, node, -CRM_SCORE_INFINITY,
XML_CONFIG_ATTR_SHUTDOWN_LOCK, data_set);
}
}
}
/*
* \internal
* \brief Stage 2 of cluster status: apply node-specific criteria
*
* Count known nodes, and apply location constraints, stickiness, and exclusive
* resource discovery.
*/
gboolean
stage2(pe_working_set_t * data_set)
{
GListPtr gIter = NULL;
if (is_set(data_set->flags, pe_flag_shutdown_lock)) {
for (gIter = data_set->resources; gIter != NULL; gIter = gIter->next) {
apply_shutdown_lock((pe_resource_t *) gIter->data, data_set);
}
}
if (is_not_set(data_set->flags, pe_flag_no_compat)) {
// @COMPAT API backward compatibility
for (gIter = data_set->nodes; gIter != NULL; gIter = gIter->next) {
pe_node_t *node = (pe_node_t *) gIter->data;
if (node && (node->weight >= 0) && node->details->online
&& (node->details->type != node_ping)) {
data_set->max_valid_nodes++;
}
}
}
crm_trace("Applying placement constraints");
apply_placement_constraints(data_set);
gIter = data_set->nodes;
for (; gIter != NULL; gIter = gIter->next) {
GListPtr gIter2 = NULL;
pe_node_t *node = (pe_node_t *) gIter->data;
gIter2 = data_set->resources;
for (; gIter2 != NULL; gIter2 = gIter2->next) {
pe_resource_t *rsc = (pe_resource_t *) gIter2->data;
common_apply_stickiness(rsc, node, data_set);
rsc_discover_filter(rsc, node);
}
}
return TRUE;
}
/*
* Create internal resource constraints before allocation
*/
gboolean
stage3(pe_working_set_t * data_set)
{
GListPtr gIter = data_set->resources;
for (; gIter != NULL; gIter = gIter->next) {
pe_resource_t *rsc = (pe_resource_t *) gIter->data;
rsc->cmds->internal_constraints(rsc, data_set);
}
return TRUE;
}
/*
* Check for orphaned or redefined actions
*/
gboolean
stage4(pe_working_set_t * data_set)
{
check_actions(data_set);
return TRUE;
}
static void *
convert_const_pointer(const void *ptr)
{
/* Worst function ever */
return (void *)ptr;
}
static gint
sort_rsc_process_order(gconstpointer a, gconstpointer b, gpointer data)
{
int rc = 0;
int r1_weight = -INFINITY;
int r2_weight = -INFINITY;
const char *reason = "existence";
const GListPtr nodes = (GListPtr) data;
const pe_resource_t *resource1 = a;
const pe_resource_t *resource2 = b;
pe_node_t *r1_node = NULL;
pe_node_t *r2_node = NULL;
GListPtr gIter = NULL;
GHashTable *r1_nodes = NULL;
GHashTable *r2_nodes = NULL;
if (a == NULL && b == NULL) {
goto done;
}
if (a == NULL) {
return 1;
}
if (b == NULL) {
return -1;
}
reason = "priority";
r1_weight = resource1->priority;
r2_weight = resource2->priority;
if (r1_weight > r2_weight) {
rc = -1;
goto done;
}
if (r1_weight < r2_weight) {
rc = 1;
goto done;
}
reason = "no node list";
if (nodes == NULL) {
goto done;
}
r1_nodes = pcmk__native_merge_weights(convert_const_pointer(resource1),
resource1->id, NULL, NULL, 1,
pe_weights_forward | pe_weights_init);
pe__show_node_weights(true, NULL, resource1->id, r1_nodes);
r2_nodes = pcmk__native_merge_weights(convert_const_pointer(resource2),
resource2->id, NULL, NULL, 1,
pe_weights_forward | pe_weights_init);
pe__show_node_weights(true, NULL, resource2->id, r2_nodes);
/* Current location score */
reason = "current location";
r1_weight = -INFINITY;
r2_weight = -INFINITY;
if (resource1->running_on) {
r1_node = pe__current_node(resource1);
r1_node = g_hash_table_lookup(r1_nodes, r1_node->details->id);
if (r1_node != NULL) {
r1_weight = r1_node->weight;
}
}
if (resource2->running_on) {
r2_node = pe__current_node(resource2);
r2_node = g_hash_table_lookup(r2_nodes, r2_node->details->id);
if (r2_node != NULL) {
r2_weight = r2_node->weight;
}
}
if (r1_weight > r2_weight) {
rc = -1;
goto done;
}
if (r1_weight < r2_weight) {
rc = 1;
goto done;
}
reason = "score";
for (gIter = nodes; gIter != NULL; gIter = gIter->next) {
pe_node_t *node = (pe_node_t *) gIter->data;
r1_node = NULL;
r2_node = NULL;
r1_weight = -INFINITY;
if (r1_nodes) {
r1_node = g_hash_table_lookup(r1_nodes, node->details->id);
}
if (r1_node) {
r1_weight = r1_node->weight;
}
r2_weight = -INFINITY;
if (r2_nodes) {
r2_node = g_hash_table_lookup(r2_nodes, node->details->id);
}
if (r2_node) {
r2_weight = r2_node->weight;
}
if (r1_weight > r2_weight) {
rc = -1;
goto done;
}
if (r1_weight < r2_weight) {
rc = 1;
goto done;
}
}
done:
crm_trace("%s (%d) on %s %c %s (%d) on %s: %s",
resource1->id, r1_weight, r1_node ? r1_node->details->id : "n/a",
rc < 0 ? '>' : rc > 0 ? '<' : '=',
resource2->id, r2_weight, r2_node ? r2_node->details->id : "n/a", reason);
if (r1_nodes) {
g_hash_table_destroy(r1_nodes);
}
if (r2_nodes) {
g_hash_table_destroy(r2_nodes);
}
return rc;
}
static void
allocate_resources(pe_working_set_t * data_set)
{
GListPtr gIter = NULL;
if (is_set(data_set->flags, pe_flag_have_remote_nodes)) {
/* Allocate remote connection resources first (which will also allocate
* any colocation dependencies). If the connection is migrating, always
* prefer the partial migration target.
*/
for (gIter = data_set->resources; gIter != NULL; gIter = gIter->next) {
pe_resource_t *rsc = (pe_resource_t *) gIter->data;
if (rsc->is_remote_node == FALSE) {
continue;
}
pe_rsc_trace(rsc, "Allocating remote connection resource '%s'",
rsc->id);
rsc->cmds->allocate(rsc, rsc->partial_migration_target, data_set);
}
}
/* now do the rest of the resources */
for (gIter = data_set->resources; gIter != NULL; gIter = gIter->next) {
pe_resource_t *rsc = (pe_resource_t *) gIter->data;
if (rsc->is_remote_node == TRUE) {
continue;
}
pe_rsc_trace(rsc, "Allocating resource '%s'", rsc->id);
rsc->cmds->allocate(rsc, NULL, data_set);
}
}
/* We always use pe_order_preserve with these convenience functions to exempt
* internally generated constraints from the prohibition of user constraints
* involving remote connection resources.
*
* The start ordering additionally uses pe_order_runnable_left so that the
* specified action is not runnable if the start is not runnable.
*/
static inline void
order_start_then_action(pe_resource_t *lh_rsc, pe_action_t *rh_action,
enum pe_ordering extra, pe_working_set_t *data_set)
{
if (lh_rsc && rh_action && data_set) {
custom_action_order(lh_rsc, start_key(lh_rsc), NULL,
rh_action->rsc, NULL, rh_action,
pe_order_preserve | pe_order_runnable_left | extra,
data_set);
}
}
static inline void
order_action_then_stop(pe_action_t *lh_action, pe_resource_t *rh_rsc,
enum pe_ordering extra, pe_working_set_t *data_set)
{
if (lh_action && rh_rsc && data_set) {
custom_action_order(lh_action->rsc, NULL, lh_action,
rh_rsc, stop_key(rh_rsc), NULL,
pe_order_preserve | extra, data_set);
}
}
// Clear fail counts for orphaned rsc on all online nodes
static void
cleanup_orphans(pe_resource_t * rsc, pe_working_set_t * data_set)
{
GListPtr gIter = NULL;
for (gIter = data_set->nodes; gIter != NULL; gIter = gIter->next) {
pe_node_t *node = (pe_node_t *) gIter->data;
if (node->details->online
&& pe_get_failcount(node, rsc, NULL, pe_fc_effective, NULL,
data_set)) {
pe_action_t *clear_op = NULL;
clear_op = pe__clear_failcount(rsc, node, "it is orphaned",
data_set);
/* We can't use order_action_then_stop() here because its
* pe_order_preserve breaks things
*/
custom_action_order(clear_op->rsc, NULL, clear_op,
rsc, stop_key(rsc), NULL,
pe_order_optional, data_set);
}
}
}
gboolean
stage5(pe_working_set_t * data_set)
{
GListPtr gIter = NULL;
int log_prio = show_utilization? LOG_STDOUT : utilization_log_level;
if (safe_str_neq(data_set->placement_strategy, "default")) {
GListPtr nodes = g_list_copy(data_set->nodes);
nodes = sort_nodes_by_weight(nodes, NULL, data_set);
data_set->resources =
g_list_sort_with_data(data_set->resources, sort_rsc_process_order, nodes);
g_list_free(nodes);
}
gIter = data_set->nodes;
for (; gIter != NULL; gIter = gIter->next) {
pe_node_t *node = (pe_node_t *) gIter->data;
dump_node_capacity(log_prio, "Original", node);
}
crm_trace("Allocating services");
/* Take (next) highest resource, assign it and create its actions */
allocate_resources(data_set);
gIter = data_set->nodes;
for (; gIter != NULL; gIter = gIter->next) {
pe_node_t *node = (pe_node_t *) gIter->data;
dump_node_capacity(log_prio, "Remaining", node);
}
// Process deferred action checks
pe__foreach_param_check(data_set, check_params);
pe__free_param_checks(data_set);
if (is_set(data_set->flags, pe_flag_startup_probes)) {
crm_trace("Calculating needed probes");
/* This code probably needs optimization
* ptest -x with 100 nodes, 100 clones and clone-max=100:
With probes:
ptest[14781]: 2010/09/27_17:56:46 notice: TRACE: do_calculations: pengine.c:258 Calculate cluster status
ptest[14781]: 2010/09/27_17:56:46 notice: TRACE: do_calculations: pengine.c:278 Applying placement constraints
ptest[14781]: 2010/09/27_17:56:47 notice: TRACE: do_calculations: pengine.c:285 Create internal constraints
ptest[14781]: 2010/09/27_17:56:47 notice: TRACE: do_calculations: pengine.c:292 Check actions
ptest[14781]: 2010/09/27_17:56:48 notice: TRACE: do_calculations: pengine.c:299 Allocate resources
ptest[14781]: 2010/09/27_17:56:48 notice: TRACE: stage5: allocate.c:881 Allocating services
ptest[14781]: 2010/09/27_17:56:49 notice: TRACE: stage5: allocate.c:894 Calculating needed probes
ptest[14781]: 2010/09/27_17:56:51 notice: TRACE: stage5: allocate.c:899 Creating actions
ptest[14781]: 2010/09/27_17:56:52 notice: TRACE: stage5: allocate.c:905 Creating done
ptest[14781]: 2010/09/27_17:56:52 notice: TRACE: do_calculations: pengine.c:306 Processing fencing and shutdown cases
ptest[14781]: 2010/09/27_17:56:52 notice: TRACE: do_calculations: pengine.c:313 Applying ordering constraints
36s
ptest[14781]: 2010/09/27_17:57:28 notice: TRACE: do_calculations: pengine.c:320 Create transition graph
Without probes:
ptest[14637]: 2010/09/27_17:56:21 notice: TRACE: do_calculations: pengine.c:258 Calculate cluster status
ptest[14637]: 2010/09/27_17:56:22 notice: TRACE: do_calculations: pengine.c:278 Applying placement constraints
ptest[14637]: 2010/09/27_17:56:22 notice: TRACE: do_calculations: pengine.c:285 Create internal constraints
ptest[14637]: 2010/09/27_17:56:22 notice: TRACE: do_calculations: pengine.c:292 Check actions
ptest[14637]: 2010/09/27_17:56:23 notice: TRACE: do_calculations: pengine.c:299 Allocate resources
ptest[14637]: 2010/09/27_17:56:23 notice: TRACE: stage5: allocate.c:881 Allocating services
ptest[14637]: 2010/09/27_17:56:24 notice: TRACE: stage5: allocate.c:899 Creating actions
ptest[14637]: 2010/09/27_17:56:25 notice: TRACE: stage5: allocate.c:905 Creating done
ptest[14637]: 2010/09/27_17:56:25 notice: TRACE: do_calculations: pengine.c:306 Processing fencing and shutdown cases
ptest[14637]: 2010/09/27_17:56:25 notice: TRACE: do_calculations: pengine.c:313 Applying ordering constraints
ptest[14637]: 2010/09/27_17:56:25 notice: TRACE: do_calculations: pengine.c:320 Create transition graph
*/
probe_resources(data_set);
}
crm_trace("Handle orphans");
if (is_set(data_set->flags, pe_flag_stop_rsc_orphans)) {
for (gIter = data_set->resources; gIter != NULL; gIter = gIter->next) {
pe_resource_t *rsc = (pe_resource_t *) gIter->data;
/* There's no need to recurse into rsc->children because those
* should just be unallocated clone instances.
*/
if (is_set(rsc->flags, pe_rsc_orphan)) {
cleanup_orphans(rsc, data_set);
}
}
}
crm_trace("Creating actions");
for (gIter = data_set->resources; gIter != NULL; gIter = gIter->next) {
pe_resource_t *rsc = (pe_resource_t *) gIter->data;
rsc->cmds->create_actions(rsc, data_set);
}
crm_trace("Creating done");
return TRUE;
}
static gboolean
is_managed(const pe_resource_t * rsc)
{
GListPtr gIter = rsc->children;
if (is_set(rsc->flags, pe_rsc_managed)) {
return TRUE;
}
for (; gIter != NULL; gIter = gIter->next) {
pe_resource_t *child_rsc = (pe_resource_t *) gIter->data;
if (is_managed(child_rsc)) {
return TRUE;
}
}
return FALSE;
}
static gboolean
any_managed_resources(pe_working_set_t * data_set)
{
GListPtr gIter = data_set->resources;
for (; gIter != NULL; gIter = gIter->next) {
pe_resource_t *rsc = (pe_resource_t *) gIter->data;
if (is_managed(rsc)) {
return TRUE;
}
}
return FALSE;
}
/*!
* \internal
* \brief Create pseudo-op for guest node fence, and order relative to it
*
* \param[in] node Guest node to fence
* \param[in] data_set Working set of CIB state
*/
static void
fence_guest(pe_node_t *node, pe_working_set_t *data_set)
{
pe_resource_t *container = node->details->remote_rsc->container;
pe_action_t *stop = NULL;
pe_action_t *stonith_op = NULL;
/* The fence action is just a label; we don't do anything differently for
* off vs. reboot. We specify it explicitly, rather than let it default to
* cluster's default action, because we are not _initiating_ fencing -- we
* are creating a pseudo-event to describe fencing that is already occurring
* by other means (container recovery).
*/
const char *fence_action = "off";
/* Check whether guest's container resource has any explicit stop or
* start (the stop may be implied by fencing of the guest's host).
*/
if (container) {
stop = find_first_action(container->actions, NULL, CRMD_ACTION_STOP, NULL);
if (find_first_action(container->actions, NULL, CRMD_ACTION_START, NULL)) {
fence_action = "reboot";
}
}
/* Create a fence pseudo-event, so we have an event to order actions
* against, and the controller can always detect it.
*/
- stonith_op = pe_fence_op(node, fence_action, FALSE, "guest is unclean", data_set);
+ stonith_op = pe_fence_op(node, fence_action, FALSE, "guest is unclean", FALSE, data_set);
update_action_flags(stonith_op, pe_action_pseudo | pe_action_runnable,
__FUNCTION__, __LINE__);
/* We want to imply stops/demotes after the guest is stopped, not wait until
* it is restarted, so we always order pseudo-fencing after stop, not start
* (even though start might be closer to what is done for a real reboot).
*/
if(stop && is_set(stop->flags, pe_action_pseudo)) {
- pe_action_t *parent_stonith_op = pe_fence_op(stop->node, NULL, FALSE, NULL, data_set);
+ pe_action_t *parent_stonith_op = pe_fence_op(stop->node, NULL, FALSE, NULL, FALSE, data_set);
crm_info("Implying guest node %s is down (action %d) after %s fencing",
node->details->uname, stonith_op->id, stop->node->details->uname);
order_actions(parent_stonith_op, stonith_op,
pe_order_runnable_left|pe_order_implies_then);
} else if (stop) {
order_actions(stop, stonith_op,
pe_order_runnable_left|pe_order_implies_then);
crm_info("Implying guest node %s is down (action %d) "
"after container %s is stopped (action %d)",
node->details->uname, stonith_op->id,
container->id, stop->id);
} else {
/* If we're fencing the guest node but there's no stop for the guest
* resource, we must think the guest is already stopped. However, we may
* think so because its resource history was just cleaned. To avoid
* unnecessarily considering the guest node down if it's really up,
* order the pseudo-fencing after any stop of the connection resource,
* which will be ordered after any container (re-)probe.
*/
stop = find_first_action(node->details->remote_rsc->actions, NULL,
RSC_STOP, NULL);
if (stop) {
order_actions(stop, stonith_op, pe_order_optional);
crm_info("Implying guest node %s is down (action %d) "
"after connection is stopped (action %d)",
node->details->uname, stonith_op->id, stop->id);
} else {
/* Not sure why we're fencing, but everything must already be
* cleanly stopped.
*/
crm_info("Implying guest node %s is down (action %d) ",
node->details->uname, stonith_op->id);
}
}
/* Order/imply other actions relative to pseudo-fence as with real fence */
pcmk__order_vs_fence(stonith_op, data_set);
}
/*
* Create dependencies for stonith and shutdown operations
*/
gboolean
stage6(pe_working_set_t * data_set)
{
pe_action_t *dc_down = NULL;
pe_action_t *stonith_op = NULL;
gboolean integrity_lost = FALSE;
gboolean need_stonith = TRUE;
GListPtr gIter;
GListPtr stonith_ops = NULL;
GList *shutdown_ops = NULL;
/* Remote ordering constraints need to happen prior to calculating fencing
* because it is one more place we will mark the node as dirty.
*
* A nice side effect of doing them early is that apply_*_ordering() can be
* simpler because pe_fence_node() has already done some of the work.
*/
crm_trace("Creating remote ordering constraints");
apply_remote_node_ordering(data_set);
crm_trace("Processing fencing and shutdown cases");
if (any_managed_resources(data_set) == FALSE) {
crm_notice("Delaying fencing operations until there are resources to manage");
need_stonith = FALSE;
}
/* Check each node for stonith/shutdown */
for (gIter = data_set->nodes; gIter != NULL; gIter = gIter->next) {
pe_node_t *node = (pe_node_t *) gIter->data;
/* Guest nodes are "fenced" by recovering their container resource,
* so handle them separately.
*/
if (pe__is_guest_node(node)) {
if (node->details->remote_requires_reset && need_stonith
&& pe_can_fence(data_set, node)) {
fence_guest(node, data_set);
}
continue;
}
stonith_op = NULL;
if (node->details->unclean
&& need_stonith && pe_can_fence(data_set, node)) {
- stonith_op = pe_fence_op(node, NULL, FALSE, "node is unclean", data_set);
+ stonith_op = pe_fence_op(node, NULL, FALSE, "node is unclean", FALSE, data_set);
pe_warn("Scheduling Node %s for STONITH", node->details->uname);
pcmk__order_vs_fence(stonith_op, data_set);
if (node->details->is_dc) {
// Remember if the DC is being fenced
dc_down = stonith_op;
} else {
if (is_not_set(data_set->flags, pe_flag_concurrent_fencing)
&& (stonith_ops != NULL)) {
/* Concurrent fencing is disabled, so order each non-DC
* fencing in a chain. If there is any DC fencing or
* shutdown, it will be ordered after the last action in the
* chain later.
*/
order_actions((pe_action_t *) stonith_ops->data,
stonith_op, pe_order_optional);
}
// Remember all non-DC fencing actions in a separate list
stonith_ops = g_list_prepend(stonith_ops, stonith_op);
}
} else if (node->details->online && node->details->shutdown &&
/* TODO define what a shutdown op means for a remote node.
* For now we do not send shutdown operations for remote nodes, but
* if we can come up with a good use for this in the future, we will. */
pe__is_guest_or_remote_node(node) == FALSE) {
pe_action_t *down_op = sched_shutdown_op(node, data_set);
if (node->details->is_dc) {
// Remember if the DC is being shut down
dc_down = down_op;
} else {
// Remember non-DC shutdowns for later ordering
shutdown_ops = g_list_prepend(shutdown_ops, down_op);
}
}
if (node->details->unclean && stonith_op == NULL) {
integrity_lost = TRUE;
pe_warn("Node %s is unclean!", node->details->uname);
}
}
if (integrity_lost) {
if (is_set(data_set->flags, pe_flag_stonith_enabled) == FALSE) {
pe_warn("YOUR RESOURCES ARE NOW LIKELY COMPROMISED");
pe_err("ENABLE STONITH TO KEEP YOUR RESOURCES SAFE");
} else if (is_set(data_set->flags, pe_flag_have_quorum) == FALSE) {
crm_notice("Cannot fence unclean nodes until quorum is"
" attained (or no-quorum-policy is set to ignore)");
}
}
if (dc_down != NULL) {
/* Order any non-DC shutdowns before any DC shutdown, to avoid repeated
* DC elections. However, we don't want to order non-DC shutdowns before
* a DC *fencing*, because even though we don't want a node that's
* shutting down to become DC, the DC fencing could be ordered before a
* clone stop that's also ordered before the shutdowns, thus leading to
* a graph loop.
*/
if (safe_str_eq(dc_down->task, CRM_OP_SHUTDOWN)) {
for (gIter = shutdown_ops; gIter != NULL; gIter = gIter->next) {
pe_action_t *node_stop = (pe_action_t *) gIter->data;
crm_debug("Ordering shutdown on %s before %s on DC %s",
node_stop->node->details->uname,
dc_down->task, dc_down->node->details->uname);
order_actions(node_stop, dc_down, pe_order_optional);
}
}
// Order any non-DC fencing before any DC fencing or shutdown
if (is_set(data_set->flags, pe_flag_concurrent_fencing)) {
/* With concurrent fencing, order each non-DC fencing action
* separately before any DC fencing or shutdown.
*/
for (gIter = stonith_ops; gIter != NULL; gIter = gIter->next) {
order_actions((pe_action_t *) gIter->data, dc_down,
pe_order_optional);
}
} else if (stonith_ops) {
/* Without concurrent fencing, the non-DC fencing actions are
* already ordered relative to each other, so we just need to order
* the DC fencing after the last action in the chain (which is the
* first item in the list).
*/
order_actions((pe_action_t *) stonith_ops->data, dc_down,
pe_order_optional);
}
}
g_list_free(stonith_ops);
g_list_free(shutdown_ops);
return TRUE;
}
/*
* Determine the sets of independent actions and the correct order for the
* actions in each set.
*
* Mark dependencies of un-runnable actions un-runnable
*
*/
static GListPtr
find_actions_by_task(GListPtr actions, pe_resource_t * rsc, const char *original_key)
{
GListPtr list = NULL;
list = find_actions(actions, original_key, NULL);
if (list == NULL) {
/* we're potentially searching a child of the original resource */
char *key = NULL;
char *task = NULL;
guint interval_ms = 0;
if (parse_op_key(original_key, NULL, &task, &interval_ms)) {
key = pcmk__op_key(rsc->id, task, interval_ms);
list = find_actions(actions, key, NULL);
} else {
crm_err("search key: %s", original_key);
}
free(key);
free(task);
}
return list;
}
static void
rsc_order_then(pe_action_t *lh_action, pe_resource_t *rsc,
pe__ordering_t *order)
{
GListPtr gIter = NULL;
GListPtr rh_actions = NULL;
pe_action_t *rh_action = NULL;
enum pe_ordering type;
CRM_CHECK(rsc != NULL, return);
CRM_CHECK(order != NULL, return);
type = order->type;
rh_action = order->rh_action;
crm_trace("Processing RH of ordering constraint %d", order->id);
if (rh_action != NULL) {
rh_actions = g_list_prepend(NULL, rh_action);
} else if (rsc != NULL) {
rh_actions = find_actions_by_task(rsc->actions, rsc, order->rh_action_task);
}
if (rh_actions == NULL) {
pe_rsc_trace(rsc, "No RH-Side (%s/%s) found for constraint..."
" ignoring", rsc->id, order->rh_action_task);
if (lh_action) {
pe_rsc_trace(rsc, "LH-Side was: %s", lh_action->uuid);
}
return;
}
if (lh_action && lh_action->rsc == rsc && is_set(lh_action->flags, pe_action_dangle)) {
pe_rsc_trace(rsc, "Detected dangling operation %s -> %s", lh_action->uuid,
order->rh_action_task);
clear_bit(type, pe_order_implies_then);
}
gIter = rh_actions;
for (; gIter != NULL; gIter = gIter->next) {
pe_action_t *rh_action_iter = (pe_action_t *) gIter->data;
if (lh_action) {
order_actions(lh_action, rh_action_iter, type);
} else if (type & pe_order_implies_then) {
update_action_flags(rh_action_iter, pe_action_runnable | pe_action_clear, __FUNCTION__, __LINE__);
crm_warn("Unrunnable %s 0x%.6x", rh_action_iter->uuid, type);
} else {
crm_warn("neither %s 0x%.6x", rh_action_iter->uuid, type);
}
}
g_list_free(rh_actions);
}
static void
rsc_order_first(pe_resource_t *lh_rsc, pe__ordering_t *order,
pe_working_set_t *data_set)
{
GListPtr gIter = NULL;
GListPtr lh_actions = NULL;
pe_action_t *lh_action = order->lh_action;
pe_resource_t *rh_rsc = order->rh_rsc;
crm_trace("Processing LH of ordering constraint %d", order->id);
CRM_ASSERT(lh_rsc != NULL);
if (lh_action != NULL) {
lh_actions = g_list_prepend(NULL, lh_action);
} else {
lh_actions = find_actions_by_task(lh_rsc->actions, lh_rsc, order->lh_action_task);
}
if (lh_actions == NULL && lh_rsc != rh_rsc) {
char *key = NULL;
char *op_type = NULL;
guint interval_ms = 0;
parse_op_key(order->lh_action_task, NULL, &op_type, &interval_ms);
key = pcmk__op_key(lh_rsc->id, op_type, interval_ms);
if (lh_rsc->fns->state(lh_rsc, TRUE) == RSC_ROLE_STOPPED && safe_str_eq(op_type, RSC_STOP)) {
free(key);
pe_rsc_trace(lh_rsc, "No LH-Side (%s/%s) found for constraint %d with %s - ignoring",
lh_rsc->id, order->lh_action_task, order->id, order->rh_action_task);
} else if (lh_rsc->fns->state(lh_rsc, TRUE) == RSC_ROLE_SLAVE && safe_str_eq(op_type, RSC_DEMOTE)) {
free(key);
pe_rsc_trace(lh_rsc, "No LH-Side (%s/%s) found for constraint %d with %s - ignoring",
lh_rsc->id, order->lh_action_task, order->id, order->rh_action_task);
} else {
pe_rsc_trace(lh_rsc, "No LH-Side (%s/%s) found for constraint %d with %s - creating",
lh_rsc->id, order->lh_action_task, order->id, order->rh_action_task);
lh_action = custom_action(lh_rsc, key, op_type, NULL, TRUE, TRUE, data_set);
lh_actions = g_list_prepend(NULL, lh_action);
}
free(op_type);
}
gIter = lh_actions;
for (; gIter != NULL; gIter = gIter->next) {
pe_action_t *lh_action_iter = (pe_action_t *) gIter->data;
if (rh_rsc == NULL && order->rh_action) {
rh_rsc = order->rh_action->rsc;
}
if (rh_rsc) {
rsc_order_then(lh_action_iter, rh_rsc, order);
} else if (order->rh_action) {
order_actions(lh_action_iter, order->rh_action, order->type);
}
}
g_list_free(lh_actions);
}
extern void update_colo_start_chain(pe_action_t *action,
pe_working_set_t *data_set);
static int
is_recurring_action(pe_action_t *action)
{
guint interval_ms;
if (pcmk__guint_from_hash(action->meta,
XML_LRM_ATTR_INTERVAL_MS, 0,
&interval_ms) != pcmk_rc_ok) {
return 0;
}
return (interval_ms > 0);
}
static void
apply_container_ordering(pe_action_t *action, pe_working_set_t *data_set)
{
/* VMs are also classified as containers for these purposes... in
* that they both involve a 'thing' running on a real or remote
* cluster node.
*
* This allows us to be smarter about the type and extent of
* recovery actions required in various scenarios
*/
pe_resource_t *remote_rsc = NULL;
pe_resource_t *container = NULL;
enum action_tasks task = text2task(action->task);
CRM_ASSERT(action->rsc);
CRM_ASSERT(action->node);
CRM_ASSERT(pe__is_guest_or_remote_node(action->node));
remote_rsc = action->node->details->remote_rsc;
CRM_ASSERT(remote_rsc);
container = remote_rsc->container;
CRM_ASSERT(container);
if(is_set(container->flags, pe_rsc_failed)) {
- pe_fence_node(data_set, action->node, "container failed");
+ pe_fence_node(data_set, action->node, "container failed", FALSE);
}
crm_trace("Order %s action %s relative to %s%s for %s%s",
action->task, action->uuid,
is_set(remote_rsc->flags, pe_rsc_failed)? "failed " : "",
remote_rsc->id,
is_set(container->flags, pe_rsc_failed)? "failed " : "",
container->id);
if (safe_str_eq(action->task, CRMD_ACTION_MIGRATE)
|| safe_str_eq(action->task, CRMD_ACTION_MIGRATED)) {
/* Migration ops map to "no_action", but we need to apply the same
* ordering as for stop or demote (see get_router_node()).
*/
task = stop_rsc;
}
switch (task) {
case start_rsc:
case action_promote:
/* Force resource recovery if the container is recovered */
order_start_then_action(container, action, pe_order_implies_then,
data_set);
/* Wait for the connection resource to be up too */
order_start_then_action(remote_rsc, action, pe_order_none,
data_set);
break;
case stop_rsc:
case action_demote:
if (is_set(container->flags, pe_rsc_failed)) {
/* When the container representing a guest node fails, any stop
* or demote actions for resources running on the guest node
* are implied by the container stopping. This is similar to
* how fencing operations work for cluster nodes and remote
* nodes.
*/
} else {
/* Ensure the operation happens before the connection is brought
* down.
*
* If we really wanted to, we could order these after the
* connection start, IFF the container's current role was
* stopped (otherwise we re-introduce an ordering loop when the
* connection is restarting).
*/
order_action_then_stop(action, remote_rsc, pe_order_none,
data_set);
}
break;
default:
/* Wait for the connection resource to be up */
if (is_recurring_action(action)) {
/* In case we ever get the recovery logic wrong, force
* recurring monitors to be restarted, even if just
* the connection was re-established
*/
if(task != no_action) {
order_start_then_action(remote_rsc, action,
pe_order_implies_then, data_set);
}
} else {
order_start_then_action(remote_rsc, action, pe_order_none,
data_set);
}
break;
}
}
static enum remote_connection_state
get_remote_node_state(pe_node_t *node)
{
pe_resource_t *remote_rsc = NULL;
pe_node_t *cluster_node = NULL;
CRM_ASSERT(node);
remote_rsc = node->details->remote_rsc;
CRM_ASSERT(remote_rsc);
cluster_node = pe__current_node(remote_rsc);
/* If the cluster node the remote connection resource resides on
* is unclean or went offline, we can't process any operations
* on that remote node until after it starts elsewhere.
*/
if(remote_rsc->next_role == RSC_ROLE_STOPPED || remote_rsc->allocated_to == NULL) {
/* The connection resource is not going to run anywhere */
if (cluster_node && cluster_node->details->unclean) {
/* The remote connection is failed because its resource is on a
* failed node and can't be recovered elsewhere, so we must fence.
*/
return remote_state_failed;
}
if (is_not_set(remote_rsc->flags, pe_rsc_failed)) {
/* Connection resource is cleanly stopped */
return remote_state_stopped;
}
/* Connection resource is failed */
if ((remote_rsc->next_role == RSC_ROLE_STOPPED)
&& remote_rsc->remote_reconnect_ms
&& node->details->remote_was_fenced
&& !pe__shutdown_requested(node)) {
/* We won't know whether the connection is recoverable until the
* reconnect interval expires and we reattempt connection.
*/
return remote_state_unknown;
}
/* The remote connection is in a failed state. If there are any
* resources known to be active on it (stop) or in an unknown state
* (probe), we must assume the worst and fence it.
*/
return remote_state_failed;
} else if (cluster_node == NULL) {
/* Connection is recoverable but not currently running anywhere, see if we can recover it first */
return remote_state_unknown;
} else if(cluster_node->details->unclean == TRUE
|| cluster_node->details->online == FALSE) {
/* Connection is running on a dead node, see if we can recover it first */
return remote_state_resting;
} else if (pcmk__list_of_multiple(remote_rsc->running_on)
&& remote_rsc->partial_migration_source
&& remote_rsc->partial_migration_target) {
/* We're in the middle of migrating a connection resource,
* wait until after the resource migrates before performing
* any actions.
*/
return remote_state_resting;
}
return remote_state_alive;
}
/*!
* \internal
* \brief Order actions on remote node relative to actions for the connection
*/
static void
apply_remote_ordering(pe_action_t *action, pe_working_set_t *data_set)
{
pe_resource_t *remote_rsc = NULL;
enum action_tasks task = text2task(action->task);
enum remote_connection_state state = get_remote_node_state(action->node);
enum pe_ordering order_opts = pe_order_none;
if (action->rsc == NULL) {
return;
}
CRM_ASSERT(action->node);
CRM_ASSERT(pe__is_guest_or_remote_node(action->node));
remote_rsc = action->node->details->remote_rsc;
CRM_ASSERT(remote_rsc);
crm_trace("Order %s action %s relative to %s%s (state: %s)",
action->task, action->uuid,
is_set(remote_rsc->flags, pe_rsc_failed)? "failed " : "",
remote_rsc->id, state2text(state));
if (safe_str_eq(action->task, CRMD_ACTION_MIGRATE)
|| safe_str_eq(action->task, CRMD_ACTION_MIGRATED)) {
/* Migration ops map to "no_action", but we need to apply the same
* ordering as for stop or demote (see get_router_node()).
*/
task = stop_rsc;
}
switch (task) {
case start_rsc:
case action_promote:
order_opts = pe_order_none;
if (state == remote_state_failed) {
/* Force recovery, by making this action required */
order_opts |= pe_order_implies_then;
}
/* Ensure connection is up before running this action */
order_start_then_action(remote_rsc, action, order_opts, data_set);
break;
case stop_rsc:
if(state == remote_state_alive) {
order_action_then_stop(action, remote_rsc,
pe_order_implies_first, data_set);
} else if(state == remote_state_failed) {
/* The resource is active on the node, but since we don't have a
* valid connection, the only way to stop the resource is by
* fencing the node. There is no need to order the stop relative
* to the remote connection, since the stop will become implied
* by the fencing.
*/
- pe_fence_node(data_set, action->node, "resources are active and the connection is unrecoverable");
+ pe_fence_node(data_set, action->node, "resources are active and the connection is unrecoverable", FALSE);
} else if(remote_rsc->next_role == RSC_ROLE_STOPPED) {
/* State must be remote_state_unknown or remote_state_stopped.
* Since the connection is not coming back up in this
* transition, stop this resource first.
*/
order_action_then_stop(action, remote_rsc,
pe_order_implies_first, data_set);
} else {
/* The connection is going to be started somewhere else, so
* stop this resource after that completes.
*/
order_start_then_action(remote_rsc, action, pe_order_none, data_set);
}
break;
case action_demote:
/* Only order this demote relative to the connection start if the
* connection isn't being torn down. Otherwise, the demote would be
* blocked because the connection start would not be allowed.
*/
if(state == remote_state_resting || state == remote_state_unknown) {
order_start_then_action(remote_rsc, action, pe_order_none,
data_set);
} /* Otherwise we can rely on the stop ordering */
break;
default:
/* Wait for the connection resource to be up */
if (is_recurring_action(action)) {
/* In case we ever get the recovery logic wrong, force
* recurring monitors to be restarted, even if just
* the connection was re-established
*/
order_start_then_action(remote_rsc, action,
pe_order_implies_then, data_set);
} else {
pe_node_t *cluster_node = pe__current_node(remote_rsc);
if(task == monitor_rsc && state == remote_state_failed) {
/* We would only be here if we do not know the
* state of the resource on the remote node.
* Since we have no way to find out, it is
* necessary to fence the node.
*/
- pe_fence_node(data_set, action->node, "resources are in an unknown state and the connection is unrecoverable");
+ pe_fence_node(data_set, action->node, "resources are in an unknown state and the connection is unrecoverable", FALSE);
}
if(cluster_node && state == remote_state_stopped) {
/* The connection is currently up, but is going
* down permanently.
*
* Make sure we check services are actually
* stopped _before_ we let the connection get
* closed
*/
order_action_then_stop(action, remote_rsc,
pe_order_runnable_left, data_set);
} else {
order_start_then_action(remote_rsc, action, pe_order_none,
data_set);
}
}
break;
}
}
static void
apply_remote_node_ordering(pe_working_set_t *data_set)
{
if (is_set(data_set->flags, pe_flag_have_remote_nodes) == FALSE) {
return;
}
for (GListPtr gIter = data_set->actions; gIter != NULL; gIter = gIter->next) {
pe_action_t *action = (pe_action_t *) gIter->data;
pe_resource_t *remote = NULL;
// We are only interested in resource actions
if (action->rsc == NULL) {
continue;
}
/* Special case: If we are clearing the failcount of an actual
* remote connection resource, then make sure this happens before
* any start of the resource in this transition.
*/
if (action->rsc->is_remote_node &&
safe_str_eq(action->task, CRM_OP_CLEAR_FAILCOUNT)) {
custom_action_order(action->rsc,
NULL,
action,
action->rsc,
pcmk__op_key(action->rsc->id, RSC_START, 0),
NULL,
pe_order_optional,
data_set);
continue;
}
// We are only interested in actions allocated to a node
if (action->node == NULL) {
continue;
}
if (!pe__is_guest_or_remote_node(action->node)) {
continue;
}
/* We are only interested in real actions.
*
* @TODO This is probably wrong; pseudo-actions might be converted to
* real actions and vice versa later in update_actions() at the end of
* stage7().
*/
if (is_set(action->flags, pe_action_pseudo)) {
continue;
}
remote = action->node->details->remote_rsc;
if (remote == NULL) {
// Orphaned
continue;
}
/* Another special case: if a resource is moving to a Pacemaker Remote
* node, order the stop on the original node after any start of the
* remote connection. This ensures that if the connection fails to
* start, we leave the resource running on the original node.
*/
if (safe_str_eq(action->task, RSC_START)) {
for (GList *item = action->rsc->actions; item != NULL;
item = item->next) {
pe_action_t *rsc_action = item->data;
if ((rsc_action->node->details != action->node->details)
&& safe_str_eq(rsc_action->task, RSC_STOP)) {
custom_action_order(remote, start_key(remote), NULL,
action->rsc, NULL, rsc_action,
pe_order_optional, data_set);
}
}
}
/* The action occurs across a remote connection, so create
* ordering constraints that guarantee the action occurs while the node
* is active (after start, before stop ... things like that).
*
* This is somewhat brittle in that we need to make sure the results of
* this ordering are compatible with the result of get_router_node().
* It would probably be better to add XML_LRM_ATTR_ROUTER_NODE as part
* of this logic rather than action2xml().
*/
if (remote->container) {
crm_trace("Container ordering for %s", action->uuid);
apply_container_ordering(action, data_set);
} else {
crm_trace("Remote ordering for %s", action->uuid);
apply_remote_ordering(action, data_set);
}
}
}
static gboolean
order_first_probe_unneeded(pe_action_t * probe, pe_action_t * rh_action)
{
/* No need to probe the resource on the node that is being
* unfenced. Otherwise it might introduce transition loop
* since probe will be performed after the node is
* unfenced.
*/
if (safe_str_eq(rh_action->task, CRM_OP_FENCE)
&& probe->node && rh_action->node
&& probe->node->details == rh_action->node->details) {
const char *op = g_hash_table_lookup(rh_action->meta, "stonith_action");
if (safe_str_eq(op, "on")) {
return TRUE;
}
}
// Shutdown waits for probe to complete only if it's on the same node
if ((safe_str_eq(rh_action->task, CRM_OP_SHUTDOWN))
&& probe->node && rh_action->node
&& probe->node->details != rh_action->node->details) {
return TRUE;
}
return FALSE;
}
static void
order_first_probes_imply_stops(pe_working_set_t * data_set)
{
GListPtr gIter = NULL;
for (gIter = data_set->ordering_constraints; gIter != NULL; gIter = gIter->next) {
pe__ordering_t *order = gIter->data;
enum pe_ordering order_type = pe_order_optional;
pe_resource_t *lh_rsc = order->lh_rsc;
pe_resource_t *rh_rsc = order->rh_rsc;
pe_action_t *lh_action = order->lh_action;
pe_action_t *rh_action = order->rh_action;
const char *lh_action_task = order->lh_action_task;
const char *rh_action_task = order->rh_action_task;
GListPtr probes = NULL;
GListPtr rh_actions = NULL;
GListPtr pIter = NULL;
if (lh_rsc == NULL) {
continue;
} else if (rh_rsc && lh_rsc == rh_rsc) {
continue;
}
if (lh_action == NULL && lh_action_task == NULL) {
continue;
}
if (rh_action == NULL && rh_action_task == NULL) {
continue;
}
/* Technically probe is expected to return "not running", which could be
* the alternative of stop action if the status of the resource is
* unknown yet.
*/
if (lh_action && safe_str_neq(lh_action->task, RSC_STOP)) {
continue;
} else if (lh_action == NULL
&& lh_action_task
&& !pcmk__ends_with(lh_action_task, "_" RSC_STOP "_0")) {
continue;
}
/* Do not probe the resource inside of a stopping container. Otherwise
* it might introduce transition loop since probe will be performed
* after the container starts again.
*/
if (rh_rsc && lh_rsc->container == rh_rsc) {
if (rh_action && safe_str_eq(rh_action->task, RSC_STOP)) {
continue;
} else if (rh_action == NULL && rh_action_task
&& pcmk__ends_with(rh_action_task,"_" RSC_STOP "_0")) {
continue;
}
}
if (order->type == pe_order_none) {
continue;
}
// Preserve the order options for future filtering
if (is_set(order->type, pe_order_apply_first_non_migratable)) {
set_bit(order_type, pe_order_apply_first_non_migratable);
}
if (is_set(order->type, pe_order_same_node)) {
set_bit(order_type, pe_order_same_node);
}
// Keep the order types for future filtering
if (order->type == pe_order_anti_colocation
|| order->type == pe_order_load) {
order_type = order->type;
}
probes = pe__resource_actions(lh_rsc, NULL, RSC_STATUS, FALSE);
if (probes == NULL) {
continue;
}
if (rh_action) {
rh_actions = g_list_prepend(rh_actions, rh_action);
} else if (rh_rsc && rh_action_task) {
rh_actions = find_actions(rh_rsc->actions, rh_action_task, NULL);
}
if (rh_actions == NULL) {
g_list_free(probes);
continue;
}
crm_trace("Processing for LH probe based on ordering constraint %s -> %s"
" (id=%d, type=%.6x)",
lh_action ? lh_action->uuid : lh_action_task,
rh_action ? rh_action->uuid : rh_action_task,
order->id, order->type);
for (pIter = probes; pIter != NULL; pIter = pIter->next) {
pe_action_t *probe = (pe_action_t *) pIter->data;
GListPtr rIter = NULL;
for (rIter = rh_actions; rIter != NULL; rIter = rIter->next) {
pe_action_t *rh_action_iter = (pe_action_t *) rIter->data;
if (order_first_probe_unneeded(probe, rh_action_iter)) {
continue;
}
order_actions(probe, rh_action_iter, order_type);
}
}
g_list_free(rh_actions);
g_list_free(probes);
}
}
static void
order_first_probe_then_restart_repromote(pe_action_t * probe,
pe_action_t * after,
pe_working_set_t * data_set)
{
GListPtr gIter = NULL;
bool interleave = FALSE;
pe_resource_t *compatible_rsc = NULL;
if (probe == NULL
|| probe->rsc == NULL
|| probe->rsc->variant != pe_native) {
return;
}
if (after == NULL
// Avoid running into any possible loop
|| is_set(after->flags, pe_action_tracking)) {
return;
}
if (safe_str_neq(probe->task, RSC_STATUS)) {
return;
}
pe_set_action_bit(after, pe_action_tracking);
crm_trace("Processing based on %s %s -> %s %s",
probe->uuid,
probe->node ? probe->node->details->uname: "",
after->uuid,
after->node ? after->node->details->uname : "");
if (after->rsc
/* Better not build a dependency directly with a clone/group.
* We are going to proceed through the ordering chain and build
* dependencies with its children.
*/
&& after->rsc->variant == pe_native
&& probe->rsc != after->rsc) {
GListPtr then_actions = NULL;
enum pe_ordering probe_order_type = pe_order_optional;
if (safe_str_eq(after->task, RSC_START)) {
then_actions = pe__resource_actions(after->rsc, NULL, RSC_STOP, FALSE);
} else if (safe_str_eq(after->task, RSC_PROMOTE)) {
then_actions = pe__resource_actions(after->rsc, NULL, RSC_DEMOTE, FALSE);
}
for (gIter = then_actions; gIter != NULL; gIter = gIter->next) {
pe_action_t *then = (pe_action_t *) gIter->data;
// Skip any pseudo action which for example is implied by fencing
if (is_set(then->flags, pe_action_pseudo)) {
continue;
}
order_actions(probe, then, probe_order_type);
}
g_list_free(then_actions);
}
if (after->rsc
&& after->rsc->variant > pe_group) {
const char *interleave_s = g_hash_table_lookup(after->rsc->meta,
XML_RSC_ATTR_INTERLEAVE);
interleave = crm_is_true(interleave_s);
if (interleave) {
/* For an interleaved clone, we should build a dependency only
* with the relevant clone child.
*/
compatible_rsc = find_compatible_child(probe->rsc,
after->rsc,
RSC_ROLE_UNKNOWN,
FALSE, data_set);
}
}
for (gIter = after->actions_after; gIter != NULL; gIter = gIter->next) {
pe_action_wrapper_t *after_wrapper = (pe_action_wrapper_t *) gIter->data;
/* pe_order_implies_then is the reason why a required A.start
* implies/enforces B.start to be required too, which is the cause of
* B.restart/re-promote.
*
* Not sure about pe_order_implies_then_on_node though. It's now only
* used for unfencing case, which tends to introduce transition
* loops...
*/
if (is_not_set(after_wrapper->type, pe_order_implies_then)) {
/* The order type between a group/clone and its child such as
* B.start-> B_child.start is:
* pe_order_implies_first_printed | pe_order_runnable_left
*
* Proceed through the ordering chain and build dependencies with
* its children.
*/
if (after->rsc == NULL
|| after->rsc->variant < pe_group
|| probe->rsc->parent == after->rsc
|| after_wrapper->action->rsc == NULL
|| after_wrapper->action->rsc->variant > pe_group
|| after->rsc != after_wrapper->action->rsc->parent) {
continue;
}
/* Proceed to the children of a group or a non-interleaved clone.
* For an interleaved clone, proceed only to the relevant child.
*/
if (after->rsc->variant > pe_group
&& interleave == TRUE
&& (compatible_rsc == NULL
|| compatible_rsc != after_wrapper->action->rsc)) {
continue;
}
}
crm_trace("Proceeding through %s %s -> %s %s (type=0x%.6x)",
after->uuid,
after->node ? after->node->details->uname: "",
after_wrapper->action->uuid,
after_wrapper->action->node ? after_wrapper->action->node->details->uname : "",
after_wrapper->type);
order_first_probe_then_restart_repromote(probe, after_wrapper->action, data_set);
}
}
static void clear_actions_tracking_flag(pe_working_set_t * data_set)
{
GListPtr gIter = NULL;
for (gIter = data_set->actions; gIter != NULL; gIter = gIter->next) {
pe_action_t *action = (pe_action_t *) gIter->data;
if (is_set(action->flags, pe_action_tracking)) {
pe_clear_action_bit(action, pe_action_tracking);
}
}
}
static void
order_first_rsc_probes(pe_resource_t * rsc, pe_working_set_t * data_set)
{
GListPtr gIter = NULL;
GListPtr probes = NULL;
for (gIter = rsc->children; gIter != NULL; gIter = gIter->next) {
pe_resource_t * child = (pe_resource_t *) gIter->data;
order_first_rsc_probes(child, data_set);
}
if (rsc->variant != pe_native) {
return;
}
probes = pe__resource_actions(rsc, NULL, RSC_STATUS, FALSE);
for (gIter = probes; gIter != NULL; gIter= gIter->next) {
pe_action_t *probe = (pe_action_t *) gIter->data;
GListPtr aIter = NULL;
for (aIter = probe->actions_after; aIter != NULL; aIter = aIter->next) {
pe_action_wrapper_t *after_wrapper = (pe_action_wrapper_t *) aIter->data;
order_first_probe_then_restart_repromote(probe, after_wrapper->action, data_set);
clear_actions_tracking_flag(data_set);
}
}
g_list_free(probes);
}
static void
order_first_probes(pe_working_set_t * data_set)
{
GListPtr gIter = NULL;
for (gIter = data_set->resources; gIter != NULL; gIter = gIter->next) {
pe_resource_t *rsc = (pe_resource_t *) gIter->data;
order_first_rsc_probes(rsc, data_set);
}
order_first_probes_imply_stops(data_set);
}
static void
order_then_probes(pe_working_set_t * data_set)
{
#if 0
GListPtr gIter = NULL;
for (gIter = data_set->resources; gIter != NULL; gIter = gIter->next) {
pe_resource_t *rsc = (pe_resource_t *) gIter->data;
/* Given "A then B", we would prefer to wait for A to be
* started before probing B.
*
* If A was a filesystem on which the binaries and data for B
* lived, it would have been useful if the author of B's agent
* could assume that A is running before B.monitor will be
* called.
*
* However we can't _only_ probe once A is running, otherwise
* we'd not detect the state of B if A could not be started
* for some reason.
*
* In practice however, we cannot even do an opportunistic
* version of this because B may be moving:
*
* B.probe -> B.start
* B.probe -> B.stop
* B.stop -> B.start
* A.stop -> A.start
* A.start -> B.probe
*
* So far so good, but if we add the result of this code:
*
* B.stop -> A.stop
*
* Then we get a loop:
*
* B.probe -> B.stop -> A.stop -> A.start -> B.probe
*
* We could kill the 'B.probe -> B.stop' dependency, but that
* could mean stopping B "too" soon, because B.start must wait
* for the probes to complete.
*
* Another option is to allow it only if A is a non-unique
* clone with clone-max == node-max (since we'll never be
* moving it). However, we could still be stopping one
* instance at the same time as starting another.
* The complexity of checking for allowed conditions combined
* with the ever narrowing usecase suggests that this code
* should remain disabled until someone gets smarter.
*/
pe_action_t *start = NULL;
GListPtr actions = NULL;
GListPtr probes = NULL;
actions = pe__resource_actions(rsc, NULL, RSC_START, FALSE);
if (actions) {
start = actions->data;
g_list_free(actions);
}
if(start == NULL) {
crm_err("No start action for %s", rsc->id);
continue;
}
probes = pe__resource_actions(rsc, NULL, RSC_STATUS, FALSE);
for (actions = start->actions_before; actions != NULL; actions = actions->next) {
pe_action_wrapper_t *before = (pe_action_wrapper_t *) actions->data;
GListPtr pIter = NULL;
pe_action_t *first = before->action;
pe_resource_t *first_rsc = first->rsc;
if(first->required_runnable_before) {
GListPtr clone_actions = NULL;
for (clone_actions = first->actions_before; clone_actions != NULL; clone_actions = clone_actions->next) {
before = (pe_action_wrapper_t *) clone_actions->data;
crm_trace("Testing %s -> %s (%p) for %s", first->uuid, before->action->uuid, before->action->rsc, start->uuid);
CRM_ASSERT(before->action->rsc);
first_rsc = before->action->rsc;
break;
}
} else if(safe_str_neq(first->task, RSC_START)) {
crm_trace("Not a start op %s for %s", first->uuid, start->uuid);
}
if(first_rsc == NULL) {
continue;
} else if(uber_parent(first_rsc) == uber_parent(start->rsc)) {
crm_trace("Same parent %s for %s", first_rsc->id, start->uuid);
continue;
} else if(FALSE && pe_rsc_is_clone(uber_parent(first_rsc)) == FALSE) {
crm_trace("Not a clone %s for %s", first_rsc->id, start->uuid);
continue;
}
crm_err("Applying %s before %s %d", first->uuid, start->uuid, uber_parent(first_rsc)->variant);
for (pIter = probes; pIter != NULL; pIter = pIter->next) {
pe_action_t *probe = (pe_action_t *) pIter->data;
crm_err("Ordering %s before %s", first->uuid, probe->uuid);
order_actions(first, probe, pe_order_optional);
}
}
}
#endif
}
static void
order_probes(pe_working_set_t * data_set)
{
order_first_probes(data_set);
order_then_probes(data_set);
}
gboolean
stage7(pe_working_set_t * data_set)
{
GList *gIter = NULL;
crm_trace("Applying ordering constraints");
/* Don't ask me why, but apparently they need to be processed in
* the order they were created in... go figure
*
* Also g_list_append() has horrendous performance characteristics
* So we need to use g_list_prepend() and then reverse the list here
*/
data_set->ordering_constraints = g_list_reverse(data_set->ordering_constraints);
for (gIter = data_set->ordering_constraints; gIter != NULL; gIter = gIter->next) {
pe__ordering_t *order = gIter->data;
pe_resource_t *rsc = order->lh_rsc;
crm_trace("Applying ordering constraint: %d", order->id);
if (rsc != NULL) {
crm_trace("rsc_action-to-*");
rsc_order_first(rsc, order, data_set);
continue;
}
rsc = order->rh_rsc;
if (rsc != NULL) {
crm_trace("action-to-rsc_action");
rsc_order_then(order->lh_action, rsc, order);
} else {
crm_trace("action-to-action");
order_actions(order->lh_action, order->rh_action, order->type);
}
}
for (gIter = data_set->actions; gIter != NULL; gIter = gIter->next) {
pe_action_t *action = (pe_action_t *) gIter->data;
update_colo_start_chain(action, data_set);
}
crm_trace("Ordering probes");
order_probes(data_set);
crm_trace("Updating %d actions", g_list_length(data_set->actions));
for (gIter = data_set->actions; gIter != NULL; gIter = gIter->next) {
pe_action_t *action = (pe_action_t *) gIter->data;
update_action(action, data_set);
}
// Check for invalid orderings
for (gIter = data_set->actions; gIter != NULL; gIter = gIter->next) {
pe_action_t *action = (pe_action_t *) gIter->data;
pe_action_wrapper_t *input = NULL;
for (GList *input_iter = action->actions_before;
input_iter != NULL; input_iter = input_iter->next) {
input = (pe_action_wrapper_t *) input_iter->data;
if (pcmk__ordering_is_invalid(action, input)) {
input->type = pe_order_none;
}
}
}
LogNodeActions(data_set, FALSE);
for (gIter = data_set->resources; gIter != NULL; gIter = gIter->next) {
pe_resource_t *rsc = (pe_resource_t *) gIter->data;
LogActions(rsc, data_set, FALSE);
}
return TRUE;
}
static int transition_id = -1;
/*!
* \internal
* \brief Log a message after calculating a transition
*
* \param[in] filename Where transition input is stored
*/
void
pcmk__log_transition_summary(const char *filename)
{
if (was_processing_error) {
crm_err("Calculated transition %d (with errors), saving inputs in %s",
transition_id, filename);
} else if (was_processing_warning) {
crm_warn("Calculated transition %d (with warnings), saving inputs in %s",
transition_id, filename);
} else {
crm_notice("Calculated transition %d, saving inputs in %s",
transition_id, filename);
}
if (crm_config_error) {
crm_notice("Configuration errors found during scheduler processing,"
" please run \"crm_verify -L\" to identify issues");
}
}
/*
* Create a dependency graph to send to the transitioner (via the controller)
*/
gboolean
stage8(pe_working_set_t * data_set)
{
GListPtr gIter = NULL;
const char *value = NULL;
transition_id++;
crm_trace("Creating transition graph %d.", transition_id);
data_set->graph = create_xml_node(NULL, XML_TAG_GRAPH);
value = pe_pref(data_set->config_hash, "cluster-delay");
crm_xml_add(data_set->graph, "cluster-delay", value);
value = pe_pref(data_set->config_hash, "stonith-timeout");
crm_xml_add(data_set->graph, "stonith-timeout", value);
crm_xml_add(data_set->graph, "failed-stop-offset", "INFINITY");
if (is_set(data_set->flags, pe_flag_start_failure_fatal)) {
crm_xml_add(data_set->graph, "failed-start-offset", "INFINITY");
} else {
crm_xml_add(data_set->graph, "failed-start-offset", "1");
}
value = pe_pref(data_set->config_hash, "batch-limit");
crm_xml_add(data_set->graph, "batch-limit", value);
crm_xml_add_int(data_set->graph, "transition_id", transition_id);
value = pe_pref(data_set->config_hash, "migration-limit");
if (crm_parse_ll(value, NULL) > 0) {
crm_xml_add(data_set->graph, "migration-limit", value);
}
if (data_set->recheck_by > 0) {
char *recheck_epoch = NULL;
recheck_epoch = crm_strdup_printf("%llu",
(long long) data_set->recheck_by);
crm_xml_add(data_set->graph, "recheck-by", recheck_epoch);
free(recheck_epoch);
}
/* errors...
slist_iter(action, pe_action_t, action_list, lpc,
if(action->optional == FALSE && action->runnable == FALSE) {
print_action("Ignoring", action, TRUE);
}
);
*/
/* The following code will de-duplicate action inputs, so nothing past this
* should rely on the action input type flags retaining their original
* values.
*/
gIter = data_set->resources;
for (; gIter != NULL; gIter = gIter->next) {
pe_resource_t *rsc = (pe_resource_t *) gIter->data;
pe_rsc_trace(rsc, "processing actions for rsc=%s", rsc->id);
rsc->cmds->expand(rsc, data_set);
}
crm_log_xml_trace(data_set->graph, "created resource-driven action list");
/* pseudo action to distribute list of nodes with maintenance state update */
add_maintenance_update(data_set);
/* catch any non-resource specific actions */
crm_trace("processing non-resource actions");
gIter = data_set->actions;
for (; gIter != NULL; gIter = gIter->next) {
pe_action_t *action = (pe_action_t *) gIter->data;
if (action->rsc
&& action->node
&& action->node->details->shutdown
&& is_not_set(action->rsc->flags, pe_rsc_maintenance)
&& is_not_set(action->flags, pe_action_optional)
&& is_not_set(action->flags, pe_action_runnable)
&& crm_str_eq(action->task, RSC_STOP, TRUE)
) {
/* Eventually we should just ignore the 'fence' case
* But for now it's the best way to detect (in CTS) when
* CIB resource updates are being lost
*/
if (is_set(data_set->flags, pe_flag_have_quorum)
|| data_set->no_quorum_policy == no_quorum_ignore) {
crm_crit("Cannot %s node '%s' because of %s:%s%s (%s)",
action->node->details->unclean ? "fence" : "shut down",
action->node->details->uname, action->rsc->id,
is_not_set(action->rsc->flags, pe_rsc_managed) ? " unmanaged" : " blocked",
is_set(action->rsc->flags, pe_rsc_failed) ? " failed" : "",
action->uuid);
}
}
graph_element_from_action(action, data_set);
}
crm_log_xml_trace(data_set->graph, "created generic action list");
crm_trace("Created transition graph %d.", transition_id);
return TRUE;
}
void
LogNodeActions(pe_working_set_t * data_set, gboolean terminal)
{
GListPtr gIter = NULL;
for (gIter = data_set->actions; gIter != NULL; gIter = gIter->next) {
char *node_name = NULL;
char *task = NULL;
pe_action_t *action = (pe_action_t *) gIter->data;
if (action->rsc != NULL) {
continue;
} else if (is_set(action->flags, pe_action_optional)) {
continue;
}
if (pe__is_guest_node(action->node)) {
node_name = crm_strdup_printf("%s (resource: %s)", action->node->details->uname, action->node->details->remote_rsc->container->id);
} else if(action->node) {
node_name = crm_strdup_printf("%s", action->node->details->uname);
}
if (safe_str_eq(action->task, CRM_OP_SHUTDOWN)) {
task = strdup("Shutdown");
} else if (safe_str_eq(action->task, CRM_OP_FENCE)) {
const char *op = g_hash_table_lookup(action->meta, "stonith_action");
task = crm_strdup_printf("Fence (%s)", op);
}
if(task == NULL) {
/* Nothing to report */
} else if(terminal && action->reason) {
printf(" * %s %s '%s'\n", task, node_name, action->reason);
} else if(terminal) {
printf(" * %s %s\n", task, node_name);
} else if(action->reason) {
crm_notice(" * %s %s '%s'\n", task, node_name, action->reason);
} else {
crm_notice(" * %s %s\n", task, node_name);
}
free(node_name);
free(task);
}
}
diff --git a/lib/pacemaker/pcmk_sched_native.c b/lib/pacemaker/pcmk_sched_native.c
index e63736af35..9ff3028765 100644
--- a/lib/pacemaker/pcmk_sched_native.c
+++ b/lib/pacemaker/pcmk_sched_native.c
@@ -1,3406 +1,3406 @@
/*
* Copyright 2004-2020 the Pacemaker project contributors
*
* The version control history for this file may have further details.
*
* This source code is licensed under the GNU General Public License version 2
* or later (GPLv2+) WITHOUT ANY WARRANTY.
*/
#include
#include
#include
#include
#include
// The controller removes the resource from the CIB, making this redundant
// #define DELETE_THEN_REFRESH 1
#define INFINITY_HACK (INFINITY * -100)
#define VARIANT_NATIVE 1
#include
void native_rsc_colocation_rh_must(pe_resource_t * rsc_lh, gboolean update_lh,
pe_resource_t * rsc_rh, gboolean update_rh);
void native_rsc_colocation_rh_mustnot(pe_resource_t * rsc_lh, gboolean update_lh,
pe_resource_t * rsc_rh, gboolean update_rh);
static void Recurring(pe_resource_t *rsc, pe_action_t *start, pe_node_t *node,
pe_working_set_t *data_set);
static void RecurringOp(pe_resource_t *rsc, pe_action_t *start, pe_node_t *node,
xmlNode *operation, pe_working_set_t *data_set);
static void Recurring_Stopped(pe_resource_t *rsc, pe_action_t *start, pe_node_t *node,
pe_working_set_t *data_set);
static void RecurringOp_Stopped(pe_resource_t *rsc, pe_action_t *start, pe_node_t *node,
xmlNode *operation, pe_working_set_t *data_set);
void ReloadRsc(pe_resource_t * rsc, pe_node_t *node, pe_working_set_t * data_set);
gboolean DeleteRsc(pe_resource_t * rsc, pe_node_t * node, gboolean optional, pe_working_set_t * data_set);
gboolean StopRsc(pe_resource_t * rsc, pe_node_t * next, gboolean optional, pe_working_set_t * data_set);
gboolean StartRsc(pe_resource_t * rsc, pe_node_t * next, gboolean optional, pe_working_set_t * data_set);
gboolean DemoteRsc(pe_resource_t * rsc, pe_node_t * next, gboolean optional, pe_working_set_t * data_set);
gboolean PromoteRsc(pe_resource_t * rsc, pe_node_t * next, gboolean optional,
pe_working_set_t * data_set);
gboolean RoleError(pe_resource_t * rsc, pe_node_t * next, gboolean optional, pe_working_set_t * data_set);
gboolean NullOp(pe_resource_t * rsc, pe_node_t * next, gboolean optional, pe_working_set_t * data_set);
/* *INDENT-OFF* */
enum rsc_role_e rsc_state_matrix[RSC_ROLE_MAX][RSC_ROLE_MAX] = {
/* Current State */
/* Next State: Unknown Stopped Started Slave Master */
/* Unknown */ { RSC_ROLE_UNKNOWN, RSC_ROLE_STOPPED, RSC_ROLE_STOPPED, RSC_ROLE_STOPPED, RSC_ROLE_STOPPED, },
/* Stopped */ { RSC_ROLE_STOPPED, RSC_ROLE_STOPPED, RSC_ROLE_STARTED, RSC_ROLE_SLAVE, RSC_ROLE_SLAVE, },
/* Started */ { RSC_ROLE_STOPPED, RSC_ROLE_STOPPED, RSC_ROLE_STARTED, RSC_ROLE_SLAVE, RSC_ROLE_MASTER, },
/* Slave */ { RSC_ROLE_STOPPED, RSC_ROLE_STOPPED, RSC_ROLE_STOPPED, RSC_ROLE_SLAVE, RSC_ROLE_MASTER, },
/* Master */ { RSC_ROLE_STOPPED, RSC_ROLE_SLAVE, RSC_ROLE_SLAVE, RSC_ROLE_SLAVE, RSC_ROLE_MASTER, },
};
gboolean (*rsc_action_matrix[RSC_ROLE_MAX][RSC_ROLE_MAX])(pe_resource_t*,pe_node_t*,gboolean,pe_working_set_t*) = {
/* Current State */
/* Next State: Unknown Stopped Started Slave Master */
/* Unknown */ { RoleError, StopRsc, RoleError, RoleError, RoleError, },
/* Stopped */ { RoleError, NullOp, StartRsc, StartRsc, RoleError, },
/* Started */ { RoleError, StopRsc, NullOp, NullOp, PromoteRsc, },
/* Slave */ { RoleError, StopRsc, StopRsc, NullOp, PromoteRsc, },
/* Master */ { RoleError, DemoteRsc, DemoteRsc, DemoteRsc, NullOp, },
};
/* *INDENT-ON* */
static gboolean
native_choose_node(pe_resource_t * rsc, pe_node_t * prefer, pe_working_set_t * data_set)
{
GListPtr nodes = NULL;
pe_node_t *chosen = NULL;
pe_node_t *best = NULL;
int multiple = 1;
int length = 0;
gboolean result = FALSE;
process_utilization(rsc, &prefer, data_set);
if (is_not_set(rsc->flags, pe_rsc_provisional)) {
return rsc->allocated_to ? TRUE : FALSE;
}
// Sort allowed nodes by weight
if (rsc->allowed_nodes) {
length = g_hash_table_size(rsc->allowed_nodes);
}
if (length > 0) {
nodes = g_hash_table_get_values(rsc->allowed_nodes);
nodes = sort_nodes_by_weight(nodes, pe__current_node(rsc), data_set);
// First node in sorted list has the best score
best = g_list_nth_data(nodes, 0);
}
if (prefer && nodes) {
chosen = g_hash_table_lookup(rsc->allowed_nodes, prefer->details->id);
if (chosen == NULL) {
pe_rsc_trace(rsc, "Preferred node %s for %s was unknown",
prefer->details->uname, rsc->id);
/* Favor the preferred node as long as its weight is at least as good as
* the best allowed node's.
*
* An alternative would be to favor the preferred node even if the best
* node is better, when the best node's weight is less than INFINITY.
*/
} else if ((chosen->weight < 0) || (chosen->weight < best->weight)) {
pe_rsc_trace(rsc, "Preferred node %s for %s was unsuitable",
chosen->details->uname, rsc->id);
chosen = NULL;
} else if (!can_run_resources(chosen)) {
pe_rsc_trace(rsc, "Preferred node %s for %s was unavailable",
chosen->details->uname, rsc->id);
chosen = NULL;
} else {
pe_rsc_trace(rsc,
"Chose preferred node %s for %s (ignoring %d candidates)",
chosen->details->uname, rsc->id, length);
}
}
if ((chosen == NULL) && nodes) {
/* Either there is no preferred node, or the preferred node is not
* available, but there are other nodes allowed to run the resource.
*/
chosen = best;
pe_rsc_trace(rsc, "Chose node %s for %s from %d candidates",
chosen ? chosen->details->uname : "", rsc->id, length);
if (!pe_rsc_is_unique_clone(rsc->parent)
&& chosen && (chosen->weight > 0) && can_run_resources(chosen)) {
/* If the resource is already running on a node, prefer that node if
* it is just as good as the chosen node.
*
* We don't do this for unique clone instances, because
* distribute_children() has already assigned instances to their
* running nodes when appropriate, and if we get here, we don't want
* remaining unallocated instances to prefer a node that's already
* running another instance.
*/
pe_node_t *running = pe__current_node(rsc);
if (running && (can_run_resources(running) == FALSE)) {
pe_rsc_trace(rsc, "Current node for %s (%s) can't run resources",
rsc->id, running->details->uname);
} else if (running) {
for (GList *iter = nodes->next; iter; iter = iter->next) {
pe_node_t *tmp = (pe_node_t *) iter->data;
if (tmp->weight != chosen->weight) {
// The nodes are sorted by weight, so no more are equal
break;
}
if (tmp->details == running->details) {
// Scores are equal, so prefer the current node
chosen = tmp;
}
multiple++;
}
}
}
}
if (multiple > 1) {
static char score[33];
int log_level = (chosen->weight >= INFINITY)? LOG_WARNING : LOG_INFO;
score2char_stack(chosen->weight, score, sizeof(score));
do_crm_log(log_level,
"Chose node %s for %s from %d nodes with score %s",
chosen->details->uname, rsc->id, multiple, score);
}
result = native_assign_node(rsc, nodes, chosen, FALSE);
g_list_free(nodes);
return result;
}
static int
node_list_attr_score(GHashTable * list, const char *attr, const char *value)
{
GHashTableIter iter;
pe_node_t *node = NULL;
int best_score = -INFINITY;
const char *best_node = NULL;
if (attr == NULL) {
attr = CRM_ATTR_UNAME;
}
g_hash_table_iter_init(&iter, list);
while (g_hash_table_iter_next(&iter, NULL, (void **)&node)) {
int weight = node->weight;
if (can_run_resources(node) == FALSE) {
weight = -INFINITY;
}
if (weight > best_score || best_node == NULL) {
const char *tmp = pe_node_attribute_raw(node, attr);
if (safe_str_eq(value, tmp)) {
best_score = weight;
best_node = node->details->uname;
}
}
}
if (safe_str_neq(attr, CRM_ATTR_UNAME)) {
crm_info("Best score for %s=%s was %s with %d",
attr, value, best_node ? best_node : "", best_score);
}
return best_score;
}
static void
node_hash_update(GHashTable * list1, GHashTable * list2, const char *attr, float factor,
gboolean only_positive)
{
int score = 0;
int new_score = 0;
GHashTableIter iter;
pe_node_t *node = NULL;
if (attr == NULL) {
attr = CRM_ATTR_UNAME;
}
g_hash_table_iter_init(&iter, list1);
while (g_hash_table_iter_next(&iter, NULL, (void **)&node)) {
float weight_f = 0;
int weight = 0;
score = node_list_attr_score(list2, attr, pe_node_attribute_raw(node, attr));
if ((factor < 0) && (score < 0)) {
/* Negative preference for a node with a negative score
* should not become a positive preference.
*
* @TODO Consider filtering only if weight == -INFINITY
*/
crm_trace("%s: Filtering %d + %f * %d (double negative disallowed)",
node->details->uname, node->weight, factor, score);
continue;
}
if (node->weight == INFINITY_HACK) {
crm_trace("%s: Filtering %d + %f * %d (node marked unusable)",
node->details->uname, node->weight, factor, score);
continue;
}
weight_f = factor * score;
// Round the number; see http://c-faq.com/fp/round.html
weight = (int) ((weight_f < 0)? (weight_f - 0.5) : (weight_f + 0.5));
/* Small factors can obliterate the small scores that are often actually
* used in configurations. If the score and factor are nonzero, ensure
* that the result is nonzero as well.
*/
if ((weight == 0) && (score != 0)) {
if (factor > 0.0) {
weight = 1;
} else if (factor < 0.0) {
weight = -1;
}
}
new_score = merge_weights(weight, node->weight);
if (only_positive && (new_score < 0) && (node->weight > 0)) {
crm_trace("%s: Filtering %d + %f * %d = %d "
"(negative disallowed, marking node unusable)",
node->details->uname, node->weight, factor, score,
new_score);
node->weight = INFINITY_HACK;
continue;
}
if (only_positive && (new_score < 0) && (node->weight == 0)) {
crm_trace("%s: Filtering %d + %f * %d = %d (negative disallowed)",
node->details->uname, node->weight, factor, score,
new_score);
continue;
}
crm_trace("%s: %d + %f * %d = %d", node->details->uname,
node->weight, factor, score, new_score);
node->weight = new_score;
}
}
GHashTable *
node_hash_dup(GHashTable * hash)
{
/* Hack! */
GListPtr list = g_hash_table_get_values(hash);
GHashTable *result = node_hash_from_list(list);
g_list_free(list);
return result;
}
/*!
* \internal
* \brief Incorporate colocation constraint scores into node weights
*
* \param[in,out] rsc Resource being placed
* \param[in] rhs ID of 'with' resource
* \param[in,out] nodes Nodes, with scores as of this point
* \param[in] attr Colocation attribute (ID by default)
* \param[in] factor Incorporate scores multiplied by this factor
* \param[in] flags Bitmask of enum pe_weights values
*
* \return Nodes, with scores modified by this constraint
* \note This function assumes ownership of the nodes argument. The caller
* should free the returned copy rather than the original.
*/
GHashTable *
pcmk__native_merge_weights(pe_resource_t *rsc, const char *rhs,
GHashTable *nodes, const char *attr, float factor,
uint32_t flags)
{
GHashTable *work = NULL;
// Avoid infinite recursion
if (is_set(rsc->flags, pe_rsc_merging)) {
pe_rsc_info(rsc, "%s: Breaking dependency loop at %s", rhs, rsc->id);
return nodes;
}
set_bit(rsc->flags, pe_rsc_merging);
if (is_set(flags, pe_weights_init)) {
if ((rsc->variant == pe_group) && (rsc->children != NULL)) {
GList *last = g_list_last(rsc->children);
pe_resource_t *last_rsc = last->data;
pe_rsc_trace(rsc, "%s: Merging scores from group %s "
"using last member %s (at %.6f)",
rhs, rsc->id, last_rsc->id, factor);
work = pcmk__native_merge_weights(last_rsc, rhs, NULL, attr, factor,
flags);
} else {
work = node_hash_dup(rsc->allowed_nodes);
}
clear_bit(flags, pe_weights_init);
} else if ((rsc->variant == pe_group) && (rsc->children != NULL)) {
pe_rsc_trace(rsc, "%s: Merging scores from %d members of group %s (at %.6f)",
rhs, g_list_length(rsc->children), rsc->id, factor);
work = node_hash_dup(nodes);
for (GList *iter = rsc->children; iter->next != NULL;
iter = iter->next) {
work = pcmk__native_merge_weights(iter->data, rhs, work, attr,
factor, flags);
}
} else {
pe_rsc_trace(rsc, "%s: Merging scores from %s (at %.6f)",
rhs, rsc->id, factor);
work = node_hash_dup(nodes);
node_hash_update(work, rsc->allowed_nodes, attr, factor,
is_set(flags, pe_weights_positive));
}
if (can_run_any(work)) {
GListPtr gIter = NULL;
int multiplier = (factor < 0)? -1 : 1;
if (is_set(flags, pe_weights_forward)) {
gIter = rsc->rsc_cons;
pe_rsc_trace(rsc,
"Checking additional %d optional '%s with' constraints",
g_list_length(gIter), rsc->id);
} else if ((rsc->variant == pe_group) && (rsc->children != NULL)) {
GList *last = g_list_last(rsc->children);
pe_resource_t *last_rsc = last->data;
gIter = last_rsc->rsc_cons_lhs;
pe_rsc_trace(rsc, "Checking additional %d optional 'with group %s' "
"constraints using last member %s",
g_list_length(gIter), rsc->id, last_rsc->id);
} else {
gIter = rsc->rsc_cons_lhs;
pe_rsc_trace(rsc,
"Checking additional %d optional 'with %s' constraints",
g_list_length(gIter), rsc->id);
}
for (; gIter != NULL; gIter = gIter->next) {
pe_resource_t *other = NULL;
rsc_colocation_t *constraint = (rsc_colocation_t *) gIter->data;
if (constraint->score == 0) {
continue;
}
if (is_set(flags, pe_weights_forward)) {
other = constraint->rsc_rh;
} else {
other = constraint->rsc_lh;
}
pe_rsc_trace(rsc, "Optionally merging score of '%s' constraint (%s with %s)",
constraint->id, constraint->rsc_lh->id,
constraint->rsc_rh->id);
work = pcmk__native_merge_weights(other, rhs, work,
constraint->node_attribute,
multiplier * constraint->score / (float) INFINITY,
flags|pe_weights_rollback);
pe__show_node_weights(true, NULL, rhs, work);
}
} else if (is_set(flags, pe_weights_rollback)) {
pe_rsc_info(rsc, "%s: Rolling back optional scores from %s",
rhs, rsc->id);
g_hash_table_destroy(work);
clear_bit(rsc->flags, pe_rsc_merging);
return nodes;
}
if (is_set(flags, pe_weights_positive)) {
pe_node_t *node = NULL;
GHashTableIter iter;
g_hash_table_iter_init(&iter, work);
while (g_hash_table_iter_next(&iter, NULL, (void **)&node)) {
if (node->weight == INFINITY_HACK) {
node->weight = 1;
}
}
}
if (nodes) {
g_hash_table_destroy(nodes);
}
clear_bit(rsc->flags, pe_rsc_merging);
return work;
}
static inline bool
node_has_been_unfenced(pe_node_t *node)
{
const char *unfenced = pe_node_attribute_raw(node, CRM_ATTR_UNFENCED);
return unfenced && strcmp("0", unfenced);
}
static inline bool
is_unfence_device(pe_resource_t *rsc, pe_working_set_t *data_set)
{
return is_set(rsc->flags, pe_rsc_fence_device)
&& is_set(data_set->flags, pe_flag_enable_unfencing);
}
pe_node_t *
pcmk__native_allocate(pe_resource_t *rsc, pe_node_t *prefer,
pe_working_set_t *data_set)
{
GListPtr gIter = NULL;
if (rsc->parent && is_not_set(rsc->parent->flags, pe_rsc_allocating)) {
/* never allocate children on their own */
pe_rsc_debug(rsc, "Escalating allocation of %s to its parent: %s", rsc->id,
rsc->parent->id);
rsc->parent->cmds->allocate(rsc->parent, prefer, data_set);
}
if (is_not_set(rsc->flags, pe_rsc_provisional)) {
return rsc->allocated_to;
}
if (is_set(rsc->flags, pe_rsc_allocating)) {
pe_rsc_debug(rsc, "Dependency loop detected involving %s", rsc->id);
return NULL;
}
set_bit(rsc->flags, pe_rsc_allocating);
pe__show_node_weights(true, rsc, "Pre-alloc", rsc->allowed_nodes);
for (gIter = rsc->rsc_cons; gIter != NULL; gIter = gIter->next) {
rsc_colocation_t *constraint = (rsc_colocation_t *) gIter->data;
GHashTable *archive = NULL;
pe_resource_t *rsc_rh = constraint->rsc_rh;
if (constraint->score == 0) {
continue;
}
if (constraint->role_lh >= RSC_ROLE_MASTER
|| (constraint->score < 0 && constraint->score > -INFINITY)) {
archive = node_hash_dup(rsc->allowed_nodes);
}
pe_rsc_trace(rsc,
"%s: Allocating %s first (constraint=%s score=%d role=%s)",
rsc->id, rsc_rh->id, constraint->id,
constraint->score, role2text(constraint->role_lh));
rsc_rh->cmds->allocate(rsc_rh, NULL, data_set);
rsc->cmds->rsc_colocation_lh(rsc, rsc_rh, constraint, data_set);
if (archive && can_run_any(rsc->allowed_nodes) == FALSE) {
pe_rsc_info(rsc, "%s: Rolling back scores from %s", rsc->id, rsc_rh->id);
g_hash_table_destroy(rsc->allowed_nodes);
rsc->allowed_nodes = archive;
archive = NULL;
}
if (archive) {
g_hash_table_destroy(archive);
}
}
pe__show_node_weights(true, rsc, "Post-coloc", rsc->allowed_nodes);
for (gIter = rsc->rsc_cons_lhs; gIter != NULL; gIter = gIter->next) {
rsc_colocation_t *constraint = (rsc_colocation_t *) gIter->data;
if (constraint->score == 0) {
continue;
}
pe_rsc_trace(rsc, "Merging score of '%s' constraint (%s with %s)",
constraint->id, constraint->rsc_lh->id,
constraint->rsc_rh->id);
rsc->allowed_nodes =
constraint->rsc_lh->cmds->merge_weights(constraint->rsc_lh, rsc->id, rsc->allowed_nodes,
constraint->node_attribute,
(float)constraint->score / INFINITY,
pe_weights_rollback);
}
if (rsc->next_role == RSC_ROLE_STOPPED) {
pe_rsc_trace(rsc, "Making sure %s doesn't get allocated", rsc->id);
/* make sure it doesn't come up again */
resource_location(rsc, NULL, -INFINITY, XML_RSC_ATTR_TARGET_ROLE, data_set);
} else if(rsc->next_role > rsc->role
&& is_set(data_set->flags, pe_flag_have_quorum) == FALSE
&& data_set->no_quorum_policy == no_quorum_freeze) {
crm_notice("Resource %s cannot be elevated from %s to %s: no-quorum-policy=freeze",
rsc->id, role2text(rsc->role), role2text(rsc->next_role));
rsc->next_role = rsc->role;
}
pe__show_node_weights(!show_scores, rsc, __FUNCTION__, rsc->allowed_nodes);
if (is_set(data_set->flags, pe_flag_stonith_enabled)
&& is_set(data_set->flags, pe_flag_have_stonith_resource) == FALSE) {
clear_bit(rsc->flags, pe_rsc_managed);
}
if (is_not_set(rsc->flags, pe_rsc_managed)) {
const char *reason = NULL;
pe_node_t *assign_to = NULL;
rsc->next_role = rsc->role;
assign_to = pe__current_node(rsc);
if (assign_to == NULL) {
reason = "inactive";
} else if (rsc->role == RSC_ROLE_MASTER) {
reason = "master";
} else if (is_set(rsc->flags, pe_rsc_failed)) {
reason = "failed";
} else {
reason = "active";
}
pe_rsc_info(rsc, "Unmanaged resource %s allocated to %s: %s", rsc->id,
(assign_to? assign_to->details->uname : "no node"), reason);
native_assign_node(rsc, NULL, assign_to, TRUE);
} else if (is_set(data_set->flags, pe_flag_stop_everything)) {
pe_rsc_debug(rsc, "Forcing %s to stop", rsc->id);
native_assign_node(rsc, NULL, NULL, TRUE);
} else if (is_set(rsc->flags, pe_rsc_provisional)
&& native_choose_node(rsc, prefer, data_set)) {
pe_rsc_trace(rsc, "Allocated resource %s to %s", rsc->id,
rsc->allocated_to->details->uname);
} else if (rsc->allocated_to == NULL) {
if (is_not_set(rsc->flags, pe_rsc_orphan)) {
pe_rsc_info(rsc, "Resource %s cannot run anywhere", rsc->id);
} else if (rsc->running_on != NULL) {
pe_rsc_info(rsc, "Stopping orphan resource %s", rsc->id);
}
} else {
pe_rsc_debug(rsc, "Pre-Allocated resource %s to %s", rsc->id,
rsc->allocated_to->details->uname);
}
clear_bit(rsc->flags, pe_rsc_allocating);
if (rsc->is_remote_node) {
pe_node_t *remote_node = pe_find_node(data_set->nodes, rsc->id);
CRM_ASSERT(remote_node != NULL);
if (rsc->allocated_to && rsc->next_role != RSC_ROLE_STOPPED) {
crm_trace("Setting Pacemaker Remote node %s to ONLINE",
remote_node->details->id);
remote_node->details->online = TRUE;
/* We shouldn't consider an unseen remote-node unclean if we are going
* to try and connect to it. Otherwise we get an unnecessary fence */
if (remote_node->details->unseen == TRUE) {
remote_node->details->unclean = FALSE;
}
} else {
crm_trace("Setting Pacemaker Remote node %s to SHUTDOWN (next role %s, %sallocated)",
remote_node->details->id, role2text(rsc->next_role),
(rsc->allocated_to? "" : "un"));
remote_node->details->shutdown = TRUE;
}
}
return rsc->allocated_to;
}
static gboolean
is_op_dup(pe_resource_t *rsc, const char *name, guint interval_ms)
{
gboolean dup = FALSE;
const char *id = NULL;
const char *value = NULL;
xmlNode *operation = NULL;
guint interval2_ms = 0;
CRM_ASSERT(rsc);
for (operation = __xml_first_child_element(rsc->ops_xml); operation != NULL;
operation = __xml_next_element(operation)) {
if (crm_str_eq((const char *)operation->name, "op", TRUE)) {
value = crm_element_value(operation, "name");
if (safe_str_neq(value, name)) {
continue;
}
value = crm_element_value(operation, XML_LRM_ATTR_INTERVAL);
interval2_ms = crm_parse_interval_spec(value);
if (interval_ms != interval2_ms) {
continue;
}
if (id == NULL) {
id = ID(operation);
} else {
pcmk__config_err("Operation %s is duplicate of %s (do not use "
"same name and interval combination more "
"than once per resource)", ID(operation), id);
dup = TRUE;
}
}
}
return dup;
}
static bool
op_cannot_recur(const char *name)
{
return safe_str_eq(name, RSC_STOP)
|| safe_str_eq(name, RSC_START)
|| safe_str_eq(name, RSC_DEMOTE)
|| safe_str_eq(name, RSC_PROMOTE);
}
static void
RecurringOp(pe_resource_t * rsc, pe_action_t * start, pe_node_t * node,
xmlNode * operation, pe_working_set_t * data_set)
{
char *key = NULL;
const char *name = NULL;
const char *role = NULL;
const char *interval_spec = NULL;
const char *node_uname = node? node->details->uname : "n/a";
guint interval_ms = 0;
pe_action_t *mon = NULL;
gboolean is_optional = TRUE;
GListPtr possible_matches = NULL;
CRM_ASSERT(rsc);
/* Only process for the operations without role="Stopped" */
role = crm_element_value(operation, "role");
if (role && text2role(role) == RSC_ROLE_STOPPED) {
return;
}
interval_spec = crm_element_value(operation, XML_LRM_ATTR_INTERVAL);
interval_ms = crm_parse_interval_spec(interval_spec);
if (interval_ms == 0) {
return;
}
name = crm_element_value(operation, "name");
if (is_op_dup(rsc, name, interval_ms)) {
crm_trace("Not creating duplicate recurring action %s for %dms %s",
ID(operation), interval_ms, name);
return;
}
if (op_cannot_recur(name)) {
pcmk__config_err("Ignoring %s because action '%s' cannot be recurring",
ID(operation), name);
return;
}
key = pcmk__op_key(rsc->id, name, interval_ms);
if (find_rsc_op_entry(rsc, key) == NULL) {
crm_trace("Not creating recurring action %s for disabled resource %s",
ID(operation), rsc->id);
free(key);
return;
}
pe_rsc_trace(rsc, "Creating recurring action %s for %s in role %s on %s",
ID(operation), rsc->id, role2text(rsc->next_role), node_uname);
if (start != NULL) {
pe_rsc_trace(rsc, "Marking %s %s due to %s",
key, is_set(start->flags, pe_action_optional) ? "optional" : "mandatory",
start->uuid);
is_optional = (rsc->cmds->action_flags(start, NULL) & pe_action_optional);
} else {
pe_rsc_trace(rsc, "Marking %s optional", key);
is_optional = TRUE;
}
/* start a monitor for an already active resource */
possible_matches = find_actions_exact(rsc->actions, key, node);
if (possible_matches == NULL) {
is_optional = FALSE;
pe_rsc_trace(rsc, "Marking %s mandatory: not active", key);
} else {
GListPtr gIter = NULL;
for (gIter = possible_matches; gIter != NULL; gIter = gIter->next) {
pe_action_t *op = (pe_action_t *) gIter->data;
if (is_set(op->flags, pe_action_reschedule)) {
is_optional = FALSE;
break;
}
}
g_list_free(possible_matches);
}
if ((rsc->next_role == RSC_ROLE_MASTER && role == NULL)
|| (role != NULL && text2role(role) != rsc->next_role)) {
int log_level = LOG_TRACE;
const char *result = "Ignoring";
if (is_optional) {
char *after_key = NULL;
pe_action_t *cancel_op = NULL;
// It's running, so cancel it
log_level = LOG_INFO;
result = "Cancelling";
cancel_op = pe_cancel_op(rsc, name, interval_ms, node, data_set);
switch (rsc->role) {
case RSC_ROLE_SLAVE:
case RSC_ROLE_STARTED:
if (rsc->next_role == RSC_ROLE_MASTER) {
after_key = promote_key(rsc);
} else if (rsc->next_role == RSC_ROLE_STOPPED) {
after_key = stop_key(rsc);
}
break;
case RSC_ROLE_MASTER:
after_key = demote_key(rsc);
break;
default:
break;
}
if (after_key) {
custom_action_order(rsc, NULL, cancel_op, rsc, after_key, NULL,
pe_order_runnable_left, data_set);
}
}
do_crm_log(log_level, "%s action %s (%s vs. %s)",
result, key, role ? role : role2text(RSC_ROLE_SLAVE),
role2text(rsc->next_role));
free(key);
return;
}
mon = custom_action(rsc, key, name, node, is_optional, TRUE, data_set);
key = mon->uuid;
if (is_optional) {
pe_rsc_trace(rsc, "%s\t %s (optional)", node_uname, mon->uuid);
}
if (start == NULL || is_set(start->flags, pe_action_runnable) == FALSE) {
pe_rsc_debug(rsc, "%s\t %s (cancelled : start un-runnable)",
node_uname, mon->uuid);
update_action_flags(mon, pe_action_runnable | pe_action_clear, __FUNCTION__, __LINE__);
} else if (node == NULL || node->details->online == FALSE || node->details->unclean) {
pe_rsc_debug(rsc, "%s\t %s (cancelled : no node available)",
node_uname, mon->uuid);
update_action_flags(mon, pe_action_runnable | pe_action_clear, __FUNCTION__, __LINE__);
} else if (is_set(mon->flags, pe_action_optional) == FALSE) {
pe_rsc_info(rsc, " Start recurring %s (%us) for %s on %s",
mon->task, interval_ms / 1000, rsc->id, node_uname);
}
if (rsc->next_role == RSC_ROLE_MASTER) {
char *running_master = crm_itoa(PCMK_OCF_RUNNING_MASTER);
add_hash_param(mon->meta, XML_ATTR_TE_TARGET_RC, running_master);
free(running_master);
}
if (node == NULL || is_set(rsc->flags, pe_rsc_managed)) {
custom_action_order(rsc, start_key(rsc), NULL,
NULL, strdup(key), mon,
pe_order_implies_then | pe_order_runnable_left, data_set);
custom_action_order(rsc, reload_key(rsc), NULL,
NULL, strdup(key), mon,
pe_order_implies_then | pe_order_runnable_left, data_set);
if (rsc->next_role == RSC_ROLE_MASTER) {
custom_action_order(rsc, promote_key(rsc), NULL,
rsc, NULL, mon,
pe_order_optional | pe_order_runnable_left, data_set);
} else if (rsc->role == RSC_ROLE_MASTER) {
custom_action_order(rsc, demote_key(rsc), NULL,
rsc, NULL, mon,
pe_order_optional | pe_order_runnable_left, data_set);
}
}
}
static void
Recurring(pe_resource_t * rsc, pe_action_t * start, pe_node_t * node, pe_working_set_t * data_set)
{
if (is_not_set(rsc->flags, pe_rsc_maintenance) &&
(node == NULL || node->details->maintenance == FALSE)) {
xmlNode *operation = NULL;
for (operation = __xml_first_child_element(rsc->ops_xml);
operation != NULL;
operation = __xml_next_element(operation)) {
if (crm_str_eq((const char *)operation->name, "op", TRUE)) {
RecurringOp(rsc, start, node, operation, data_set);
}
}
}
}
static void
RecurringOp_Stopped(pe_resource_t * rsc, pe_action_t * start, pe_node_t * node,
xmlNode * operation, pe_working_set_t * data_set)
{
char *key = NULL;
const char *name = NULL;
const char *role = NULL;
const char *interval_spec = NULL;
const char *node_uname = node? node->details->uname : "n/a";
guint interval_ms = 0;
GListPtr possible_matches = NULL;
GListPtr gIter = NULL;
/* Only process for the operations with role="Stopped" */
role = crm_element_value(operation, "role");
if (role == NULL || text2role(role) != RSC_ROLE_STOPPED) {
return;
}
interval_spec = crm_element_value(operation, XML_LRM_ATTR_INTERVAL);
interval_ms = crm_parse_interval_spec(interval_spec);
if (interval_ms == 0) {
return;
}
name = crm_element_value(operation, "name");
if (is_op_dup(rsc, name, interval_ms)) {
crm_trace("Not creating duplicate recurring action %s for %dms %s",
ID(operation), interval_ms, name);
return;
}
if (op_cannot_recur(name)) {
pcmk__config_err("Ignoring %s because action '%s' cannot be recurring",
ID(operation), name);
return;
}
key = pcmk__op_key(rsc->id, name, interval_ms);
if (find_rsc_op_entry(rsc, key) == NULL) {
crm_trace("Not creating recurring action %s for disabled resource %s",
ID(operation), rsc->id);
free(key);
return;
}
// @TODO add support
if (is_set(rsc->flags, pe_rsc_unique) == FALSE) {
crm_notice("Ignoring %s (recurring monitors for Stopped role are "
"not supported for anonymous clones)",
ID(operation));
return;
}
pe_rsc_trace(rsc,
"Creating recurring action %s for %s in role %s on nodes where it should not be running",
ID(operation), rsc->id, role2text(rsc->next_role));
/* if the monitor exists on the node where the resource will be running, cancel it */
if (node != NULL) {
possible_matches = find_actions_exact(rsc->actions, key, node);
if (possible_matches) {
pe_action_t *cancel_op = NULL;
g_list_free(possible_matches);
cancel_op = pe_cancel_op(rsc, name, interval_ms, node, data_set);
if (rsc->next_role == RSC_ROLE_STARTED || rsc->next_role == RSC_ROLE_SLAVE) {
/* rsc->role == RSC_ROLE_STOPPED: cancel the monitor before start */
/* rsc->role == RSC_ROLE_STARTED: for a migration, cancel the monitor on the target node before start */
custom_action_order(rsc, NULL, cancel_op, rsc, start_key(rsc), NULL,
pe_order_runnable_left, data_set);
}
pe_rsc_info(rsc, "Cancel action %s (%s vs. %s) on %s",
key, role, role2text(rsc->next_role), node_uname);
}
}
for (gIter = data_set->nodes; gIter != NULL; gIter = gIter->next) {
pe_node_t *stop_node = (pe_node_t *) gIter->data;
const char *stop_node_uname = stop_node->details->uname;
gboolean is_optional = TRUE;
gboolean probe_is_optional = TRUE;
gboolean stop_is_optional = TRUE;
pe_action_t *stopped_mon = NULL;
char *rc_inactive = NULL;
GListPtr probe_complete_ops = NULL;
GListPtr stop_ops = NULL;
GListPtr local_gIter = NULL;
if (node && safe_str_eq(stop_node_uname, node_uname)) {
continue;
}
pe_rsc_trace(rsc, "Creating recurring action %s for %s on %s",
ID(operation), rsc->id, crm_str(stop_node_uname));
/* start a monitor for an already stopped resource */
possible_matches = find_actions_exact(rsc->actions, key, stop_node);
if (possible_matches == NULL) {
pe_rsc_trace(rsc, "Marking %s mandatory on %s: not active", key,
crm_str(stop_node_uname));
is_optional = FALSE;
} else {
pe_rsc_trace(rsc, "Marking %s optional on %s: already active", key,
crm_str(stop_node_uname));
is_optional = TRUE;
g_list_free(possible_matches);
}
stopped_mon = custom_action(rsc, strdup(key), name, stop_node, is_optional, TRUE, data_set);
rc_inactive = crm_itoa(PCMK_OCF_NOT_RUNNING);
add_hash_param(stopped_mon->meta, XML_ATTR_TE_TARGET_RC, rc_inactive);
free(rc_inactive);
if (is_set(rsc->flags, pe_rsc_managed)) {
GList *probes = pe__resource_actions(rsc, stop_node, RSC_STATUS,
FALSE);
GListPtr pIter = NULL;
for (pIter = probes; pIter != NULL; pIter = pIter->next) {
pe_action_t *probe = (pe_action_t *) pIter->data;
order_actions(probe, stopped_mon, pe_order_runnable_left);
crm_trace("%s then %s on %s", probe->uuid, stopped_mon->uuid, stop_node->details->uname);
}
g_list_free(probes);
}
if (probe_complete_ops) {
g_list_free(probe_complete_ops);
}
stop_ops = pe__resource_actions(rsc, stop_node, RSC_STOP, TRUE);
for (local_gIter = stop_ops; local_gIter != NULL; local_gIter = local_gIter->next) {
pe_action_t *stop = (pe_action_t *) local_gIter->data;
if (is_set(stop->flags, pe_action_optional) == FALSE) {
stop_is_optional = FALSE;
}
if (is_set(stop->flags, pe_action_runnable) == FALSE) {
crm_debug("%s\t %s (cancelled : stop un-runnable)",
crm_str(stop_node_uname), stopped_mon->uuid);
update_action_flags(stopped_mon, pe_action_runnable | pe_action_clear, __FUNCTION__, __LINE__);
}
if (is_set(rsc->flags, pe_rsc_managed)) {
custom_action_order(rsc, stop_key(rsc), stop,
NULL, strdup(key), stopped_mon,
pe_order_implies_then | pe_order_runnable_left, data_set);
}
}
if (stop_ops) {
g_list_free(stop_ops);
}
if (is_optional == FALSE && probe_is_optional && stop_is_optional
&& is_set(rsc->flags, pe_rsc_managed) == FALSE) {
pe_rsc_trace(rsc, "Marking %s optional on %s due to unmanaged",
key, crm_str(stop_node_uname));
update_action_flags(stopped_mon, pe_action_optional, __FUNCTION__, __LINE__);
}
if (is_set(stopped_mon->flags, pe_action_optional)) {
pe_rsc_trace(rsc, "%s\t %s (optional)", crm_str(stop_node_uname), stopped_mon->uuid);
}
if (stop_node->details->online == FALSE || stop_node->details->unclean) {
pe_rsc_debug(rsc, "%s\t %s (cancelled : no node available)",
crm_str(stop_node_uname), stopped_mon->uuid);
update_action_flags(stopped_mon, pe_action_runnable | pe_action_clear, __FUNCTION__, __LINE__);
}
if (is_set(stopped_mon->flags, pe_action_runnable)
&& is_set(stopped_mon->flags, pe_action_optional) == FALSE) {
crm_notice(" Start recurring %s (%us) for %s on %s", stopped_mon->task,
interval_ms / 1000, rsc->id, crm_str(stop_node_uname));
}
}
free(key);
}
static void
Recurring_Stopped(pe_resource_t * rsc, pe_action_t * start, pe_node_t * node, pe_working_set_t * data_set)
{
if (is_not_set(rsc->flags, pe_rsc_maintenance) &&
(node == NULL || node->details->maintenance == FALSE)) {
xmlNode *operation = NULL;
for (operation = __xml_first_child_element(rsc->ops_xml);
operation != NULL;
operation = __xml_next_element(operation)) {
if (crm_str_eq((const char *)operation->name, "op", TRUE)) {
RecurringOp_Stopped(rsc, start, node, operation, data_set);
}
}
}
}
static void
handle_migration_actions(pe_resource_t * rsc, pe_node_t *current, pe_node_t *chosen, pe_working_set_t * data_set)
{
pe_action_t *migrate_to = NULL;
pe_action_t *migrate_from = NULL;
pe_action_t *start = NULL;
pe_action_t *stop = NULL;
gboolean partial = rsc->partial_migration_target ? TRUE : FALSE;
pe_rsc_trace(rsc, "Processing migration actions %s moving from %s to %s . partial migration = %s",
rsc->id, current->details->id, chosen->details->id, partial ? "TRUE" : "FALSE");
start = start_action(rsc, chosen, TRUE);
stop = stop_action(rsc, current, TRUE);
if (partial == FALSE) {
migrate_to = custom_action(rsc, pcmk__op_key(rsc->id, RSC_MIGRATE, 0),
RSC_MIGRATE, current, TRUE, TRUE, data_set);
}
migrate_from = custom_action(rsc, pcmk__op_key(rsc->id, RSC_MIGRATED, 0),
RSC_MIGRATED, chosen, TRUE, TRUE, data_set);
if ((migrate_to && migrate_from) || (migrate_from && partial)) {
set_bit(start->flags, pe_action_migrate_runnable);
set_bit(stop->flags, pe_action_migrate_runnable);
update_action_flags(start, pe_action_pseudo, __FUNCTION__, __LINE__); /* easier than trying to delete it from the graph */
/* order probes before migrations */
if (partial) {
set_bit(migrate_from->flags, pe_action_migrate_runnable);
migrate_from->needs = start->needs;
custom_action_order(rsc, pcmk__op_key(rsc->id, RSC_STATUS, 0), NULL,
rsc, pcmk__op_key(rsc->id, RSC_MIGRATED, 0),
NULL, pe_order_optional, data_set);
} else {
set_bit(migrate_from->flags, pe_action_migrate_runnable);
set_bit(migrate_to->flags, pe_action_migrate_runnable);
migrate_to->needs = start->needs;
custom_action_order(rsc, pcmk__op_key(rsc->id, RSC_STATUS, 0), NULL,
rsc, pcmk__op_key(rsc->id, RSC_MIGRATE, 0),
NULL, pe_order_optional, data_set);
custom_action_order(rsc, pcmk__op_key(rsc->id, RSC_MIGRATE, 0),
NULL, rsc,
pcmk__op_key(rsc->id, RSC_MIGRATED, 0), NULL,
pe_order_optional|pe_order_implies_first_migratable,
data_set);
}
custom_action_order(rsc, pcmk__op_key(rsc->id, RSC_MIGRATED, 0), NULL,
rsc, pcmk__op_key(rsc->id, RSC_STOP, 0), NULL,
pe_order_optional|pe_order_implies_first_migratable,
data_set);
custom_action_order(rsc, pcmk__op_key(rsc->id, RSC_MIGRATED, 0), NULL,
rsc, pcmk__op_key(rsc->id, RSC_START, 0), NULL,
pe_order_optional|pe_order_implies_first_migratable|pe_order_pseudo_left,
data_set);
}
if (migrate_to) {
add_hash_param(migrate_to->meta, XML_LRM_ATTR_MIGRATE_SOURCE, current->details->uname);
add_hash_param(migrate_to->meta, XML_LRM_ATTR_MIGRATE_TARGET, chosen->details->uname);
/* Pacemaker Remote connections don't require pending to be recorded in
* the CIB. We can reduce CIB writes by not setting PENDING for them.
*/
if (rsc->is_remote_node == FALSE) {
/* migrate_to takes place on the source node, but can
* have an effect on the target node depending on how
* the agent is written. Because of this, we have to maintain
* a record that the migrate_to occurred, in case the source node
* loses membership while the migrate_to action is still in-flight.
*/
add_hash_param(migrate_to->meta, XML_OP_ATTR_PENDING, "true");
}
}
if (migrate_from) {
add_hash_param(migrate_from->meta, XML_LRM_ATTR_MIGRATE_SOURCE, current->details->uname);
add_hash_param(migrate_from->meta, XML_LRM_ATTR_MIGRATE_TARGET, chosen->details->uname);
}
}
void
native_create_actions(pe_resource_t * rsc, pe_working_set_t * data_set)
{
pe_action_t *start = NULL;
pe_node_t *chosen = NULL;
pe_node_t *current = NULL;
gboolean need_stop = FALSE;
gboolean is_moving = FALSE;
gboolean allow_migrate = is_set(rsc->flags, pe_rsc_allow_migrate) ? TRUE : FALSE;
GListPtr gIter = NULL;
unsigned int num_all_active = 0;
unsigned int num_clean_active = 0;
bool multiply_active = FALSE;
enum rsc_role_e role = RSC_ROLE_UNKNOWN;
enum rsc_role_e next_role = RSC_ROLE_UNKNOWN;
CRM_ASSERT(rsc);
chosen = rsc->allocated_to;
if (chosen != NULL && rsc->next_role == RSC_ROLE_UNKNOWN) {
rsc->next_role = RSC_ROLE_STARTED;
pe_rsc_trace(rsc, "Fixed next_role: unknown -> %s", role2text(rsc->next_role));
} else if (rsc->next_role == RSC_ROLE_UNKNOWN) {
rsc->next_role = RSC_ROLE_STOPPED;
pe_rsc_trace(rsc, "Fixed next_role: unknown -> %s", role2text(rsc->next_role));
}
pe_rsc_trace(rsc, "Processing state transition for %s %p: %s->%s", rsc->id, rsc,
role2text(rsc->role), role2text(rsc->next_role));
current = pe__find_active_on(rsc, &num_all_active, &num_clean_active);
for (gIter = rsc->dangling_migrations; gIter != NULL; gIter = gIter->next) {
pe_node_t *dangling_source = (pe_node_t *) gIter->data;
pe_action_t *stop = stop_action(rsc, dangling_source, FALSE);
set_bit(stop->flags, pe_action_dangle);
pe_rsc_trace(rsc, "Forcing a cleanup of %s on %s",
rsc->id, dangling_source->details->uname);
if (is_set(data_set->flags, pe_flag_remove_after_stop)) {
DeleteRsc(rsc, dangling_source, FALSE, data_set);
}
}
if ((num_all_active == 2) && (num_clean_active == 2) && chosen
&& rsc->partial_migration_source && rsc->partial_migration_target
&& (current->details == rsc->partial_migration_source->details)
&& (chosen->details == rsc->partial_migration_target->details)) {
/* The chosen node is still the migration target from a partial
* migration. Attempt to continue the migration instead of recovering
* by stopping the resource everywhere and starting it on a single node.
*/
pe_rsc_trace(rsc,
"Will attempt to continue with a partial migration to target %s from %s",
rsc->partial_migration_target->details->id,
rsc->partial_migration_source->details->id);
} else if (is_not_set(rsc->flags, pe_rsc_needs_fencing)) {
/* If a resource has "requires" set to nothing or quorum, don't consider
* it active on unclean nodes (similar to how all resources behave when
* stonith-enabled is false). We can start such resources elsewhere
* before fencing completes, and if we considered the resource active on
* the failed node, we would attempt recovery for being active on
* multiple nodes.
*/
multiply_active = (num_clean_active > 1);
} else {
multiply_active = (num_all_active > 1);
}
if (multiply_active) {
if (rsc->partial_migration_target && rsc->partial_migration_source) {
// Migration was in progress, but we've chosen a different target
crm_notice("Resource %s can no longer migrate to %s. Stopping on %s too",
rsc->id, rsc->partial_migration_target->details->uname,
rsc->partial_migration_source->details->uname);
} else {
// Resource was incorrectly multiply active
pe_proc_err("Resource %s is active on %u nodes (%s)",
rsc->id, num_all_active,
recovery2text(rsc->recovery_type));
crm_notice("See https://wiki.clusterlabs.org/wiki/FAQ#Resource_is_Too_Active for more information");
}
if (rsc->recovery_type == recovery_stop_start) {
need_stop = TRUE;
}
/* If by chance a partial migration is in process, but the migration
* target is not chosen still, clear all partial migration data.
*/
rsc->partial_migration_source = rsc->partial_migration_target = NULL;
allow_migrate = FALSE;
}
if (is_set(rsc->flags, pe_rsc_start_pending)) {
start = start_action(rsc, chosen, TRUE);
set_bit(start->flags, pe_action_print_always);
}
if (current && chosen && current->details != chosen->details) {
pe_rsc_trace(rsc, "Moving %s", rsc->id);
is_moving = TRUE;
need_stop = TRUE;
} else if (is_set(rsc->flags, pe_rsc_failed)) {
pe_rsc_trace(rsc, "Recovering %s", rsc->id);
need_stop = TRUE;
} else if (is_set(rsc->flags, pe_rsc_block)) {
pe_rsc_trace(rsc, "Block %s", rsc->id);
need_stop = TRUE;
} else if (rsc->role > RSC_ROLE_STARTED && current != NULL && chosen != NULL) {
/* Recovery of a promoted resource */
start = start_action(rsc, chosen, TRUE);
if (is_set(start->flags, pe_action_optional) == FALSE) {
pe_rsc_trace(rsc, "Forced start %s", rsc->id);
need_stop = TRUE;
}
}
pe_rsc_trace(rsc, "Creating actions for %s: %s->%s", rsc->id,
role2text(rsc->role), role2text(rsc->next_role));
/* Create any additional actions required when bringing resource down and
* back up to same level.
*/
role = rsc->role;
while (role != RSC_ROLE_STOPPED) {
next_role = rsc_state_matrix[role][RSC_ROLE_STOPPED];
pe_rsc_trace(rsc, "Down: Executing: %s->%s (%s)%s", role2text(role), role2text(next_role),
rsc->id, need_stop ? " required" : "");
if (rsc_action_matrix[role][next_role] (rsc, current, !need_stop, data_set) == FALSE) {
break;
}
role = next_role;
}
while (rsc->role <= rsc->next_role && role != rsc->role && is_not_set(rsc->flags, pe_rsc_block)) {
next_role = rsc_state_matrix[role][rsc->role];
pe_rsc_trace(rsc, "Up: Executing: %s->%s (%s)%s", role2text(role), role2text(next_role),
rsc->id, need_stop ? " required" : "");
if (rsc_action_matrix[role][next_role] (rsc, chosen, !need_stop, data_set) == FALSE) {
break;
}
role = next_role;
}
role = rsc->role;
/* Required steps from this role to the next */
while (role != rsc->next_role) {
next_role = rsc_state_matrix[role][rsc->next_role];
pe_rsc_trace(rsc, "Role: Executing: %s->%s = (%s on %s)", role2text(role), role2text(next_role), rsc->id, chosen?chosen->details->uname:"NA");
if (rsc_action_matrix[role][next_role] (rsc, chosen, FALSE, data_set) == FALSE) {
break;
}
role = next_role;
}
if(is_set(rsc->flags, pe_rsc_block)) {
pe_rsc_trace(rsc, "No monitor additional ops for blocked resource");
} else if (rsc->next_role != RSC_ROLE_STOPPED || is_set(rsc->flags, pe_rsc_managed) == FALSE) {
pe_rsc_trace(rsc, "Monitor ops for active resource");
start = start_action(rsc, chosen, TRUE);
Recurring(rsc, start, chosen, data_set);
Recurring_Stopped(rsc, start, chosen, data_set);
} else {
pe_rsc_trace(rsc, "Monitor ops for inactive resource");
Recurring_Stopped(rsc, NULL, NULL, data_set);
}
/* if we are stuck in a partial migration, where the target
* of the partial migration no longer matches the chosen target.
* A full stop/start is required */
if (rsc->partial_migration_target && (chosen == NULL || rsc->partial_migration_target->details != chosen->details)) {
pe_rsc_trace(rsc, "Not allowing partial migration to continue. %s", rsc->id);
allow_migrate = FALSE;
} else if (is_moving == FALSE ||
is_not_set(rsc->flags, pe_rsc_managed) ||
is_set(rsc->flags, pe_rsc_failed) ||
is_set(rsc->flags, pe_rsc_start_pending) ||
(current && current->details->unclean) ||
rsc->next_role < RSC_ROLE_STARTED) {
allow_migrate = FALSE;
}
if (allow_migrate) {
handle_migration_actions(rsc, current, chosen, data_set);
}
}
static void
rsc_avoids_remote_nodes(pe_resource_t *rsc)
{
GHashTableIter iter;
pe_node_t *node = NULL;
g_hash_table_iter_init(&iter, rsc->allowed_nodes);
while (g_hash_table_iter_next(&iter, NULL, (void **)&node)) {
if (node->details->remote_rsc) {
node->weight = -INFINITY;
}
}
}
/*!
* \internal
* \brief Return allowed nodes as (possibly sorted) list
*
* Convert a resource's hash table of allowed nodes to a list. If printing to
* stdout, sort the list, to keep action ID numbers consistent for regression
* test output (while avoiding the performance hit on a live cluster).
*
* \param[in] rsc Resource to check for allowed nodes
* \param[in] data_set Cluster working set
*
* \return List of resource's allowed nodes
* \note Callers should take care not to rely on the list being sorted.
*/
static GList *
allowed_nodes_as_list(pe_resource_t *rsc, pe_working_set_t *data_set)
{
GList *allowed_nodes = NULL;
if (rsc->allowed_nodes) {
allowed_nodes = g_hash_table_get_values(rsc->allowed_nodes);
}
if (is_set(data_set->flags, pe_flag_stdout)) {
allowed_nodes = g_list_sort(allowed_nodes, sort_node_uname);
}
return allowed_nodes;
}
void
native_internal_constraints(pe_resource_t * rsc, pe_working_set_t * data_set)
{
/* This function is on the critical path and worth optimizing as much as possible */
pe_resource_t *top = NULL;
GList *allowed_nodes = NULL;
bool check_unfencing = FALSE;
bool check_utilization = FALSE;
if (is_not_set(rsc->flags, pe_rsc_managed)) {
pe_rsc_trace(rsc,
"Skipping native constraints for unmanaged resource: %s",
rsc->id);
return;
}
top = uber_parent(rsc);
// Whether resource requires unfencing
check_unfencing = is_not_set(rsc->flags, pe_rsc_fence_device)
&& is_set(data_set->flags, pe_flag_enable_unfencing)
&& is_set(rsc->flags, pe_rsc_needs_unfencing);
// Whether a non-default placement strategy is used
check_utilization = (g_hash_table_size(rsc->utilization) > 0)
&& safe_str_neq(data_set->placement_strategy, "default");
// Order stops before starts (i.e. restart)
custom_action_order(rsc, pcmk__op_key(rsc->id, RSC_STOP, 0), NULL,
rsc, pcmk__op_key(rsc->id, RSC_START, 0), NULL,
pe_order_optional|pe_order_implies_then|pe_order_restart,
data_set);
// Promotable ordering: demote before stop, start before promote
if (is_set(top->flags, pe_rsc_promotable) || (rsc->role > RSC_ROLE_SLAVE)) {
custom_action_order(rsc, pcmk__op_key(rsc->id, RSC_DEMOTE, 0), NULL,
rsc, pcmk__op_key(rsc->id, RSC_STOP, 0), NULL,
pe_order_implies_first_master, data_set);
custom_action_order(rsc, pcmk__op_key(rsc->id, RSC_START, 0), NULL,
rsc, pcmk__op_key(rsc->id, RSC_PROMOTE, 0), NULL,
pe_order_runnable_left, data_set);
}
// Don't clear resource history if probing on same node
custom_action_order(rsc, pcmk__op_key(rsc->id, CRM_OP_LRM_DELETE, 0),
NULL, rsc, pcmk__op_key(rsc->id, RSC_STATUS, 0),
NULL, pe_order_same_node|pe_order_then_cancels_first,
data_set);
// Certain checks need allowed nodes
if (check_unfencing || check_utilization || rsc->container) {
allowed_nodes = allowed_nodes_as_list(rsc, data_set);
}
if (check_unfencing) {
/* Check if the node needs to be unfenced first */
for (GList *item = allowed_nodes; item; item = item->next) {
pe_node_t *node = item->data;
- pe_action_t *unfence = pe_fence_op(node, "on", TRUE, NULL, data_set);
+ pe_action_t *unfence = pe_fence_op(node, "on", TRUE, NULL, FALSE, data_set);
crm_debug("Ordering any stops of %s before %s, and any starts after",
rsc->id, unfence->uuid);
/*
* It would be more efficient to order clone resources once,
* rather than order each instance, but ordering the instance
* allows us to avoid unnecessary dependencies that might conflict
* with user constraints.
*
* @TODO: This constraint can still produce a transition loop if the
* resource has a stop scheduled on the node being unfenced, and
* there is a user ordering constraint to start some other resource
* (which will be ordered after the unfence) before stopping this
* resource. An example is "start some slow-starting cloned service
* before stopping an associated virtual IP that may be moving to
* it":
* stop this -> unfencing -> start that -> stop this
*/
custom_action_order(rsc, stop_key(rsc), NULL,
NULL, strdup(unfence->uuid), unfence,
pe_order_optional|pe_order_same_node, data_set);
custom_action_order(NULL, strdup(unfence->uuid), unfence,
rsc, start_key(rsc), NULL,
pe_order_implies_then_on_node|pe_order_same_node,
data_set);
}
}
if (check_utilization) {
GListPtr gIter = NULL;
pe_rsc_trace(rsc, "Creating utilization constraints for %s - strategy: %s",
rsc->id, data_set->placement_strategy);
for (gIter = rsc->running_on; gIter != NULL; gIter = gIter->next) {
pe_node_t *current = (pe_node_t *) gIter->data;
char *load_stopped_task = crm_strdup_printf(LOAD_STOPPED "_%s",
current->details->uname);
pe_action_t *load_stopped = get_pseudo_op(load_stopped_task, data_set);
if (load_stopped->node == NULL) {
load_stopped->node = node_copy(current);
update_action_flags(load_stopped, pe_action_optional | pe_action_clear, __FUNCTION__, __LINE__);
}
custom_action_order(rsc, stop_key(rsc), NULL,
NULL, load_stopped_task, load_stopped, pe_order_load, data_set);
}
for (GList *item = allowed_nodes; item; item = item->next) {
pe_node_t *next = item->data;
char *load_stopped_task = crm_strdup_printf(LOAD_STOPPED "_%s",
next->details->uname);
pe_action_t *load_stopped = get_pseudo_op(load_stopped_task, data_set);
if (load_stopped->node == NULL) {
load_stopped->node = node_copy(next);
update_action_flags(load_stopped, pe_action_optional | pe_action_clear, __FUNCTION__, __LINE__);
}
custom_action_order(NULL, strdup(load_stopped_task), load_stopped,
rsc, start_key(rsc), NULL, pe_order_load, data_set);
custom_action_order(NULL, strdup(load_stopped_task), load_stopped,
rsc, pcmk__op_key(rsc->id, RSC_MIGRATE, 0),
NULL, pe_order_load, data_set);
free(load_stopped_task);
}
}
if (rsc->container) {
pe_resource_t *remote_rsc = NULL;
if (rsc->is_remote_node) {
// rsc is the implicit remote connection for a guest or bundle node
/* Do not allow a guest resource to live on a Pacemaker Remote node,
* to avoid nesting remotes. However, allow bundles to run on remote
* nodes.
*/
if (is_not_set(rsc->flags, pe_rsc_allow_remote_remotes)) {
rsc_avoids_remote_nodes(rsc->container);
}
/* If someone cleans up a guest or bundle node's container, we will
* likely schedule a (re-)probe of the container and recovery of the
* connection. Order the connection stop after the container probe,
* so that if we detect the container running, we will trigger a new
* transition and avoid the unnecessary recovery.
*/
new_rsc_order(rsc->container, RSC_STATUS, rsc, RSC_STOP,
pe_order_optional, data_set);
/* A user can specify that a resource must start on a Pacemaker Remote
* node by explicitly configuring it with the container=NODENAME
* meta-attribute. This is of questionable merit, since location
* constraints can accomplish the same thing. But we support it, so here
* we check whether a resource (that is not itself a remote connection)
* has container set to a remote node or guest node resource.
*/
} else if (rsc->container->is_remote_node) {
remote_rsc = rsc->container;
} else {
remote_rsc = pe__resource_contains_guest_node(data_set,
rsc->container);
}
if (remote_rsc) {
/* Force the resource on the Pacemaker Remote node instead of
* colocating the resource with the container resource.
*/
for (GList *item = allowed_nodes; item; item = item->next) {
pe_node_t *node = item->data;
if (node->details->remote_rsc != remote_rsc) {
node->weight = -INFINITY;
}
}
} else {
/* This resource is either a filler for a container that does NOT
* represent a Pacemaker Remote node, or a Pacemaker Remote
* connection resource for a guest node or bundle.
*/
int score;
crm_trace("Order and colocate %s relative to its container %s",
rsc->id, rsc->container->id);
custom_action_order(rsc->container,
pcmk__op_key(rsc->container->id, RSC_START, 0),
NULL, rsc, pcmk__op_key(rsc->id, RSC_START, 0),
NULL,
pe_order_implies_then|pe_order_runnable_left,
data_set);
custom_action_order(rsc, pcmk__op_key(rsc->id, RSC_STOP, 0), NULL,
rsc->container,
pcmk__op_key(rsc->container->id, RSC_STOP, 0),
NULL, pe_order_implies_first, data_set);
if (is_set(rsc->flags, pe_rsc_allow_remote_remotes)) {
score = 10000; /* Highly preferred but not essential */
} else {
score = INFINITY; /* Force them to run on the same host */
}
rsc_colocation_new("resource-with-container", NULL, score, rsc,
rsc->container, NULL, NULL, data_set);
}
}
if (rsc->is_remote_node || is_set(rsc->flags, pe_rsc_fence_device)) {
/* don't allow remote nodes to run stonith devices
* or remote connection resources.*/
rsc_avoids_remote_nodes(rsc);
}
g_list_free(allowed_nodes);
}
void
native_rsc_colocation_lh(pe_resource_t *rsc_lh, pe_resource_t *rsc_rh,
rsc_colocation_t *constraint,
pe_working_set_t *data_set)
{
if (rsc_lh == NULL) {
pe_err("rsc_lh was NULL for %s", constraint->id);
return;
} else if (constraint->rsc_rh == NULL) {
pe_err("rsc_rh was NULL for %s", constraint->id);
return;
}
if (constraint->score == 0) {
return;
}
pe_rsc_trace(rsc_lh, "Processing colocation constraint between %s and %s", rsc_lh->id,
rsc_rh->id);
rsc_rh->cmds->rsc_colocation_rh(rsc_lh, rsc_rh, constraint, data_set);
}
enum filter_colocation_res
filter_colocation_constraint(pe_resource_t * rsc_lh, pe_resource_t * rsc_rh,
rsc_colocation_t * constraint, gboolean preview)
{
if (constraint->score == 0) {
return influence_nothing;
}
/* rh side must be allocated before we can process constraint */
if (preview == FALSE && is_set(rsc_rh->flags, pe_rsc_provisional)) {
return influence_nothing;
}
if ((constraint->role_lh >= RSC_ROLE_SLAVE) &&
rsc_lh->parent && is_set(rsc_lh->parent->flags, pe_rsc_promotable)
&& is_not_set(rsc_lh->flags, pe_rsc_provisional)) {
/* LH and RH resources have already been allocated, place the correct
* priority on LH rsc for the given promotable clone resource role */
return influence_rsc_priority;
}
if (preview == FALSE && is_not_set(rsc_lh->flags, pe_rsc_provisional)) {
// Log an error if we violated a mandatory colocation constraint
const pe_node_t *rh_node = rsc_rh->allocated_to;
if (rsc_lh->allocated_to == NULL) {
// Dependent resource isn't allocated, so constraint doesn't matter
return influence_nothing;
}
if (constraint->score >= INFINITY) {
// Dependent resource must colocate with rh_node
if ((rh_node == NULL)
|| (rh_node->details != rsc_lh->allocated_to->details)) {
crm_err("%s must be colocated with %s but is not (%s vs. %s)",
rsc_lh->id, rsc_rh->id,
rsc_lh->allocated_to->details->uname,
(rh_node? rh_node->details->uname : "unallocated"));
}
} else if (constraint->score <= -INFINITY) {
// Dependent resource must anti-colocate with rh_node
if ((rh_node != NULL)
&& (rsc_lh->allocated_to->details == rh_node->details)) {
crm_err("%s and %s must be anti-colocated but are allocated "
"to the same node (%s)",
rsc_lh->id, rsc_rh->id, rh_node->details->uname);
}
}
return influence_nothing;
}
if (constraint->score > 0
&& constraint->role_lh != RSC_ROLE_UNKNOWN && constraint->role_lh != rsc_lh->next_role) {
crm_trace("LH: Skipping constraint: \"%s\" state filter nextrole is %s",
role2text(constraint->role_lh), role2text(rsc_lh->next_role));
return influence_nothing;
}
if (constraint->score > 0
&& constraint->role_rh != RSC_ROLE_UNKNOWN && constraint->role_rh != rsc_rh->next_role) {
crm_trace("RH: Skipping constraint: \"%s\" state filter", role2text(constraint->role_rh));
return FALSE;
}
if (constraint->score < 0
&& constraint->role_lh != RSC_ROLE_UNKNOWN && constraint->role_lh == rsc_lh->next_role) {
crm_trace("LH: Skipping negative constraint: \"%s\" state filter",
role2text(constraint->role_lh));
return influence_nothing;
}
if (constraint->score < 0
&& constraint->role_rh != RSC_ROLE_UNKNOWN && constraint->role_rh == rsc_rh->next_role) {
crm_trace("RH: Skipping negative constraint: \"%s\" state filter",
role2text(constraint->role_rh));
return influence_nothing;
}
return influence_rsc_location;
}
static void
influence_priority(pe_resource_t * rsc_lh, pe_resource_t * rsc_rh, rsc_colocation_t * constraint)
{
const char *rh_value = NULL;
const char *lh_value = NULL;
const char *attribute = CRM_ATTR_ID;
int score_multiplier = 1;
if (constraint->score == 0) {
return;
}
if (!rsc_rh->allocated_to || !rsc_lh->allocated_to) {
return;
}
if (constraint->node_attribute != NULL) {
attribute = constraint->node_attribute;
}
lh_value = pe_node_attribute_raw(rsc_lh->allocated_to, attribute);
rh_value = pe_node_attribute_raw(rsc_rh->allocated_to, attribute);
if (!safe_str_eq(lh_value, rh_value)) {
if(constraint->score == INFINITY && constraint->role_lh == RSC_ROLE_MASTER) {
rsc_lh->priority = -INFINITY;
}
return;
}
if (constraint->role_rh && (constraint->role_rh != rsc_rh->next_role)) {
return;
}
if (constraint->role_lh == RSC_ROLE_SLAVE) {
score_multiplier = -1;
}
rsc_lh->priority = merge_weights(score_multiplier * constraint->score, rsc_lh->priority);
}
static void
colocation_match(pe_resource_t * rsc_lh, pe_resource_t * rsc_rh, rsc_colocation_t * constraint)
{
const char *tmp = NULL;
const char *value = NULL;
const char *attribute = CRM_ATTR_ID;
GHashTable *work = NULL;
gboolean do_check = FALSE;
GHashTableIter iter;
pe_node_t *node = NULL;
if (constraint->score == 0) {
return;
}
if (constraint->node_attribute != NULL) {
attribute = constraint->node_attribute;
}
if (rsc_rh->allocated_to) {
value = pe_node_attribute_raw(rsc_rh->allocated_to, attribute);
do_check = TRUE;
} else if (constraint->score < 0) {
/* nothing to do:
* anti-colocation with something that is not running
*/
return;
}
work = node_hash_dup(rsc_lh->allowed_nodes);
g_hash_table_iter_init(&iter, work);
while (g_hash_table_iter_next(&iter, NULL, (void **)&node)) {
tmp = pe_node_attribute_raw(node, attribute);
if (do_check && safe_str_eq(tmp, value)) {
if (constraint->score < INFINITY) {
pe_rsc_trace(rsc_lh, "%s: %s.%s += %d", constraint->id, rsc_lh->id,
node->details->uname, constraint->score);
node->weight = merge_weights(constraint->score, node->weight);
}
} else if (do_check == FALSE || constraint->score >= INFINITY) {
pe_rsc_trace(rsc_lh, "%s: %s.%s -= %d (%s)", constraint->id, rsc_lh->id,
node->details->uname, constraint->score,
do_check ? "failed" : "unallocated");
node->weight = merge_weights(-constraint->score, node->weight);
}
}
if (can_run_any(work)
|| constraint->score <= -INFINITY || constraint->score >= INFINITY) {
g_hash_table_destroy(rsc_lh->allowed_nodes);
rsc_lh->allowed_nodes = work;
work = NULL;
} else {
static char score[33];
score2char_stack(constraint->score, score, sizeof(score));
pe_rsc_info(rsc_lh, "%s: Rolling back scores from %s (%d, %s)",
rsc_lh->id, rsc_rh->id, do_check, score);
}
if (work) {
g_hash_table_destroy(work);
}
}
void
native_rsc_colocation_rh(pe_resource_t *rsc_lh, pe_resource_t *rsc_rh,
rsc_colocation_t *constraint,
pe_working_set_t *data_set)
{
enum filter_colocation_res filter_results;
CRM_ASSERT(rsc_lh);
CRM_ASSERT(rsc_rh);
filter_results = filter_colocation_constraint(rsc_lh, rsc_rh, constraint, FALSE);
pe_rsc_trace(rsc_lh, "%sColocating %s with %s (%s, weight=%d, filter=%d)",
constraint->score >= 0 ? "" : "Anti-",
rsc_lh->id, rsc_rh->id, constraint->id, constraint->score, filter_results);
switch (filter_results) {
case influence_rsc_priority:
influence_priority(rsc_lh, rsc_rh, constraint);
break;
case influence_rsc_location:
pe_rsc_trace(rsc_lh, "%sColocating %s with %s (%s, weight=%d)",
constraint->score >= 0 ? "" : "Anti-",
rsc_lh->id, rsc_rh->id, constraint->id, constraint->score);
colocation_match(rsc_lh, rsc_rh, constraint);
break;
case influence_nothing:
default:
return;
}
}
static gboolean
filter_rsc_ticket(pe_resource_t * rsc_lh, rsc_ticket_t * rsc_ticket)
{
if (rsc_ticket->role_lh != RSC_ROLE_UNKNOWN && rsc_ticket->role_lh != rsc_lh->role) {
pe_rsc_trace(rsc_lh, "LH: Skipping constraint: \"%s\" state filter",
role2text(rsc_ticket->role_lh));
return FALSE;
}
return TRUE;
}
void
rsc_ticket_constraint(pe_resource_t * rsc_lh, rsc_ticket_t * rsc_ticket, pe_working_set_t * data_set)
{
if (rsc_ticket == NULL) {
pe_err("rsc_ticket was NULL");
return;
}
if (rsc_lh == NULL) {
pe_err("rsc_lh was NULL for %s", rsc_ticket->id);
return;
}
if (rsc_ticket->ticket->granted && rsc_ticket->ticket->standby == FALSE) {
return;
}
if (rsc_lh->children) {
GListPtr gIter = rsc_lh->children;
pe_rsc_trace(rsc_lh, "Processing ticket dependencies from %s", rsc_lh->id);
for (; gIter != NULL; gIter = gIter->next) {
pe_resource_t *child_rsc = (pe_resource_t *) gIter->data;
rsc_ticket_constraint(child_rsc, rsc_ticket, data_set);
}
return;
}
pe_rsc_trace(rsc_lh, "%s: Processing ticket dependency on %s (%s, %s)",
rsc_lh->id, rsc_ticket->ticket->id, rsc_ticket->id,
role2text(rsc_ticket->role_lh));
if ((rsc_ticket->ticket->granted == FALSE)
&& (rsc_lh->running_on != NULL)) {
GListPtr gIter = NULL;
switch (rsc_ticket->loss_policy) {
case loss_ticket_stop:
resource_location(rsc_lh, NULL, -INFINITY, "__loss_of_ticket__", data_set);
break;
case loss_ticket_demote:
// Promotion score will be set to -INFINITY in promotion_order()
if (rsc_ticket->role_lh != RSC_ROLE_MASTER) {
resource_location(rsc_lh, NULL, -INFINITY, "__loss_of_ticket__", data_set);
}
break;
case loss_ticket_fence:
if (filter_rsc_ticket(rsc_lh, rsc_ticket) == FALSE) {
return;
}
resource_location(rsc_lh, NULL, -INFINITY, "__loss_of_ticket__", data_set);
for (gIter = rsc_lh->running_on; gIter != NULL; gIter = gIter->next) {
pe_node_t *node = (pe_node_t *) gIter->data;
- pe_fence_node(data_set, node, "deadman ticket was lost");
+ pe_fence_node(data_set, node, "deadman ticket was lost", FALSE);
}
break;
case loss_ticket_freeze:
if (filter_rsc_ticket(rsc_lh, rsc_ticket) == FALSE) {
return;
}
if (rsc_lh->running_on != NULL) {
clear_bit(rsc_lh->flags, pe_rsc_managed);
set_bit(rsc_lh->flags, pe_rsc_block);
}
break;
}
} else if (rsc_ticket->ticket->granted == FALSE) {
if (rsc_ticket->role_lh != RSC_ROLE_MASTER || rsc_ticket->loss_policy == loss_ticket_stop) {
resource_location(rsc_lh, NULL, -INFINITY, "__no_ticket__", data_set);
}
} else if (rsc_ticket->ticket->standby) {
if (rsc_ticket->role_lh != RSC_ROLE_MASTER || rsc_ticket->loss_policy == loss_ticket_stop) {
resource_location(rsc_lh, NULL, -INFINITY, "__ticket_standby__", data_set);
}
}
}
enum pe_action_flags
native_action_flags(pe_action_t * action, pe_node_t * node)
{
return action->flags;
}
static inline bool
is_primitive_action(pe_action_t *action)
{
return action && action->rsc && (action->rsc->variant == pe_native);
}
/*!
* \internal
* \brief Set action bits appropriately when pe_restart_order is used
*
* \param[in] first 'First' action in an ordering with pe_restart_order
* \param[in] then 'Then' action in an ordering with pe_restart_order
* \param[in] filter What ordering flags to care about
*
* \note pe_restart_order is set for "stop resource before starting it" and
* "stop later group member before stopping earlier group member"
*/
static void
handle_restart_ordering(pe_action_t *first, pe_action_t *then,
enum pe_action_flags filter)
{
const char *reason = NULL;
CRM_ASSERT(is_primitive_action(first));
CRM_ASSERT(is_primitive_action(then));
// We need to update the action in two cases:
// ... if 'then' is required
if (is_set(filter, pe_action_optional)
&& is_not_set(then->flags, pe_action_optional)) {
reason = "restart";
}
/* ... if 'then' is unrunnable start of managed resource (if a resource
* should restart but can't start, we still want to stop)
*/
if (is_set(filter, pe_action_runnable)
&& is_not_set(then->flags, pe_action_runnable)
&& is_set(then->rsc->flags, pe_rsc_managed)
&& safe_str_eq(then->task, RSC_START)) {
reason = "stop";
}
if (reason == NULL) {
return;
}
pe_rsc_trace(first->rsc, "Handling %s -> %s for %s",
first->uuid, then->uuid, reason);
// Make 'first' required if it is runnable
if (is_set(first->flags, pe_action_runnable)) {
pe_action_implies(first, then, pe_action_optional);
}
// Make 'first' required if 'then' is required
if (is_not_set(then->flags, pe_action_optional)) {
pe_action_implies(first, then, pe_action_optional);
}
// Make 'first' unmigratable if 'then' is unmigratable
if (is_not_set(then->flags, pe_action_migrate_runnable)) {
pe_action_implies(first, then, pe_action_migrate_runnable);
}
// Make 'then' unrunnable if 'first' is required but unrunnable
if (is_not_set(first->flags, pe_action_optional)
&& is_not_set(first->flags, pe_action_runnable)) {
pe_action_implies(then, first, pe_action_runnable);
}
}
enum pe_graph_flags
native_update_actions(pe_action_t *first, pe_action_t *then, pe_node_t *node,
enum pe_action_flags flags, enum pe_action_flags filter,
enum pe_ordering type, pe_working_set_t *data_set)
{
/* flags == get_action_flags(first, then_node) called from update_action() */
enum pe_graph_flags changed = pe_graph_none;
enum pe_action_flags then_flags = then->flags;
enum pe_action_flags first_flags = first->flags;
crm_trace( "Testing %s on %s (0x%.6x) with %s 0x%.6x",
first->uuid, first->node ? first->node->details->uname : "[none]",
first->flags, then->uuid, then->flags);
if (type & pe_order_asymmetrical) {
pe_resource_t *then_rsc = then->rsc;
enum rsc_role_e then_rsc_role = then_rsc ? then_rsc->fns->state(then_rsc, TRUE) : 0;
if (!then_rsc) {
/* ignore */
} else if ((then_rsc_role == RSC_ROLE_STOPPED) && safe_str_eq(then->task, RSC_STOP)) {
/* ignore... if 'then' is supposed to be stopped after 'first', but
* then is already stopped, there is nothing to be done when non-symmetrical. */
} else if ((then_rsc_role >= RSC_ROLE_STARTED)
&& safe_str_eq(then->task, RSC_START)
&& is_set(then->flags, pe_action_optional)
&& then->node
&& pcmk__list_of_1(then_rsc->running_on)
&& then->node->details == ((pe_node_t *) then_rsc->running_on->data)->details) {
/* Ignore. If 'then' is supposed to be started after 'first', but
* 'then' is already started, there is nothing to be done when
* asymmetrical -- unless the start is mandatory, which indicates
* the resource is restarting, and the ordering is still needed.
*/
} else if (!(first->flags & pe_action_runnable)) {
/* prevent 'then' action from happening if 'first' is not runnable and
* 'then' has not yet occurred. */
pe_action_implies(then, first, pe_action_optional);
pe_action_implies(then, first, pe_action_runnable);
pe_rsc_trace(then->rsc, "Unset optional and runnable on %s", then->uuid);
} else {
/* ignore... then is allowed to start/stop if it wants to. */
}
}
if (type & pe_order_implies_first) {
if (is_set(filter, pe_action_optional) && is_not_set(flags /* Should be then_flags? */, pe_action_optional)) {
// Needs is_set(first_flags, pe_action_optional) too?
pe_rsc_trace(first->rsc, "Unset optional on %s because of %s", first->uuid, then->uuid);
pe_action_implies(first, then, pe_action_optional);
}
if (is_set(flags, pe_action_migrate_runnable) &&
is_set(then->flags, pe_action_migrate_runnable) == FALSE &&
is_set(then->flags, pe_action_optional) == FALSE) {
pe_rsc_trace(first->rsc, "Unset migrate runnable on %s because of %s",
first->uuid, then->uuid);
pe_action_implies(first, then, pe_action_migrate_runnable);
}
}
if (type & pe_order_implies_first_master) {
if ((filter & pe_action_optional) &&
((then->flags & pe_action_optional) == FALSE) &&
then->rsc && (then->rsc->role == RSC_ROLE_MASTER)) {
pe_action_implies(first, then, pe_action_optional);
if (is_set(first->flags, pe_action_migrate_runnable) &&
is_set(then->flags, pe_action_migrate_runnable) == FALSE) {
pe_rsc_trace(first->rsc, "Unset migrate runnable on %s because of %s", first->uuid, then->uuid);
pe_action_implies(first, then, pe_action_migrate_runnable);
}
pe_rsc_trace(then->rsc, "Unset optional on %s because of %s", first->uuid, then->uuid);
}
}
if ((type & pe_order_implies_first_migratable)
&& is_set(filter, pe_action_optional)) {
if (((then->flags & pe_action_migrate_runnable) == FALSE) ||
((then->flags & pe_action_runnable) == FALSE)) {
pe_rsc_trace(then->rsc, "Unset runnable on %s because %s is neither runnable or migratable", first->uuid, then->uuid);
pe_action_implies(first, then, pe_action_runnable);
}
if ((then->flags & pe_action_optional) == 0) {
pe_rsc_trace(then->rsc, "Unset optional on %s because %s is not optional", first->uuid, then->uuid);
pe_action_implies(first, then, pe_action_optional);
}
}
if ((type & pe_order_pseudo_left)
&& is_set(filter, pe_action_optional)) {
if ((first->flags & pe_action_runnable) == FALSE) {
pe_action_implies(then, first, pe_action_migrate_runnable);
pe_clear_action_bit(then, pe_action_pseudo);
pe_rsc_trace(then->rsc, "Unset pseudo on %s because %s is not runnable", then->uuid, first->uuid);
}
}
if (is_set(type, pe_order_runnable_left)
&& is_set(filter, pe_action_runnable)
&& is_set(then->flags, pe_action_runnable)
&& is_set(flags, pe_action_runnable) == FALSE) {
pe_rsc_trace(then->rsc, "Unset runnable on %s because of %s", then->uuid, first->uuid);
pe_action_implies(then, first, pe_action_runnable);
pe_action_implies(then, first, pe_action_migrate_runnable);
}
if (is_set(type, pe_order_implies_then)
&& is_set(filter, pe_action_optional)
&& is_set(then->flags, pe_action_optional)
&& is_set(flags, pe_action_optional) == FALSE) {
/* in this case, treat migrate_runnable as if first is optional */
if (is_set(first->flags, pe_action_migrate_runnable) == FALSE) {
pe_rsc_trace(then->rsc, "Unset optional on %s because of %s", then->uuid, first->uuid);
pe_action_implies(then, first, pe_action_optional);
}
}
if (is_set(type, pe_order_restart)) {
handle_restart_ordering(first, then, filter);
}
if (then_flags != then->flags) {
changed |= pe_graph_updated_then;
pe_rsc_trace(then->rsc,
"Then: Flags for %s on %s are now 0x%.6x (was 0x%.6x) because of %s 0x%.6x",
then->uuid, then->node ? then->node->details->uname : "[none]", then->flags,
then_flags, first->uuid, first->flags);
if(then->rsc && then->rsc->parent) {
/* "X_stop then X_start" doesn't get handled for cloned groups unless we do this */
update_action(then, data_set);
}
}
if (first_flags != first->flags) {
changed |= pe_graph_updated_first;
pe_rsc_trace(first->rsc,
"First: Flags for %s on %s are now 0x%.6x (was 0x%.6x) because of %s 0x%.6x",
first->uuid, first->node ? first->node->details->uname : "[none]",
first->flags, first_flags, then->uuid, then->flags);
}
return changed;
}
void
native_rsc_location(pe_resource_t *rsc, pe__location_t *constraint)
{
GListPtr gIter = NULL;
GHashTableIter iter;
pe_node_t *node = NULL;
if (constraint == NULL) {
pe_err("Constraint is NULL");
return;
} else if (rsc == NULL) {
pe_err("LHS of rsc_to_node (%s) is NULL", constraint->id);
return;
}
pe_rsc_trace(rsc, "Applying %s (%s) to %s", constraint->id,
role2text(constraint->role_filter), rsc->id);
/* take "lifetime" into account */
if (constraint->role_filter > RSC_ROLE_UNKNOWN && constraint->role_filter != rsc->next_role) {
pe_rsc_debug(rsc, "Constraint (%s) is not active (role : %s vs. %s)",
constraint->id, role2text(constraint->role_filter), role2text(rsc->next_role));
return;
}
if (constraint->node_list_rh == NULL) {
pe_rsc_trace(rsc, "RHS of constraint %s is NULL", constraint->id);
return;
}
for (gIter = constraint->node_list_rh; gIter != NULL; gIter = gIter->next) {
pe_node_t *node = (pe_node_t *) gIter->data;
pe_node_t *other_node = NULL;
other_node = (pe_node_t *) pe_hash_table_lookup(rsc->allowed_nodes, node->details->id);
if (other_node != NULL) {
pe_rsc_trace(rsc, "%s + %s: %d + %d",
node->details->uname,
other_node->details->uname, node->weight, other_node->weight);
other_node->weight = merge_weights(other_node->weight, node->weight);
} else {
other_node = node_copy(node);
pe_rsc_trace(rsc, "%s: %d (insert %d)", other_node->details->uname, other_node->weight, constraint->discover_mode);
g_hash_table_insert(rsc->allowed_nodes, (gpointer) other_node->details->id, other_node);
}
if (other_node->rsc_discover_mode < constraint->discover_mode) {
if (constraint->discover_mode == pe_discover_exclusive) {
rsc->exclusive_discover = TRUE;
}
/* exclusive > never > always... always is default */
other_node->rsc_discover_mode = constraint->discover_mode;
}
}
g_hash_table_iter_init(&iter, rsc->allowed_nodes);
while (g_hash_table_iter_next(&iter, NULL, (void **)&node)) {
pe_rsc_trace(rsc, "%s + %s : %d", rsc->id, node->details->uname, node->weight);
}
}
void
native_expand(pe_resource_t * rsc, pe_working_set_t * data_set)
{
GListPtr gIter = NULL;
CRM_ASSERT(rsc);
pe_rsc_trace(rsc, "Processing actions from %s", rsc->id);
for (gIter = rsc->actions; gIter != NULL; gIter = gIter->next) {
pe_action_t *action = (pe_action_t *) gIter->data;
crm_trace("processing action %d for rsc=%s", action->id, rsc->id);
graph_element_from_action(action, data_set);
}
for (gIter = rsc->children; gIter != NULL; gIter = gIter->next) {
pe_resource_t *child_rsc = (pe_resource_t *) gIter->data;
child_rsc->cmds->expand(child_rsc, data_set);
}
}
#define log_change(a, fmt, args...) do { \
if(a && a->reason && terminal) { \
printf(" * "fmt" \tdue to %s\n", ##args, a->reason); \
} else if(a && a->reason) { \
crm_notice(fmt" \tdue to %s", ##args, a->reason); \
} else if(terminal) { \
printf(" * "fmt"\n", ##args); \
} else { \
crm_notice(fmt, ##args); \
} \
} while(0)
#define STOP_SANITY_ASSERT(lineno) do { \
if(current && current->details->unclean) { \
/* It will be a pseudo op */ \
} else if(stop == NULL) { \
crm_err("%s:%d: No stop action exists for %s", __FUNCTION__, lineno, rsc->id); \
CRM_ASSERT(stop != NULL); \
} else if(is_set(stop->flags, pe_action_optional)) { \
crm_err("%s:%d: Action %s is still optional", __FUNCTION__, lineno, stop->uuid); \
CRM_ASSERT(is_not_set(stop->flags, pe_action_optional)); \
} \
} while(0)
static int rsc_width = 5;
static int detail_width = 5;
static void
LogAction(const char *change, pe_resource_t *rsc, pe_node_t *origin, pe_node_t *destination, pe_action_t *action, pe_action_t *source, gboolean terminal)
{
int len = 0;
char *reason = NULL;
char *details = NULL;
bool same_host = FALSE;
bool same_role = FALSE;
bool need_role = FALSE;
CRM_ASSERT(action);
CRM_ASSERT(destination != NULL || origin != NULL);
if(source == NULL) {
source = action;
}
len = strlen(rsc->id);
if(len > rsc_width) {
rsc_width = len + 2;
}
if(rsc->role > RSC_ROLE_STARTED || rsc->next_role > RSC_ROLE_SLAVE) {
need_role = TRUE;
}
if(origin != NULL && destination != NULL && origin->details == destination->details) {
same_host = TRUE;
}
if(rsc->role == rsc->next_role) {
same_role = TRUE;
}
if(need_role && origin == NULL) {
/* Promoting from Stopped */
details = crm_strdup_printf("%s -> %s %s", role2text(rsc->role), role2text(rsc->next_role), destination->details->uname);
} else if(need_role && destination == NULL) {
/* Demoting a Master or Stopping a Slave */
details = crm_strdup_printf("%s %s", role2text(rsc->role), origin->details->uname);
} else if(origin == NULL || destination == NULL) {
/* Starting or stopping a resource */
details = crm_strdup_printf("%s", origin?origin->details->uname:destination->details->uname);
} else if(need_role && same_role && same_host) {
/* Recovering or restarting a promotable clone resource */
details = crm_strdup_printf("%s %s", role2text(rsc->role), origin->details->uname);
} else if(same_role && same_host) {
/* Recovering or Restarting a normal resource */
details = crm_strdup_printf("%s", origin->details->uname);
} else if(same_role && need_role) {
/* Moving a promotable clone resource */
details = crm_strdup_printf("%s -> %s %s", origin->details->uname, destination->details->uname, role2text(rsc->role));
} else if(same_role) {
/* Moving a normal resource */
details = crm_strdup_printf("%s -> %s", origin->details->uname, destination->details->uname);
} else if(same_host) {
/* Promoting or demoting a promotable clone resource */
details = crm_strdup_printf("%s -> %s %s", role2text(rsc->role), role2text(rsc->next_role), origin->details->uname);
} else {
/* Moving and promoting/demoting */
details = crm_strdup_printf("%s %s -> %s %s", role2text(rsc->role), origin->details->uname, role2text(rsc->next_role), destination->details->uname);
}
len = strlen(details);
if(len > detail_width) {
detail_width = len;
}
if(source->reason && is_not_set(action->flags, pe_action_runnable)) {
reason = crm_strdup_printf(" due to %s (blocked)", source->reason);
} else if(source->reason) {
reason = crm_strdup_printf(" due to %s", source->reason);
} else if(is_not_set(action->flags, pe_action_runnable)) {
reason = strdup(" blocked");
} else {
reason = strdup("");
}
if(terminal) {
printf(" * %-8s %-*s ( %*s ) %s\n", change, rsc_width, rsc->id, detail_width, details, reason);
} else {
crm_notice(" * %-8s %-*s ( %*s ) %s", change, rsc_width, rsc->id, detail_width, details, reason);
}
free(details);
free(reason);
}
void
LogActions(pe_resource_t * rsc, pe_working_set_t * data_set, gboolean terminal)
{
pe_node_t *next = NULL;
pe_node_t *current = NULL;
pe_node_t *start_node = NULL;
pe_action_t *stop = NULL;
pe_action_t *start = NULL;
pe_action_t *demote = NULL;
pe_action_t *promote = NULL;
char *key = NULL;
gboolean moving = FALSE;
GListPtr possible_matches = NULL;
if(rsc->variant == pe_container) {
pcmk__bundle_log_actions(rsc, data_set, terminal);
return;
}
if (rsc->children) {
GListPtr gIter = NULL;
for (gIter = rsc->children; gIter != NULL; gIter = gIter->next) {
pe_resource_t *child_rsc = (pe_resource_t *) gIter->data;
LogActions(child_rsc, data_set, terminal);
}
return;
}
next = rsc->allocated_to;
if (rsc->running_on) {
current = pe__current_node(rsc);
if (rsc->role == RSC_ROLE_STOPPED) {
/*
* This can occur when resources are being recovered
* We fiddle with the current role in native_create_actions()
*/
rsc->role = RSC_ROLE_STARTED;
}
}
if (current == NULL && is_set(rsc->flags, pe_rsc_orphan)) {
/* Don't log stopped orphans */
return;
}
if (is_not_set(rsc->flags, pe_rsc_managed)
|| (current == NULL && next == NULL)) {
pe_rsc_info(rsc, "Leave %s\t(%s%s)",
rsc->id, role2text(rsc->role), is_not_set(rsc->flags,
pe_rsc_managed) ? " unmanaged" : "");
return;
}
if (current != NULL && next != NULL && safe_str_neq(current->details->id, next->details->id)) {
moving = TRUE;
}
possible_matches = pe__resource_actions(rsc, next, RSC_START, FALSE);
if (possible_matches) {
start = possible_matches->data;
g_list_free(possible_matches);
}
if ((start == NULL) || is_not_set(start->flags, pe_action_runnable)) {
start_node = NULL;
} else {
start_node = current;
}
possible_matches = pe__resource_actions(rsc, start_node, RSC_STOP, FALSE);
if (possible_matches) {
stop = possible_matches->data;
g_list_free(possible_matches);
}
possible_matches = pe__resource_actions(rsc, next, RSC_PROMOTE, FALSE);
if (possible_matches) {
promote = possible_matches->data;
g_list_free(possible_matches);
}
possible_matches = pe__resource_actions(rsc, next, RSC_DEMOTE, FALSE);
if (possible_matches) {
demote = possible_matches->data;
g_list_free(possible_matches);
}
if (rsc->role == rsc->next_role) {
pe_action_t *migrate_op = NULL;
possible_matches = pe__resource_actions(rsc, next, RSC_MIGRATED, FALSE);
if (possible_matches) {
migrate_op = possible_matches->data;
}
CRM_CHECK(next != NULL,);
if (next == NULL) {
} else if (migrate_op && is_set(migrate_op->flags, pe_action_runnable) && current) {
LogAction("Migrate", rsc, current, next, start, NULL, terminal);
} else if (is_set(rsc->flags, pe_rsc_reload)) {
LogAction("Reload", rsc, current, next, start, NULL, terminal);
} else if (start == NULL || is_set(start->flags, pe_action_optional)) {
pe_rsc_info(rsc, "Leave %s\t(%s %s)", rsc->id, role2text(rsc->role),
next->details->uname);
} else if (start && is_set(start->flags, pe_action_runnable) == FALSE) {
LogAction("Stop", rsc, current, NULL, stop,
(stop && stop->reason)? stop : start, terminal);
STOP_SANITY_ASSERT(__LINE__);
} else if (moving && current) {
LogAction(is_set(rsc->flags, pe_rsc_failed) ? "Recover" : "Move",
rsc, current, next, stop, NULL, terminal);
} else if (is_set(rsc->flags, pe_rsc_failed)) {
LogAction("Recover", rsc, current, NULL, stop, NULL, terminal);
STOP_SANITY_ASSERT(__LINE__);
} else {
LogAction("Restart", rsc, current, next, start, NULL, terminal);
/* STOP_SANITY_ASSERT(__LINE__); False positive for migrate-fail-7 */
}
g_list_free(possible_matches);
return;
}
if(stop
&& (rsc->next_role == RSC_ROLE_STOPPED
|| (start && is_not_set(start->flags, pe_action_runnable)))) {
GListPtr gIter = NULL;
key = stop_key(rsc);
for (gIter = rsc->running_on; gIter != NULL; gIter = gIter->next) {
pe_node_t *node = (pe_node_t *) gIter->data;
pe_action_t *stop_op = NULL;
possible_matches = find_actions(rsc->actions, key, node);
if (possible_matches) {
stop_op = possible_matches->data;
g_list_free(possible_matches);
}
if (stop_op && (stop_op->flags & pe_action_runnable)) {
STOP_SANITY_ASSERT(__LINE__);
}
LogAction("Stop", rsc, node, NULL, stop_op,
(stop_op && stop_op->reason)? stop_op : start, terminal);
}
free(key);
} else if (stop && is_set(rsc->flags, pe_rsc_failed)) {
/* 'stop' may be NULL if the failure was ignored */
LogAction("Recover", rsc, current, next, stop, start, terminal);
STOP_SANITY_ASSERT(__LINE__);
} else if (moving) {
LogAction("Move", rsc, current, next, stop, NULL, terminal);
STOP_SANITY_ASSERT(__LINE__);
} else if (is_set(rsc->flags, pe_rsc_reload)) {
LogAction("Reload", rsc, current, next, start, NULL, terminal);
} else if (stop != NULL && is_not_set(stop->flags, pe_action_optional)) {
LogAction("Restart", rsc, current, next, start, NULL, terminal);
STOP_SANITY_ASSERT(__LINE__);
} else if (rsc->role == RSC_ROLE_MASTER) {
CRM_LOG_ASSERT(current != NULL);
LogAction("Demote", rsc, current, next, demote, NULL, terminal);
} else if(rsc->next_role == RSC_ROLE_MASTER) {
CRM_LOG_ASSERT(next);
LogAction("Promote", rsc, current, next, promote, NULL, terminal);
} else if (rsc->role == RSC_ROLE_STOPPED && rsc->next_role > RSC_ROLE_STOPPED) {
LogAction("Start", rsc, current, next, start, NULL, terminal);
}
}
gboolean
StopRsc(pe_resource_t * rsc, pe_node_t * next, gboolean optional, pe_working_set_t * data_set)
{
GListPtr gIter = NULL;
CRM_ASSERT(rsc);
pe_rsc_trace(rsc, "%s", rsc->id);
for (gIter = rsc->running_on; gIter != NULL; gIter = gIter->next) {
pe_node_t *current = (pe_node_t *) gIter->data;
pe_action_t *stop;
if (rsc->partial_migration_target) {
if (rsc->partial_migration_target->details == current->details) {
pe_rsc_trace(rsc, "Filtered %s -> %s %s", current->details->uname,
next->details->uname, rsc->id);
continue;
} else {
pe_rsc_trace(rsc, "Forced on %s %s", current->details->uname, rsc->id);
optional = FALSE;
}
}
pe_rsc_trace(rsc, "%s on %s", rsc->id, current->details->uname);
stop = stop_action(rsc, current, optional);
if(rsc->allocated_to == NULL) {
pe_action_set_reason(stop, "node availability", TRUE);
}
if (is_not_set(rsc->flags, pe_rsc_managed)) {
update_action_flags(stop, pe_action_runnable | pe_action_clear, __FUNCTION__, __LINE__);
}
if (is_set(data_set->flags, pe_flag_remove_after_stop)) {
DeleteRsc(rsc, current, optional, data_set);
}
if(is_set(rsc->flags, pe_rsc_needs_unfencing)) {
- pe_action_t *unfence = pe_fence_op(current, "on", TRUE, NULL, data_set);
+ pe_action_t *unfence = pe_fence_op(current, "on", TRUE, NULL, FALSE, data_set);
order_actions(stop, unfence, pe_order_implies_first);
if (!node_has_been_unfenced(current)) {
pe_proc_err("Stopping %s until %s can be unfenced", rsc->id, current->details->uname);
}
}
}
return TRUE;
}
static void
order_after_unfencing(pe_resource_t *rsc, pe_node_t *node, pe_action_t *action,
enum pe_ordering order, pe_working_set_t *data_set)
{
/* When unfencing is in use, we order unfence actions before any probe or
* start of resources that require unfencing, and also of fence devices.
*
* This might seem to violate the principle that fence devices require
* only quorum. However, fence agents that unfence often don't have enough
* information to even probe or start unless the node is first unfenced.
*/
if (is_unfence_device(rsc, data_set)
|| is_set(rsc->flags, pe_rsc_needs_unfencing)) {
/* Start with an optional ordering. Requiring unfencing would result in
* the node being unfenced, and all its resources being stopped,
* whenever a new resource is added -- which would be highly suboptimal.
*/
- pe_action_t *unfence = pe_fence_op(node, "on", TRUE, NULL, data_set);
+ pe_action_t *unfence = pe_fence_op(node, "on", TRUE, NULL, FALSE, data_set);
order_actions(unfence, action, order);
if (!node_has_been_unfenced(node)) {
// But unfencing is required if it has never been done
char *reason = crm_strdup_printf("required by %s %s",
rsc->id, action->task);
trigger_unfencing(NULL, node, reason, NULL, data_set);
free(reason);
}
}
}
gboolean
StartRsc(pe_resource_t * rsc, pe_node_t * next, gboolean optional, pe_working_set_t * data_set)
{
pe_action_t *start = NULL;
CRM_ASSERT(rsc);
pe_rsc_trace(rsc, "%s on %s %d %d", rsc->id, next ? next->details->uname : "N/A", optional, next ? next->weight : 0);
start = start_action(rsc, next, TRUE);
order_after_unfencing(rsc, next, start, pe_order_implies_then, data_set);
if (is_set(start->flags, pe_action_runnable) && optional == FALSE) {
update_action_flags(start, pe_action_optional | pe_action_clear, __FUNCTION__, __LINE__);
}
return TRUE;
}
gboolean
PromoteRsc(pe_resource_t * rsc, pe_node_t * next, gboolean optional, pe_working_set_t * data_set)
{
GListPtr gIter = NULL;
gboolean runnable = TRUE;
GListPtr action_list = NULL;
CRM_ASSERT(rsc);
CRM_CHECK(next != NULL, return FALSE);
pe_rsc_trace(rsc, "%s on %s", rsc->id, next->details->uname);
action_list = pe__resource_actions(rsc, next, RSC_START, TRUE);
for (gIter = action_list; gIter != NULL; gIter = gIter->next) {
pe_action_t *start = (pe_action_t *) gIter->data;
if (is_set(start->flags, pe_action_runnable) == FALSE) {
runnable = FALSE;
}
}
g_list_free(action_list);
if (runnable) {
promote_action(rsc, next, optional);
return TRUE;
}
pe_rsc_debug(rsc, "%s\tPromote %s (canceled)", next->details->uname, rsc->id);
action_list = pe__resource_actions(rsc, next, RSC_PROMOTE, TRUE);
for (gIter = action_list; gIter != NULL; gIter = gIter->next) {
pe_action_t *promote = (pe_action_t *) gIter->data;
update_action_flags(promote, pe_action_runnable | pe_action_clear, __FUNCTION__, __LINE__);
}
g_list_free(action_list);
return TRUE;
}
gboolean
DemoteRsc(pe_resource_t * rsc, pe_node_t * next, gboolean optional, pe_working_set_t * data_set)
{
GListPtr gIter = NULL;
CRM_ASSERT(rsc);
pe_rsc_trace(rsc, "%s", rsc->id);
/* CRM_CHECK(rsc->next_role == RSC_ROLE_SLAVE, return FALSE); */
for (gIter = rsc->running_on; gIter != NULL; gIter = gIter->next) {
pe_node_t *current = (pe_node_t *) gIter->data;
pe_rsc_trace(rsc, "%s on %s", rsc->id, next ? next->details->uname : "N/A");
demote_action(rsc, current, optional);
}
return TRUE;
}
gboolean
RoleError(pe_resource_t * rsc, pe_node_t * next, gboolean optional, pe_working_set_t * data_set)
{
CRM_ASSERT(rsc);
crm_err("%s on %s", rsc->id, next ? next->details->uname : "N/A");
CRM_CHECK(FALSE, return FALSE);
return FALSE;
}
gboolean
NullOp(pe_resource_t * rsc, pe_node_t * next, gboolean optional, pe_working_set_t * data_set)
{
CRM_ASSERT(rsc);
pe_rsc_trace(rsc, "%s", rsc->id);
return FALSE;
}
gboolean
DeleteRsc(pe_resource_t * rsc, pe_node_t * node, gboolean optional, pe_working_set_t * data_set)
{
if (is_set(rsc->flags, pe_rsc_failed)) {
pe_rsc_trace(rsc, "Resource %s not deleted from %s: failed", rsc->id, node->details->uname);
return FALSE;
} else if (node == NULL) {
pe_rsc_trace(rsc, "Resource %s not deleted: NULL node", rsc->id);
return FALSE;
} else if (node->details->unclean || node->details->online == FALSE) {
pe_rsc_trace(rsc, "Resource %s not deleted from %s: unrunnable", rsc->id,
node->details->uname);
return FALSE;
}
crm_notice("Removing %s from %s", rsc->id, node->details->uname);
delete_action(rsc, node, optional);
new_rsc_order(rsc, RSC_STOP, rsc, RSC_DELETE,
optional ? pe_order_implies_then : pe_order_optional, data_set);
new_rsc_order(rsc, RSC_DELETE, rsc, RSC_START,
optional ? pe_order_implies_then : pe_order_optional, data_set);
return TRUE;
}
gboolean
native_create_probe(pe_resource_t * rsc, pe_node_t * node, pe_action_t * complete,
gboolean force, pe_working_set_t * data_set)
{
enum pe_ordering flags = pe_order_optional;
char *key = NULL;
pe_action_t *probe = NULL;
pe_node_t *running = NULL;
pe_node_t *allowed = NULL;
pe_resource_t *top = uber_parent(rsc);
static const char *rc_master = NULL;
static const char *rc_inactive = NULL;
if (rc_inactive == NULL) {
rc_inactive = crm_itoa(PCMK_OCF_NOT_RUNNING);
rc_master = crm_itoa(PCMK_OCF_RUNNING_MASTER);
}
CRM_CHECK(node != NULL, return FALSE);
if (force == FALSE && is_not_set(data_set->flags, pe_flag_startup_probes)) {
pe_rsc_trace(rsc, "Skipping active resource detection for %s", rsc->id);
return FALSE;
}
if (pe__is_guest_or_remote_node(node)) {
const char *class = crm_element_value(rsc->xml, XML_AGENT_ATTR_CLASS);
if (safe_str_eq(class, PCMK_RESOURCE_CLASS_STONITH)) {
pe_rsc_trace(rsc,
"Skipping probe for %s on %s because Pacemaker Remote nodes cannot run stonith agents",
rsc->id, node->details->id);
return FALSE;
} else if (pe__is_guest_node(node)
&& pe__resource_contains_guest_node(data_set, rsc)) {
pe_rsc_trace(rsc,
"Skipping probe for %s on %s because guest nodes cannot run resources containing guest nodes",
rsc->id, node->details->id);
return FALSE;
} else if (rsc->is_remote_node) {
pe_rsc_trace(rsc,
"Skipping probe for %s on %s because Pacemaker Remote nodes cannot host remote connections",
rsc->id, node->details->id);
return FALSE;
}
}
if (rsc->children) {
GListPtr gIter = NULL;
gboolean any_created = FALSE;
for (gIter = rsc->children; gIter != NULL; gIter = gIter->next) {
pe_resource_t *child_rsc = (pe_resource_t *) gIter->data;
any_created = child_rsc->cmds->create_probe(child_rsc, node, complete, force, data_set)
|| any_created;
}
return any_created;
} else if ((rsc->container) && (!rsc->is_remote_node)) {
pe_rsc_trace(rsc, "Skipping %s: it is within container %s", rsc->id, rsc->container->id);
return FALSE;
}
if (is_set(rsc->flags, pe_rsc_orphan)) {
pe_rsc_trace(rsc, "Skipping orphan: %s", rsc->id);
return FALSE;
}
// Check whether resource is already known on node
if (!force && g_hash_table_lookup(rsc->known_on, node->details->id)) {
pe_rsc_trace(rsc, "Skipping known: %s on %s", rsc->id, node->details->uname);
return FALSE;
}
allowed = g_hash_table_lookup(rsc->allowed_nodes, node->details->id);
if (rsc->exclusive_discover || top->exclusive_discover) {
if (allowed == NULL) {
/* exclusive discover is enabled and this node is not in the allowed list. */
pe_rsc_trace(rsc, "Skipping probe for %s on node %s, A", rsc->id, node->details->id);
return FALSE;
} else if (allowed->rsc_discover_mode != pe_discover_exclusive) {
/* exclusive discover is enabled and this node is not marked
* as a node this resource should be discovered on */
pe_rsc_trace(rsc, "Skipping probe for %s on node %s, B", rsc->id, node->details->id);
return FALSE;
}
}
if(allowed == NULL && node->rsc_discover_mode == pe_discover_never) {
/* If this node was allowed to host this resource it would
* have been explicitly added to the 'allowed_nodes' list.
* However it wasn't and the node has discovery disabled, so
* no need to probe for this resource.
*/
pe_rsc_trace(rsc, "Skipping probe for %s on node %s, C", rsc->id, node->details->id);
return FALSE;
}
if (allowed && allowed->rsc_discover_mode == pe_discover_never) {
/* this resource is marked as not needing to be discovered on this node */
pe_rsc_trace(rsc, "Skipping probe for %s on node %s, discovery mode", rsc->id, node->details->id);
return FALSE;
}
if (pe__is_guest_node(node)) {
pe_resource_t *remote = node->details->remote_rsc->container;
if(remote->role == RSC_ROLE_STOPPED) {
/* If the container is stopped, then we know anything that
* might have been inside it is also stopped and there is
* no need to probe.
*
* If we don't know the container's state on the target
* either:
*
* - the container is running, the transition will abort
* and we'll end up in a different case next time, or
*
* - the container is stopped
*
* Either way there is no need to probe.
*
*/
if(remote->allocated_to
&& g_hash_table_lookup(remote->known_on, remote->allocated_to->details->id) == NULL) {
/* For safety, we order the 'rsc' start after 'remote'
* has been probed.
*
* Using 'top' helps for groups, but we may need to
* follow the start's ordering chain backwards.
*/
custom_action_order(remote,
pcmk__op_key(remote->id, RSC_STATUS, 0),
NULL, top,
pcmk__op_key(top->id, RSC_START, 0), NULL,
pe_order_optional, data_set);
}
pe_rsc_trace(rsc, "Skipping probe for %s on node %s, %s is stopped",
rsc->id, node->details->id, remote->id);
return FALSE;
/* Here we really we want to check if remote->stop is required,
* but that information doesn't exist yet
*/
} else if(node->details->remote_requires_reset
|| node->details->unclean
|| is_set(remote->flags, pe_rsc_failed)
|| remote->next_role == RSC_ROLE_STOPPED
|| (remote->allocated_to
&& pe_find_node(remote->running_on, remote->allocated_to->details->uname) == NULL)
) {
/* The container is stopping or restarting, don't start
* 'rsc' until 'remote' stops as this also implies that
* 'rsc' is stopped - avoiding the need to probe
*/
custom_action_order(remote, pcmk__op_key(remote->id, RSC_STOP, 0),
NULL, top, pcmk__op_key(top->id, RSC_START, 0),
NULL, pe_order_optional, data_set);
pe_rsc_trace(rsc, "Skipping probe for %s on node %s, %s is stopping, restarting or moving",
rsc->id, node->details->id, remote->id);
return FALSE;
/* } else {
* The container is running so there is no problem probing it
*/
}
}
key = pcmk__op_key(rsc->id, RSC_STATUS, 0);
probe = custom_action(rsc, key, RSC_STATUS, node, FALSE, TRUE, data_set);
update_action_flags(probe, pe_action_optional | pe_action_clear, __FUNCTION__, __LINE__);
order_after_unfencing(rsc, node, probe, pe_order_optional, data_set);
/*
* We need to know if it's running_on (not just known_on) this node
* to correctly determine the target rc.
*/
running = pe_find_node_id(rsc->running_on, node->details->id);
if (running == NULL) {
add_hash_param(probe->meta, XML_ATTR_TE_TARGET_RC, rc_inactive);
} else if (rsc->role == RSC_ROLE_MASTER) {
add_hash_param(probe->meta, XML_ATTR_TE_TARGET_RC, rc_master);
}
crm_debug("Probing %s on %s (%s) %d %p", rsc->id, node->details->uname, role2text(rsc->role),
is_set(probe->flags, pe_action_runnable), rsc->running_on);
if (is_unfence_device(rsc, data_set) || !pe_rsc_is_clone(top)) {
top = rsc;
} else {
crm_trace("Probing %s on %s (%s) as %s", rsc->id, node->details->uname, role2text(rsc->role), top->id);
}
if(is_not_set(probe->flags, pe_action_runnable) && rsc->running_on == NULL) {
/* Prevent the start from occurring if rsc isn't active, but
* don't cause it to stop if it was active already
*/
flags |= pe_order_runnable_left;
}
custom_action_order(rsc, NULL, probe,
top, pcmk__op_key(top->id, RSC_START, 0), NULL,
flags, data_set);
/* Before any reloads, if they exist */
custom_action_order(rsc, NULL, probe,
top, reload_key(rsc), NULL,
pe_order_optional, data_set);
#if 0
// complete is always null currently
if (!is_unfence_device(rsc, data_set)) {
/* Normally rsc.start depends on probe complete which depends
* on rsc.probe. But this can't be the case for fence devices
* with unfencing, as it would create graph loops.
*
* So instead we explicitly order 'rsc.probe then rsc.start'
*/
order_actions(probe, complete, pe_order_implies_then);
}
#endif
return TRUE;
}
/*!
* \internal
* \brief Check whether a resource is known on a particular node
*
* \param[in] rsc Resource to check
* \param[in] node Node to check
*
* \return TRUE if resource (or parent if an anonymous clone) is known
*/
static bool
rsc_is_known_on(pe_resource_t *rsc, const pe_node_t *node)
{
if (pe_hash_table_lookup(rsc->known_on, node->details->id)) {
return TRUE;
} else if ((rsc->variant == pe_native)
&& pe_rsc_is_anon_clone(rsc->parent)
&& pe_hash_table_lookup(rsc->parent->known_on, node->details->id)) {
/* We check only the parent, not the uber-parent, because we cannot
* assume that the resource is known if it is in an anonymously cloned
* group (which may be only partially known).
*/
return TRUE;
}
return FALSE;
}
/*!
* \internal
* \brief Order a resource's start and promote actions relative to fencing
*
* \param[in] rsc Resource to be ordered
* \param[in] stonith_op Fence action
* \param[in] data_set Cluster information
*/
static void
native_start_constraints(pe_resource_t * rsc, pe_action_t * stonith_op, pe_working_set_t * data_set)
{
pe_node_t *target;
GListPtr gIter = NULL;
CRM_CHECK(stonith_op && stonith_op->node, return);
target = stonith_op->node;
for (gIter = rsc->actions; gIter != NULL; gIter = gIter->next) {
pe_action_t *action = (pe_action_t *) gIter->data;
switch (action->needs) {
case rsc_req_nothing:
// Anything other than start or promote requires nothing
break;
case rsc_req_stonith:
order_actions(stonith_op, action, pe_order_optional);
break;
case rsc_req_quorum:
if (safe_str_eq(action->task, RSC_START)
&& pe_hash_table_lookup(rsc->allowed_nodes, target->details->id)
&& !rsc_is_known_on(rsc, target)) {
/* If we don't know the status of the resource on the node
* we're about to shoot, we have to assume it may be active
* there. Order the resource start after the fencing. This
* is analogous to waiting for all the probes for a resource
* to complete before starting it.
*
* The most likely explanation is that the DC died and took
* its status with it.
*/
pe_rsc_debug(rsc, "Ordering %s after %s recovery", action->uuid,
target->details->uname);
order_actions(stonith_op, action,
pe_order_optional | pe_order_runnable_left);
}
break;
}
}
}
static void
native_stop_constraints(pe_resource_t * rsc, pe_action_t * stonith_op, pe_working_set_t * data_set)
{
GListPtr gIter = NULL;
GListPtr action_list = NULL;
bool order_implicit = false;
pe_resource_t *top = uber_parent(rsc);
pe_action_t *parent_stop = NULL;
pe_node_t *target;
CRM_CHECK(stonith_op && stonith_op->node, return);
target = stonith_op->node;
/* Get a list of stop actions potentially implied by the fencing */
action_list = pe__resource_actions(rsc, target, RSC_STOP, FALSE);
/* If resource requires fencing, implicit actions must occur after fencing.
*
* Implied stops and demotes of resources running on guest nodes are always
* ordered after fencing, even if the resource does not require fencing,
* because guest node "fencing" is actually just a resource stop.
*/
if (is_set(rsc->flags, pe_rsc_needs_fencing) || pe__is_guest_node(target)) {
order_implicit = true;
}
if (action_list && order_implicit) {
parent_stop = find_first_action(top->actions, NULL, RSC_STOP, NULL);
}
for (gIter = action_list; gIter != NULL; gIter = gIter->next) {
pe_action_t *action = (pe_action_t *) gIter->data;
// The stop would never complete, so convert it into a pseudo-action.
update_action_flags(action, pe_action_pseudo|pe_action_runnable,
__FUNCTION__, __LINE__);
if (order_implicit) {
update_action_flags(action, pe_action_implied_by_stonith,
__FUNCTION__, __LINE__);
/* Order the stonith before the parent stop (if any).
*
* Also order the stonith before the resource stop, unless the
* resource is inside a bundle -- that would cause a graph loop.
* We can rely on the parent stop's ordering instead.
*
* User constraints must not order a resource in a guest node
* relative to the guest node container resource. The
* pe_order_preserve flag marks constraints as generated by the
* cluster and thus immune to that check (and is irrelevant if
* target is not a guest).
*/
if (!pe_rsc_is_bundled(rsc)) {
order_actions(stonith_op, action, pe_order_preserve);
}
order_actions(stonith_op, parent_stop, pe_order_preserve);
}
if (is_set(rsc->flags, pe_rsc_failed)) {
crm_notice("Stop of failed resource %s is implicit %s %s is fenced",
rsc->id, (order_implicit? "after" : "because"),
target->details->uname);
} else {
crm_info("%s is implicit %s %s is fenced",
action->uuid, (order_implicit? "after" : "because"),
target->details->uname);
}
if (is_set(rsc->flags, pe_rsc_notify)) {
/* Create a second notification that will be delivered
* immediately after the node is fenced
*
* Basic problem:
* - C is a clone active on the node to be shot and stopping on another
* - R is a resource that depends on C
*
* + C.stop depends on R.stop
* + C.stopped depends on STONITH
* + C.notify depends on C.stopped
* + C.healthy depends on C.notify
* + R.stop depends on C.healthy
*
* The extra notification here changes
* + C.healthy depends on C.notify
* into:
* + C.healthy depends on C.notify'
* + C.notify' depends on STONITH'
* thus breaking the loop
*/
create_secondary_notification(action, rsc, stonith_op, data_set);
}
/* From Bug #1601, successful fencing must be an input to a failed resources stop action.
However given group(rA, rB) running on nodeX and B.stop has failed,
A := stop healthy resource (rA.stop)
B := stop failed resource (pseudo operation B.stop)
C := stonith nodeX
A requires B, B requires C, C requires A
This loop would prevent the cluster from making progress.
This block creates the "C requires A" dependency and therefore must (at least
for now) be disabled.
Instead, run the block above and treat all resources on nodeX as B would be
(marked as a pseudo op depending on the STONITH).
TODO: Break the "A requires B" dependency in update_action() and re-enable this block
} else if(is_stonith == FALSE) {
crm_info("Moving healthy resource %s"
" off %s before fencing",
rsc->id, node->details->uname);
* stop healthy resources before the
* stonith op
*
custom_action_order(
rsc, stop_key(rsc), NULL,
NULL,strdup(CRM_OP_FENCE),stonith_op,
pe_order_optional, data_set);
*/
}
g_list_free(action_list);
/* Get a list of demote actions potentially implied by the fencing */
action_list = pe__resource_actions(rsc, target, RSC_DEMOTE, FALSE);
for (gIter = action_list; gIter != NULL; gIter = gIter->next) {
pe_action_t *action = (pe_action_t *) gIter->data;
if (action->node->details->online == FALSE || action->node->details->unclean == TRUE
|| is_set(rsc->flags, pe_rsc_failed)) {
if (is_set(rsc->flags, pe_rsc_failed)) {
pe_rsc_info(rsc,
"Demote of failed resource %s is implicit after %s is fenced",
rsc->id, target->details->uname);
} else {
pe_rsc_info(rsc, "%s is implicit after %s is fenced",
action->uuid, target->details->uname);
}
/* The demote would never complete and is now implied by the
* fencing, so convert it into a pseudo-action.
*/
update_action_flags(action, pe_action_pseudo|pe_action_runnable,
__FUNCTION__, __LINE__);
if (pe_rsc_is_bundled(rsc)) {
/* Do nothing, let the recovery be ordered after the parent's implied stop */
} else if (order_implicit) {
order_actions(stonith_op, action, pe_order_preserve|pe_order_optional);
}
}
}
g_list_free(action_list);
}
void
rsc_stonith_ordering(pe_resource_t * rsc, pe_action_t * stonith_op, pe_working_set_t * data_set)
{
if (rsc->children) {
GListPtr gIter = NULL;
for (gIter = rsc->children; gIter != NULL; gIter = gIter->next) {
pe_resource_t *child_rsc = (pe_resource_t *) gIter->data;
rsc_stonith_ordering(child_rsc, stonith_op, data_set);
}
} else if (is_not_set(rsc->flags, pe_rsc_managed)) {
pe_rsc_trace(rsc, "Skipping fencing constraints for unmanaged resource: %s", rsc->id);
} else {
native_start_constraints(rsc, stonith_op, data_set);
native_stop_constraints(rsc, stonith_op, data_set);
}
}
void
ReloadRsc(pe_resource_t * rsc, pe_node_t *node, pe_working_set_t * data_set)
{
GListPtr gIter = NULL;
pe_action_t *reload = NULL;
if (rsc->children) {
for (gIter = rsc->children; gIter != NULL; gIter = gIter->next) {
pe_resource_t *child_rsc = (pe_resource_t *) gIter->data;
ReloadRsc(child_rsc, node, data_set);
}
return;
} else if (rsc->variant > pe_native) {
/* Complex resource with no children */
return;
} else if (is_not_set(rsc->flags, pe_rsc_managed)) {
pe_rsc_trace(rsc, "%s: unmanaged", rsc->id);
return;
} else if (is_set(rsc->flags, pe_rsc_failed) || is_set(rsc->flags, pe_rsc_start_pending)) {
pe_rsc_trace(rsc, "%s: general resource state: flags=0x%.16llx", rsc->id, rsc->flags);
stop_action(rsc, node, FALSE); /* Force a full restart, overkill? */
return;
} else if (node == NULL) {
pe_rsc_trace(rsc, "%s: not active", rsc->id);
return;
}
pe_rsc_trace(rsc, "Processing %s", rsc->id);
set_bit(rsc->flags, pe_rsc_reload);
reload = custom_action(
rsc, reload_key(rsc), CRMD_ACTION_RELOAD, node, FALSE, TRUE, data_set);
pe_action_set_reason(reload, "resource definition change", FALSE);
custom_action_order(NULL, NULL, reload, rsc, stop_key(rsc), NULL,
pe_order_optional|pe_order_then_cancels_first,
data_set);
custom_action_order(NULL, NULL, reload, rsc, demote_key(rsc), NULL,
pe_order_optional|pe_order_then_cancels_first,
data_set);
}
void
native_append_meta(pe_resource_t * rsc, xmlNode * xml)
{
char *value = g_hash_table_lookup(rsc->meta, XML_RSC_ATTR_INCARNATION);
pe_resource_t *parent;
if (value) {
char *name = NULL;
name = crm_meta_name(XML_RSC_ATTR_INCARNATION);
crm_xml_add(xml, name, value);
free(name);
}
value = g_hash_table_lookup(rsc->meta, XML_RSC_ATTR_REMOTE_NODE);
if (value) {
char *name = NULL;
name = crm_meta_name(XML_RSC_ATTR_REMOTE_NODE);
crm_xml_add(xml, name, value);
free(name);
}
for (parent = rsc; parent != NULL; parent = parent->parent) {
if (parent->container) {
crm_xml_add(xml, CRM_META"_"XML_RSC_ATTR_CONTAINER, parent->container->id);
}
}
}
diff --git a/lib/pengine/common.c b/lib/pengine/common.c
index b642d40074..b9a622d42c 100644
--- a/lib/pengine/common.c
+++ b/lib/pengine/common.c
@@ -1,585 +1,602 @@
/*
* Copyright 2004-2020 the Pacemaker project contributors
*
* The version control history for this file may have further details.
*
* This source code is licensed under the GNU Lesser General Public License
* version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY.
*/
#include
#include
#include
#include
#include
#include
#include
gboolean was_processing_error = FALSE;
gboolean was_processing_warning = FALSE;
static bool
check_health(const char *value)
{
return safe_str_eq(value, "none")
|| safe_str_eq(value, "custom")
|| safe_str_eq(value, "only-green")
|| safe_str_eq(value, "progressive")
|| safe_str_eq(value, "migrate-on-red");
}
static bool
check_stonith_action(const char *value)
{
return safe_str_eq(value, "reboot")
|| safe_str_eq(value, "poweroff")
|| safe_str_eq(value, "off");
}
static bool
check_placement_strategy(const char *value)
{
return safe_str_eq(value, "default")
|| safe_str_eq(value, "utilization")
|| safe_str_eq(value, "minimal")
|| safe_str_eq(value, "balanced");
}
static pcmk__cluster_option_t pe_opts[] = {
/* name, old name, type, allowed values,
* default value, validator,
* short description,
* long description
*/
{
"no-quorum-policy", NULL, "enum", "stop, freeze, ignore, suicide",
"stop", pcmk__valid_quorum,
"What to do when the cluster does not have quorum",
NULL
},
{
"symmetric-cluster", NULL, "boolean", NULL,
"true", pcmk__valid_boolean,
"Whether resources can run on any node by default",
NULL
},
{
"maintenance-mode", NULL, "boolean", NULL,
"false", pcmk__valid_boolean,
"Whether the cluster should refrain from monitoring, starting, "
"and stopping resources",
NULL
},
{
"start-failure-is-fatal", NULL, "boolean", NULL,
"true", pcmk__valid_boolean,
"Whether a start failure should prevent a resource from being "
"recovered on the same node",
"When true, the cluster will immediately ban a resource from a node "
"if it fails to start there. When false, the cluster will instead "
"check the resource's fail count against its migration-threshold."
},
{
"enable-startup-probes", NULL, "boolean", NULL,
"true", pcmk__valid_boolean,
"Whether the cluster should check for active resources during start-up",
NULL
},
{
XML_CONFIG_ATTR_SHUTDOWN_LOCK, NULL, "boolean", NULL,
"false", pcmk__valid_boolean,
"Whether to lock resources to a cleanly shut down node",
"When true, resources active on a node when it is cleanly shut down "
"are kept \"locked\" to that node (not allowed to run elsewhere) "
"until they start again on that node after it rejoins (or for at "
"most shutdown-lock-limit, if set). Stonith resources and "
"Pacemaker Remote connections are never locked. Clone and bundle "
"instances and the master role of promotable clones are currently "
"never locked, though support could be added in a future release."
},
{
XML_CONFIG_ATTR_SHUTDOWN_LOCK_LIMIT, NULL, "time", NULL,
"0", pcmk__valid_interval_spec,
"Do not lock resources to a cleanly shut down node longer than this",
"If shutdown-lock is true and this is set to a nonzero time duration, "
"shutdown locks will expire after this much time has passed since "
"the shutdown was initiated, even if the node has not rejoined."
},
// Fencing-related options
{
"stonith-enabled", NULL, "boolean", NULL,
"true", pcmk__valid_boolean,
"*** Advanced Use Only *** "
"Whether nodes may be fenced as part of recovery",
"If false, unresponsive nodes are immediately assumed to be harmless, "
"and resources that were active on them may be recovered "
"elsewhere. This can result in a \"split-brain\" situation, "
"potentially leading to data loss and/or service unavailability."
},
{
"stonith-action", NULL, "enum", "reboot, off, poweroff",
"reboot", check_stonith_action,
"Action to send to fence device when a node needs to be fenced "
"(\"poweroff\" is a deprecated alias for \"off\")",
NULL
},
{
"stonith-timeout", NULL, "time", NULL,
"60s", pcmk__valid_interval_spec,
"*** Advanced Use Only *** Unused by Pacemaker",
"This value is not used by Pacemaker, but is kept for backward "
"compatibility, and certain legacy fence agents might use it."
},
{
XML_ATTR_HAVE_WATCHDOG, NULL, "boolean", NULL,
"false", pcmk__valid_boolean,
"Whether watchdog integration is enabled",
"This is set automatically by the cluster according to whether SBD "
"is detected to be in use. User-configured values are ignored."
},
{
"concurrent-fencing", NULL, "boolean", NULL,
#ifdef DEFAULT_CONCURRENT_FENCING_TRUE
"true",
#else
"false",
#endif
pcmk__valid_boolean,
"Allow performing fencing operations in parallel",
NULL
},
{
"startup-fencing", NULL, "boolean", NULL,
"true", pcmk__valid_boolean,
"*** Advanced Use Only *** Whether to fence unseen nodes at start-up",
"Setting this to false may lead to a \"split-brain\" situation,"
"potentially leading to data loss and/or service unavailability."
},
+ {
+ XML_CONFIG_ATTR_PRIORITY_FENCING_DELAY, NULL, "time", NULL,
+ NULL, pcmk__valid_interval_spec,
+ "Enforced fencing delay targeting the lost nodes with the highest total resource priority",
+ "Enforce specified delay for the fencings that are targeting the lost "
+ "nodes with the highest total resource priority in case we don't "
+ "have the majority of the nodes in our cluster partition, so that "
+ "the more significant nodes potentially win any fencing match, "
+ "which is especially meaningful under split-brain of 2-node "
+ "cluster. A promoted resource instance takes the base priority + 1 "
+ "on calculation if the base priority is not 0. If all the nodes "
+ "have equal priority, then any pcmk_delay_base/max configured for "
+ "the corresponding fencing resources will be applied. Otherwise as "
+ "long as it's set, even if to 0, it takes precedence over any "
+ "configured pcmk_delay_base/max. By default, priority fencing "
+ "delay is disabled."
+ },
{
"cluster-delay", NULL, "time", NULL,
"60s", pcmk__valid_interval_spec,
"Maximum time for node-to-node communication",
"The node elected Designated Controller (DC) will consider an action "
"failed if it does not get a response from the node executing the "
"action within this time (after considering the action's own "
"timeout). The \"correct\" value will depend on the speed and "
"load of your network and cluster nodes."
},
{
"batch-limit", NULL, "integer", NULL,
"0", pcmk__valid_number,
"Maximum number of jobs that the cluster may execute in parallel "
"across all nodes",
"The \"correct\" value will depend on the speed and load of your "
"network and cluster nodes. If set to 0, the cluster will "
"impose a dynamically calculated limit when any node has a "
"high load."
},
{
"migration-limit", NULL, "integer", NULL,
"-1", pcmk__valid_number,
"The number of live migration actions that the cluster is allowed "
"to execute in parallel on a node (-1 means no limit)"
},
/* Orphans and stopping */
{
"stop-all-resources", NULL, "boolean", NULL,
"false", pcmk__valid_boolean,
"Whether the cluster should stop all active resources",
NULL
},
{
"stop-orphan-resources", NULL, "boolean", NULL,
"true", pcmk__valid_boolean,
"Whether to stop resources that were removed from the configuration",
NULL
},
{
"stop-orphan-actions", NULL, "boolean", NULL,
"true", pcmk__valid_boolean,
"Whether to cancel recurring actions removed from the configuration",
NULL
},
{
"remove-after-stop", NULL, "boolean", NULL,
"false", pcmk__valid_boolean,
"*** Advanced Use Only *** Whether to remove stopped resources from "
"the executor",
"Values other than default are poorly tested and potentially dangerous."
},
/* Storing inputs */
{
"pe-error-series-max", NULL, "integer", NULL,
"-1", pcmk__valid_number,
"The number of scheduler inputs resulting in errors to save",
"Zero to disable, -1 to store unlimited."
},
{
"pe-warn-series-max", NULL, "integer", NULL,
"5000", pcmk__valid_number,
"The number of scheduler inputs resulting in warnings to save",
"Zero to disable, -1 to store unlimited."
},
{
"pe-input-series-max", NULL, "integer", NULL,
"4000", pcmk__valid_number,
"The number of scheduler inputs without errors or warnings to save",
"Zero to disable, -1 to store unlimited."
},
/* Node health */
{
"node-health-strategy", NULL, "enum",
"none, migrate-on-red, only-green, progressive, custom",
"none", check_health,
"How cluster should react to node health attributes",
"Requires external entities to create node attributes (named with "
"the prefix \"#health\") with values \"red\", \"yellow\" or "
"\"green\"."
},
{
"node-health-base", NULL, "integer", NULL,
"0", pcmk__valid_number,
"Base health score assigned to a node",
"Only used when node-health-strategy is set to progressive."
},
{
"node-health-green", NULL, "integer", NULL,
"0", pcmk__valid_number,
"The score to use for a node health attribute whose value is \"green\"",
"Only used when node-health-strategy is set to custom or progressive."
},
{
"node-health-yellow", NULL, "integer", NULL,
"0", pcmk__valid_number,
"The score to use for a node health attribute whose value is \"yellow\"",
"Only used when node-health-strategy is set to custom or progressive."
},
{
"node-health-red", NULL, "integer", NULL,
"-INFINITY", pcmk__valid_number,
"The score to use for a node health attribute whose value is \"red\"",
"Only used when node-health-strategy is set to custom or progressive."
},
/*Placement Strategy*/
{
"placement-strategy", NULL, "enum",
"default, utilization, minimal, balanced",
"default", check_placement_strategy,
"How the cluster should allocate resources to nodes",
NULL
},
};
void
pe_metadata(void)
{
pcmk__print_option_metadata("pacemaker-schedulerd", "1.0",
"Pacemaker scheduler options",
"Cluster options used by Pacemaker's scheduler"
" (formerly called pengine)",
pe_opts, DIMOF(pe_opts));
}
void
verify_pe_options(GHashTable * options)
{
pcmk__validate_cluster_options(options, pe_opts, DIMOF(pe_opts));
}
const char *
pe_pref(GHashTable * options, const char *name)
{
return pcmk__cluster_option(options, pe_opts, DIMOF(pe_opts), name);
}
const char *
fail2text(enum action_fail_response fail)
{
const char *result = "";
switch (fail) {
case action_fail_ignore:
result = "ignore";
break;
case action_fail_block:
result = "block";
break;
case action_fail_recover:
result = "recover";
break;
case action_fail_migrate:
result = "migrate";
break;
case action_fail_stop:
result = "stop";
break;
case action_fail_fence:
result = "fence";
break;
case action_fail_standby:
result = "standby";
break;
case action_fail_restart_container:
result = "restart-container";
break;
case action_fail_reset_remote:
result = "reset-remote";
break;
}
return result;
}
enum action_tasks
text2task(const char *task)
{
if (safe_str_eq(task, CRMD_ACTION_STOP)) {
return stop_rsc;
} else if (safe_str_eq(task, CRMD_ACTION_STOPPED)) {
return stopped_rsc;
} else if (safe_str_eq(task, CRMD_ACTION_START)) {
return start_rsc;
} else if (safe_str_eq(task, CRMD_ACTION_STARTED)) {
return started_rsc;
} else if (safe_str_eq(task, CRM_OP_SHUTDOWN)) {
return shutdown_crm;
} else if (safe_str_eq(task, CRM_OP_FENCE)) {
return stonith_node;
} else if (safe_str_eq(task, CRMD_ACTION_STATUS)) {
return monitor_rsc;
} else if (safe_str_eq(task, CRMD_ACTION_NOTIFY)) {
return action_notify;
} else if (safe_str_eq(task, CRMD_ACTION_NOTIFIED)) {
return action_notified;
} else if (safe_str_eq(task, CRMD_ACTION_PROMOTE)) {
return action_promote;
} else if (safe_str_eq(task, CRMD_ACTION_DEMOTE)) {
return action_demote;
} else if (safe_str_eq(task, CRMD_ACTION_PROMOTED)) {
return action_promoted;
} else if (safe_str_eq(task, CRMD_ACTION_DEMOTED)) {
return action_demoted;
}
#if SUPPORT_TRACING
if (safe_str_eq(task, CRMD_ACTION_CANCEL)) {
return no_action;
} else if (safe_str_eq(task, CRMD_ACTION_DELETE)) {
return no_action;
} else if (safe_str_eq(task, CRMD_ACTION_STATUS)) {
return no_action;
} else if (safe_str_eq(task, CRM_OP_PROBED)) {
return no_action;
} else if (safe_str_eq(task, CRM_OP_LRM_REFRESH)) {
return no_action;
} else if (safe_str_eq(task, CRMD_ACTION_MIGRATE)) {
return no_action;
} else if (safe_str_eq(task, CRMD_ACTION_MIGRATED)) {
return no_action;
}
crm_trace("Unsupported action: %s", task);
#endif
return no_action;
}
const char *
task2text(enum action_tasks task)
{
const char *result = "";
switch (task) {
case no_action:
result = "no_action";
break;
case stop_rsc:
result = CRMD_ACTION_STOP;
break;
case stopped_rsc:
result = CRMD_ACTION_STOPPED;
break;
case start_rsc:
result = CRMD_ACTION_START;
break;
case started_rsc:
result = CRMD_ACTION_STARTED;
break;
case shutdown_crm:
result = CRM_OP_SHUTDOWN;
break;
case stonith_node:
result = CRM_OP_FENCE;
break;
case monitor_rsc:
result = CRMD_ACTION_STATUS;
break;
case action_notify:
result = CRMD_ACTION_NOTIFY;
break;
case action_notified:
result = CRMD_ACTION_NOTIFIED;
break;
case action_promote:
result = CRMD_ACTION_PROMOTE;
break;
case action_promoted:
result = CRMD_ACTION_PROMOTED;
break;
case action_demote:
result = CRMD_ACTION_DEMOTE;
break;
case action_demoted:
result = CRMD_ACTION_DEMOTED;
break;
}
return result;
}
const char *
role2text(enum rsc_role_e role)
{
switch (role) {
case RSC_ROLE_UNKNOWN:
return RSC_ROLE_UNKNOWN_S;
case RSC_ROLE_STOPPED:
return RSC_ROLE_STOPPED_S;
case RSC_ROLE_STARTED:
return RSC_ROLE_STARTED_S;
case RSC_ROLE_SLAVE:
return RSC_ROLE_SLAVE_S;
case RSC_ROLE_MASTER:
return RSC_ROLE_MASTER_S;
}
CRM_CHECK(role >= RSC_ROLE_UNKNOWN, return RSC_ROLE_UNKNOWN_S);
CRM_CHECK(role < RSC_ROLE_MAX, return RSC_ROLE_UNKNOWN_S);
// coverity[dead_error_line]
return RSC_ROLE_UNKNOWN_S;
}
enum rsc_role_e
text2role(const char *role)
{
CRM_ASSERT(role != NULL);
if (safe_str_eq(role, RSC_ROLE_STOPPED_S)) {
return RSC_ROLE_STOPPED;
} else if (safe_str_eq(role, RSC_ROLE_STARTED_S)) {
return RSC_ROLE_STARTED;
} else if (safe_str_eq(role, RSC_ROLE_SLAVE_S)) {
return RSC_ROLE_SLAVE;
} else if (safe_str_eq(role, RSC_ROLE_MASTER_S)) {
return RSC_ROLE_MASTER;
} else if (safe_str_eq(role, RSC_ROLE_UNKNOWN_S)) {
return RSC_ROLE_UNKNOWN;
}
crm_err("Unknown role: %s", role);
return RSC_ROLE_UNKNOWN;
}
int
merge_weights(int w1, int w2)
{
int result = w1 + w2;
if (w1 <= -INFINITY || w2 <= -INFINITY) {
if (w1 >= INFINITY || w2 >= INFINITY) {
crm_trace("-INFINITY + INFINITY == -INFINITY");
}
return -INFINITY;
} else if (w1 >= INFINITY || w2 >= INFINITY) {
return INFINITY;
}
/* detect wrap-around */
if (result > 0) {
if (w1 <= 0 && w2 < 0) {
result = -INFINITY;
}
} else if (w1 > 0 && w2 > 0) {
result = INFINITY;
}
/* detect +/- INFINITY */
if (result >= INFINITY) {
result = INFINITY;
} else if (result <= -INFINITY) {
result = -INFINITY;
}
crm_trace("%d + %d = %d", w1, w2, result);
return result;
}
void
add_hash_param(GHashTable * hash, const char *name, const char *value)
{
CRM_CHECK(hash != NULL, return);
crm_trace("adding: name=%s value=%s", crm_str(name), crm_str(value));
if (name == NULL || value == NULL) {
return;
} else if (safe_str_eq(value, "#default")) {
return;
} else if (g_hash_table_lookup(hash, name) == NULL) {
g_hash_table_insert(hash, strdup(name), strdup(value));
}
}
const char *
pe_node_attribute_calculated(const pe_node_t *node, const char *name,
const pe_resource_t *rsc)
{
const char *source;
if(node == NULL) {
return NULL;
} else if(rsc == NULL) {
return g_hash_table_lookup(node->details->attrs, name);
}
source = g_hash_table_lookup(rsc->meta, XML_RSC_ATTR_TARGET);
if(source == NULL || safe_str_eq("host", source) == FALSE) {
return g_hash_table_lookup(node->details->attrs, name);
}
/* Use attributes set for the containers location
* instead of for the container itself
*
* Useful when the container is using the host's local
* storage
*/
CRM_ASSERT(node->details->remote_rsc);
CRM_ASSERT(node->details->remote_rsc->container);
if(node->details->remote_rsc->container->running_on) {
pe_node_t *host = node->details->remote_rsc->container->running_on->data;
pe_rsc_trace(rsc, "%s: Looking for %s on the container host %s", rsc->id, name, host->details->uname);
return g_hash_table_lookup(host->details->attrs, name);
}
pe_rsc_trace(rsc, "%s: Not looking for %s on the container host: %s is inactive",
rsc->id, name, node->details->remote_rsc->container->id);
return NULL;
}
const char *
pe_node_attribute_raw(pe_node_t *node, const char *name)
{
if(node == NULL) {
return NULL;
}
return g_hash_table_lookup(node->details->attrs, name);
}
diff --git a/lib/pengine/native.c b/lib/pengine/native.c
index 8308d88667..3a6dc24da7 100644
--- a/lib/pengine/native.c
+++ b/lib/pengine/native.c
@@ -1,1397 +1,1444 @@
/*
* Copyright 2004-2020 the Pacemaker project contributors
*
* The version control history for this file may have further details.
*
* This source code is licensed under the GNU Lesser General Public License
* version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY.
*/
#include
#include
#include
#include
#include
#include
#include
#define VARIANT_NATIVE 1
#include "./variant.h"
/*!
* \internal
* \brief Check whether a resource is active on multiple nodes
*/
static bool
is_multiply_active(pe_resource_t *rsc)
{
unsigned int count = 0;
if (rsc->variant == pe_native) {
pe__find_active_requires(rsc, &count);
}
return count > 1;
}
+static void
+native_priority_to_node(pe_resource_t * rsc, pe_node_t * node)
+{
+ int priority = 0;
+
+ if (rsc->priority == 0) {
+ return;
+ }
+
+ if (rsc->role == RSC_ROLE_MASTER) {
+ // Promoted instance takes base priority + 1
+ priority = rsc->priority + 1;
+
+ } else {
+ priority = rsc->priority;
+ }
+
+ node->details->priority += priority;
+ pe_rsc_trace(rsc, "Node '%s' now has priority %d with %s'%s' (priority: %d%s)",
+ node->details->uname, node->details->priority,
+ rsc->role == RSC_ROLE_MASTER ? "promoted " : "",
+ rsc->id, rsc->priority,
+ rsc->role == RSC_ROLE_MASTER ? " + 1" : "");
+
+ /* Priority of a resource running on a guest node is added to the cluster
+ * node as well. */
+ if (node->details->remote_rsc
+ && node->details->remote_rsc->container) {
+ GListPtr gIter = node->details->remote_rsc->container->running_on;
+
+ for (; gIter != NULL; gIter = gIter->next) {
+ pe_node_t *a_node = gIter->data;
+
+ a_node->details->priority += priority;
+ pe_rsc_trace(rsc, "Node '%s' now has priority %d with %s'%s' (priority: %d%s) "
+ "from guest node '%s'",
+ a_node->details->uname, a_node->details->priority,
+ rsc->role == RSC_ROLE_MASTER ? "promoted " : "",
+ rsc->id, rsc->priority,
+ rsc->role == RSC_ROLE_MASTER ? " + 1" : "",
+ node->details->uname);
+ }
+ }
+}
+
void
native_add_running(pe_resource_t * rsc, pe_node_t * node, pe_working_set_t * data_set)
{
GListPtr gIter = rsc->running_on;
CRM_CHECK(node != NULL, return);
for (; gIter != NULL; gIter = gIter->next) {
pe_node_t *a_node = (pe_node_t *) gIter->data;
CRM_CHECK(a_node != NULL, return);
if (safe_str_eq(a_node->details->id, node->details->id)) {
return;
}
}
pe_rsc_trace(rsc, "Adding %s to %s %s", rsc->id, node->details->uname,
is_set(rsc->flags, pe_rsc_managed)?"":"(unmanaged)");
rsc->running_on = g_list_append(rsc->running_on, node);
if (rsc->variant == pe_native) {
node->details->running_rsc = g_list_append(node->details->running_rsc, rsc);
+
+ native_priority_to_node(rsc, node);
}
if (rsc->variant == pe_native && node->details->maintenance) {
clear_bit(rsc->flags, pe_rsc_managed);
}
if (is_not_set(rsc->flags, pe_rsc_managed)) {
pe_resource_t *p = rsc->parent;
pe_rsc_info(rsc, "resource %s isn't managed", rsc->id);
resource_location(rsc, node, INFINITY, "not_managed_default", data_set);
while(p && node->details->online) {
/* add without the additional location constraint */
p->running_on = g_list_append(p->running_on, node);
p = p->parent;
}
return;
}
if (is_multiply_active(rsc)) {
switch (rsc->recovery_type) {
case recovery_stop_only:
{
GHashTableIter gIter;
pe_node_t *local_node = NULL;
/* make sure it doesn't come up again */
if (rsc->allowed_nodes != NULL) {
g_hash_table_destroy(rsc->allowed_nodes);
}
rsc->allowed_nodes = node_hash_from_list(data_set->nodes);
g_hash_table_iter_init(&gIter, rsc->allowed_nodes);
while (g_hash_table_iter_next(&gIter, NULL, (void **)&local_node)) {
local_node->weight = -INFINITY;
}
}
break;
case recovery_stop_start:
break;
case recovery_block:
clear_bit(rsc->flags, pe_rsc_managed);
set_bit(rsc->flags, pe_rsc_block);
/* If the resource belongs to a group or bundle configured with
* multiple-active=block, block the entire entity.
*/
if (rsc->parent
&& (rsc->parent->variant == pe_group || rsc->parent->variant == pe_container)
&& rsc->parent->recovery_type == recovery_block) {
GListPtr gIter = rsc->parent->children;
for (; gIter != NULL; gIter = gIter->next) {
pe_resource_t *child = (pe_resource_t *) gIter->data;
clear_bit(child->flags, pe_rsc_managed);
set_bit(child->flags, pe_rsc_block);
}
}
break;
}
crm_debug("%s is active on multiple nodes including %s: %s",
rsc->id, node->details->uname,
recovery2text(rsc->recovery_type));
} else {
pe_rsc_trace(rsc, "Resource %s is active on: %s", rsc->id, node->details->uname);
}
if (rsc->parent != NULL) {
native_add_running(rsc->parent, node, data_set);
}
}
static void
recursive_clear_unique(pe_resource_t *rsc)
{
clear_bit(rsc->flags, pe_rsc_unique);
add_hash_param(rsc->meta, XML_RSC_ATTR_UNIQUE, XML_BOOLEAN_FALSE);
for (GList *child = rsc->children; child != NULL; child = child->next) {
recursive_clear_unique((pe_resource_t *) child->data);
}
}
gboolean
native_unpack(pe_resource_t * rsc, pe_working_set_t * data_set)
{
pe_resource_t *parent = uber_parent(rsc);
native_variant_data_t *native_data = NULL;
const char *standard = crm_element_value(rsc->xml, XML_AGENT_ATTR_CLASS);
uint32_t ra_caps = pcmk_get_ra_caps(standard);
pe_rsc_trace(rsc, "Processing resource %s...", rsc->id);
native_data = calloc(1, sizeof(native_variant_data_t));
rsc->variant_opaque = native_data;
// Only some agent standards support unique and promotable clones
if (is_not_set(ra_caps, pcmk_ra_cap_unique)
&& is_set(rsc->flags, pe_rsc_unique) && pe_rsc_is_clone(parent)) {
/* @COMPAT We should probably reject this situation as an error (as we
* do for promotable below) rather than warn and convert, but that would
* be a backward-incompatible change that we should probably do with a
* transform at a schema major version bump.
*/
pe__force_anon(standard, parent, rsc->id, data_set);
/* Clear globally-unique on the parent and all its descendents unpacked
* so far (clearing the parent should make any future children unpacking
* correct). We have to clear this resource explicitly because it isn't
* hooked into the parent's children yet.
*/
recursive_clear_unique(parent);
recursive_clear_unique(rsc);
}
if (is_not_set(ra_caps, pcmk_ra_cap_promotable)
&& is_set(parent->flags, pe_rsc_promotable)) {
pe_err("Resource %s is of type %s and therefore "
"cannot be used as a promotable clone resource",
rsc->id, standard);
return FALSE;
}
return TRUE;
}
static bool
rsc_is_on_node(pe_resource_t *rsc, const pe_node_t *node, int flags)
{
pe_rsc_trace(rsc, "Checking whether %s is on %s",
rsc->id, node->details->uname);
if (is_set(flags, pe_find_current) && rsc->running_on) {
for (GListPtr iter = rsc->running_on; iter; iter = iter->next) {
pe_node_t *loc = (pe_node_t *) iter->data;
if (loc->details == node->details) {
return TRUE;
}
}
} else if (is_set(flags, pe_find_inactive) && (rsc->running_on == NULL)) {
return TRUE;
} else if (is_not_set(flags, pe_find_current) && rsc->allocated_to
&& (rsc->allocated_to->details == node->details)) {
return TRUE;
}
return FALSE;
}
pe_resource_t *
native_find_rsc(pe_resource_t * rsc, const char *id, const pe_node_t *on_node,
int flags)
{
bool match = FALSE;
pe_resource_t *result = NULL;
CRM_CHECK(id && rsc && rsc->id, return NULL);
if (flags & pe_find_clone) {
const char *rid = ID(rsc->xml);
if (!pe_rsc_is_clone(uber_parent(rsc))) {
match = FALSE;
} else if (!strcmp(id, rsc->id) || safe_str_eq(id, rid)) {
match = TRUE;
}
} else if (!strcmp(id, rsc->id)) {
match = TRUE;
} else if (is_set(flags, pe_find_renamed)
&& rsc->clone_name && strcmp(rsc->clone_name, id) == 0) {
match = TRUE;
} else if (is_set(flags, pe_find_any)
|| (is_set(flags, pe_find_anon)
&& is_not_set(rsc->flags, pe_rsc_unique))) {
match = pe_base_name_eq(rsc, id);
}
if (match && on_node) {
bool match_node = rsc_is_on_node(rsc, on_node, flags);
if (match_node == FALSE) {
match = FALSE;
}
}
if (match) {
return rsc;
}
for (GListPtr gIter = rsc->children; gIter != NULL; gIter = gIter->next) {
pe_resource_t *child = (pe_resource_t *) gIter->data;
result = rsc->fns->find_rsc(child, id, on_node, flags);
if (result) {
return result;
}
}
return NULL;
}
char *
native_parameter(pe_resource_t * rsc, pe_node_t * node, gboolean create, const char *name,
pe_working_set_t * data_set)
{
char *value_copy = NULL;
const char *value = NULL;
GHashTable *hash = NULL;
GHashTable *local_hash = NULL;
CRM_CHECK(rsc != NULL, return NULL);
CRM_CHECK(name != NULL && strlen(name) != 0, return NULL);
pe_rsc_trace(rsc, "Looking up %s in %s", name, rsc->id);
if (create || g_hash_table_size(rsc->parameters) == 0) {
if (node != NULL) {
pe_rsc_trace(rsc, "Creating hash with node %s", node->details->uname);
} else {
pe_rsc_trace(rsc, "Creating default hash");
}
local_hash = crm_str_table_new();
get_rsc_attributes(local_hash, rsc, node, data_set);
hash = local_hash;
} else {
hash = rsc->parameters;
}
value = g_hash_table_lookup(hash, name);
if (value == NULL) {
/* try meta attributes instead */
value = g_hash_table_lookup(rsc->meta, name);
}
if (value != NULL) {
value_copy = strdup(value);
}
if (local_hash != NULL) {
g_hash_table_destroy(local_hash);
}
return value_copy;
}
gboolean
native_active(pe_resource_t * rsc, gboolean all)
{
GListPtr gIter = rsc->running_on;
for (; gIter != NULL; gIter = gIter->next) {
pe_node_t *a_node = (pe_node_t *) gIter->data;
if (a_node->details->unclean) {
crm_debug("Resource %s: node %s is unclean", rsc->id, a_node->details->uname);
return TRUE;
} else if (a_node->details->online == FALSE) {
crm_debug("Resource %s: node %s is offline", rsc->id, a_node->details->uname);
} else {
crm_debug("Resource %s active on %s", rsc->id, a_node->details->uname);
return TRUE;
}
}
return FALSE;
}
struct print_data_s {
long options;
void *print_data;
};
static void
native_print_attr(gpointer key, gpointer value, gpointer user_data)
{
long options = ((struct print_data_s *)user_data)->options;
void *print_data = ((struct print_data_s *)user_data)->print_data;
status_print("Option: %s = %s\n", (char *)key, (char *)value);
}
static const char *
native_pending_state(pe_resource_t * rsc)
{
const char *pending_state = NULL;
if (safe_str_eq(rsc->pending_task, CRMD_ACTION_START)) {
pending_state = "Starting";
} else if (safe_str_eq(rsc->pending_task, CRMD_ACTION_STOP)) {
pending_state = "Stopping";
} else if (safe_str_eq(rsc->pending_task, CRMD_ACTION_MIGRATE)) {
pending_state = "Migrating";
} else if (safe_str_eq(rsc->pending_task, CRMD_ACTION_MIGRATED)) {
/* Work might be done in here. */
pending_state = "Migrating";
} else if (safe_str_eq(rsc->pending_task, CRMD_ACTION_PROMOTE)) {
pending_state = "Promoting";
} else if (safe_str_eq(rsc->pending_task, CRMD_ACTION_DEMOTE)) {
pending_state = "Demoting";
}
return pending_state;
}
static const char *
native_pending_task(pe_resource_t * rsc)
{
const char *pending_task = NULL;
if (safe_str_eq(rsc->pending_task, CRMD_ACTION_STATUS)) {
pending_task = "Monitoring";
/* Pending probes are not printed, even if pending
* operations are requested. If someone ever requests that
* behavior, uncomment this and the corresponding part of
* unpack.c:unpack_rsc_op().
*/
/*
} else if (safe_str_eq(rsc->pending_task, "probe")) {
pending_task = "Checking";
*/
}
return pending_task;
}
static enum rsc_role_e
native_displayable_role(pe_resource_t *rsc)
{
enum rsc_role_e role = rsc->role;
if ((role == RSC_ROLE_STARTED)
&& is_set(uber_parent(rsc)->flags, pe_rsc_promotable)) {
role = RSC_ROLE_SLAVE;
}
return role;
}
static const char *
native_displayable_state(pe_resource_t *rsc, long options)
{
const char *rsc_state = NULL;
if (options & pe_print_pending) {
rsc_state = native_pending_state(rsc);
}
if (rsc_state == NULL) {
rsc_state = role2text(native_displayable_role(rsc));
}
return rsc_state;
}
static void
native_print_xml(pe_resource_t * rsc, const char *pre_text, long options, void *print_data)
{
const char *class = crm_element_value(rsc->xml, XML_AGENT_ATTR_CLASS);
const char *prov = crm_element_value(rsc->xml, XML_AGENT_ATTR_PROVIDER);
const char *rsc_state = native_displayable_state(rsc, options);
const char *target_role = NULL;
/* resource information. */
status_print("%sxml, XML_ATTR_TYPE));
status_print("role=\"%s\" ", rsc_state);
if (rsc->meta) {
target_role = g_hash_table_lookup(rsc->meta, XML_RSC_ATTR_TARGET_ROLE);
}
if (target_role) {
status_print("target_role=\"%s\" ", target_role);
}
status_print("active=\"%s\" ", rsc->fns->active(rsc, TRUE) ? "true" : "false");
status_print("orphaned=\"%s\" ", is_set(rsc->flags, pe_rsc_orphan) ? "true" : "false");
status_print("blocked=\"%s\" ", is_set(rsc->flags, pe_rsc_block) ? "true" : "false");
status_print("managed=\"%s\" ", is_set(rsc->flags, pe_rsc_managed) ? "true" : "false");
status_print("failed=\"%s\" ", is_set(rsc->flags, pe_rsc_failed) ? "true" : "false");
status_print("failure_ignored=\"%s\" ",
is_set(rsc->flags, pe_rsc_failure_ignored) ? "true" : "false");
status_print("nodes_running_on=\"%d\" ", g_list_length(rsc->running_on));
if (options & pe_print_pending) {
const char *pending_task = native_pending_task(rsc);
if (pending_task) {
status_print("pending=\"%s\" ", pending_task);
}
}
if (options & pe_print_dev) {
status_print("provisional=\"%s\" ",
is_set(rsc->flags, pe_rsc_provisional) ? "true" : "false");
status_print("runnable=\"%s\" ", is_set(rsc->flags, pe_rsc_runnable) ? "true" : "false");
status_print("priority=\"%f\" ", (double)rsc->priority);
status_print("variant=\"%s\" ", crm_element_name(rsc->xml));
}
/* print out the nodes this resource is running on */
if (options & pe_print_rsconly) {
status_print("/>\n");
/* do nothing */
} else if (rsc->running_on != NULL) {
GListPtr gIter = rsc->running_on;
status_print(">\n");
for (; gIter != NULL; gIter = gIter->next) {
pe_node_t *node = (pe_node_t *) gIter->data;
status_print("%s \n", pre_text,
node->details->uname, node->details->id,
node->details->online ? "false" : "true");
}
status_print("%s\n", pre_text);
} else {
status_print("/>\n");
}
}
// Append a flag to resource description string's flags list
static bool
add_output_flag(GString *s, const char *flag_desc, bool have_flags)
{
g_string_append(s, (have_flags? ", " : " ("));
g_string_append(s, flag_desc);
return true;
}
// Append a node name to resource description string's node list
static bool
add_output_node(GString *s, const char *node, bool have_nodes)
{
g_string_append(s, (have_nodes? " " : " [ "));
g_string_append(s, node);
return true;
}
/*!
* \internal
* \brief Create a string description of a resource
*
* \param[in] rsc Resource to describe
* \param[in] name Desired identifier for the resource
* \param[in] node If not NULL, node that resource is "on"
* \param[in] options Bitmask of pe_print_*
* \param[in] target_role Resource's target role
* \param[in] show_nodes Whether to display nodes when multiply active
*
* \return Newly allocated string description of resource
* \note Caller must free the result with g_free().
*/
static gchar *
native_output_string(pe_resource_t *rsc, const char *name, pe_node_t *node,
long options, const char *target_role, bool show_nodes)
{
const char *class = crm_element_value(rsc->xml, XML_AGENT_ATTR_CLASS);
const char *provider = NULL;
const char *kind = crm_element_value(rsc->xml, XML_ATTR_TYPE);
char *retval = NULL;
GString *outstr = NULL;
bool have_flags = false;
CRM_CHECK(name != NULL, name = "unknown");
CRM_CHECK(kind != NULL, kind = "unknown");
CRM_CHECK(class != NULL, class = "unknown");
if (is_set(pcmk_get_ra_caps(class), pcmk_ra_cap_provider)) {
provider = crm_element_value(rsc->xml, XML_AGENT_ATTR_PROVIDER);
}
if ((node == NULL) && (rsc->lock_node != NULL)) {
node = rsc->lock_node;
}
if (is_set(options, pe_print_rsconly)
|| pcmk__list_of_multiple(rsc->running_on)) {
node = NULL;
}
// We need a string of at least this size
outstr = g_string_sized_new(strlen(name) + strlen(class) + strlen(kind)
+ (provider? (strlen(provider) + 2) : 0)
+ (node? strlen(node->details->uname) + 1 : 0)
+ 11);
// Resource name and agent
g_string_printf(outstr, "%s\t(%s%s%s:%s):\t", name, class,
/* @COMPAT This should be a single ':' (see CLBZ#5395) but
* to avoid breaking anything relying on it, we're keeping
* it like this until the next minor version bump.
*/
(provider? "::" : ""), (provider? provider : ""), kind);
// State on node
if (is_set(rsc->flags, pe_rsc_orphan)) {
g_string_append(outstr, " ORPHANED");
}
if (is_set(rsc->flags, pe_rsc_failed)) {
enum rsc_role_e role = native_displayable_role(rsc);
if (role > RSC_ROLE_SLAVE) {
g_string_append_printf(outstr, " FAILED %s", role2text(role));
} else {
g_string_append(outstr, " FAILED");
}
} else {
g_string_append(outstr, native_displayable_state(rsc, options));
}
if (node) {
g_string_append_printf(outstr, " %s", node->details->uname);
}
// Flags, as: ( [...])
if (node && !(node->details->online) && node->details->unclean) {
have_flags = add_output_flag(outstr, "UNCLEAN", have_flags);
}
if (node && (node == rsc->lock_node)) {
have_flags = add_output_flag(outstr, "LOCKED", have_flags);
}
if (is_set(options, pe_print_pending)) {
const char *pending_task = native_pending_task(rsc);
if (pending_task) {
have_flags = add_output_flag(outstr, pending_task, have_flags);
}
}
if (target_role) {
enum rsc_role_e target_role_e = text2role(target_role);
/* Only show target role if it limits our abilities (i.e. ignore
* Started, as it is the default anyways, and doesn't prevent the
* resource from becoming Master).
*/
if (target_role_e == RSC_ROLE_STOPPED) {
have_flags = add_output_flag(outstr, "disabled", have_flags);
} else if (is_set(uber_parent(rsc)->flags, pe_rsc_promotable)
&& target_role_e == RSC_ROLE_SLAVE) {
have_flags = add_output_flag(outstr, "target-role:", have_flags);
g_string_append(outstr, target_role);
}
}
if (is_set(rsc->flags, pe_rsc_block)) {
have_flags = add_output_flag(outstr, "blocked", have_flags);
} else if (is_not_set(rsc->flags, pe_rsc_managed)) {
have_flags = add_output_flag(outstr, "unmanaged", have_flags);
}
if (is_set(rsc->flags, pe_rsc_failure_ignored)) {
have_flags = add_output_flag(outstr, "failure ignored", have_flags);
}
if (is_set(options, pe_print_dev)) {
if (is_set(options, pe_rsc_provisional)) {
have_flags = add_output_flag(outstr, "provisional", have_flags);
}
if (is_not_set(options, pe_rsc_runnable)) {
have_flags = add_output_flag(outstr, "non-startable", have_flags);
}
have_flags = add_output_flag(outstr, "variant:", have_flags);
g_string_append_printf(outstr, "%s priority:%f",
crm_element_name(rsc->xml),
(double) (rsc->priority));
}
if (have_flags) {
g_string_append(outstr, ")");
}
// User-supplied description
if (is_set(options, pe_print_rsconly)
|| pcmk__list_of_multiple(rsc->running_on)) {
const char *desc = crm_element_value(rsc->xml, XML_ATTR_DESC);
if (desc) {
g_string_append_printf(outstr, " %s", desc);
}
}
if (show_nodes && is_not_set(options, pe_print_rsconly)
&& pcmk__list_of_multiple(rsc->running_on)) {
bool have_nodes = false;
for (GList *iter = rsc->running_on; iter != NULL; iter = iter->next) {
pe_node_t *n = (pe_node_t *) iter->data;
have_nodes = add_output_node(outstr, n->details->uname, have_nodes);
}
if (have_nodes) {
g_string_append(outstr, " ]");
}
}
retval = outstr->str;
g_string_free(outstr, FALSE);
return retval;
}
int
pe__common_output_html(pcmk__output_t *out, pe_resource_t * rsc,
const char *name, pe_node_t *node, long options)
{
const char *kind = crm_element_value(rsc->xml, XML_ATTR_TYPE);
const char *target_role = NULL;
xmlNodePtr list_node = NULL;
const char *cl = NULL;
CRM_ASSERT(rsc->variant == pe_native);
CRM_ASSERT(kind != NULL);
if (rsc->meta) {
const char *is_internal = g_hash_table_lookup(rsc->meta, XML_RSC_ATTR_INTERNAL_RSC);
if (crm_is_true(is_internal) && is_not_set(options, pe_print_implicit)) {
crm_trace("skipping print of internal resource %s", rsc->id);
return pcmk_rc_no_output;
}
target_role = g_hash_table_lookup(rsc->meta, XML_RSC_ATTR_TARGET_ROLE);
}
if (is_not_set(rsc->flags, pe_rsc_managed)) {
cl = "rsc-managed";
} else if (is_set(rsc->flags, pe_rsc_failed)) {
cl = "rsc-failed";
} else if (rsc->variant == pe_native && (rsc->running_on == NULL)) {
cl = "rsc-failed";
} else if (pcmk__list_of_multiple(rsc->running_on)) {
cl = "rsc-multiple";
} else if (is_set(rsc->flags, pe_rsc_failure_ignored)) {
cl = "rsc-failure-ignored";
} else {
cl = "rsc-ok";
}
{
gchar *s = native_output_string(rsc, name, node, options, target_role,
true);
list_node = pcmk__output_create_html_node(out, "li", NULL, NULL, NULL);
pcmk_create_html_node(list_node, "span", NULL, cl, s);
g_free(s);
}
if (is_set(options, pe_print_details)) {
GHashTableIter iter;
gpointer key, value;
out->begin_list(out, NULL, NULL, "Options");
g_hash_table_iter_init(&iter, rsc->parameters);
while (g_hash_table_iter_next(&iter, &key, &value)) {
out->list_item(out, NULL, "Option: %s = %s", (char *) key, (char *) value);
}
out->end_list(out);
}
if (is_set(options, pe_print_dev)) {
GHashTableIter iter;
pe_node_t *n = NULL;
out->begin_list(out, NULL, NULL, "Allowed Nodes");
g_hash_table_iter_init(&iter, rsc->allowed_nodes);
while (g_hash_table_iter_next(&iter, NULL, (void **)&n)) {
out->list_item(out, NULL, "%s %d", n->details->uname, n->weight);
}
out->end_list(out);
}
if (is_set(options, pe_print_max_details)) {
GHashTableIter iter;
pe_node_t *n = NULL;
out->begin_list(out, NULL, NULL, "=== Allowed Nodes");
g_hash_table_iter_init(&iter, rsc->allowed_nodes);
while (g_hash_table_iter_next(&iter, NULL, (void **)&n)) {
pe__output_node(n, FALSE, out);
}
out->end_list(out);
}
return pcmk_rc_ok;
}
int
pe__common_output_text(pcmk__output_t *out, pe_resource_t * rsc,
const char *name, pe_node_t *node, long options)
{
const char *target_role = NULL;
CRM_ASSERT(rsc->variant == pe_native);
if (rsc->meta) {
const char *is_internal = g_hash_table_lookup(rsc->meta, XML_RSC_ATTR_INTERNAL_RSC);
if (crm_is_true(is_internal) && is_not_set(options, pe_print_implicit)) {
crm_trace("skipping print of internal resource %s", rsc->id);
return pcmk_rc_no_output;
}
target_role = g_hash_table_lookup(rsc->meta, XML_RSC_ATTR_TARGET_ROLE);
}
{
gchar *s = native_output_string(rsc, name, node, options, target_role,
true);
out->list_item(out, NULL, "%s", s);
g_free(s);
}
if (is_set(options, pe_print_details)) {
GHashTableIter iter;
gpointer key, value;
out->begin_list(out, NULL, NULL, "Options");
g_hash_table_iter_init(&iter, rsc->parameters);
while (g_hash_table_iter_next(&iter, &key, &value)) {
out->list_item(out, NULL, "Option: %s = %s", (char *) key, (char *) value);
}
out->end_list(out);
}
if (is_set(options, pe_print_dev)) {
GHashTableIter iter;
pe_node_t *n = NULL;
out->begin_list(out, NULL, NULL, "Allowed Nodes");
g_hash_table_iter_init(&iter, rsc->allowed_nodes);
while (g_hash_table_iter_next(&iter, NULL, (void **)&n)) {
out->list_item(out, NULL, "%s %d", n->details->uname, n->weight);
}
out->end_list(out);
}
if (is_set(options, pe_print_max_details)) {
GHashTableIter iter;
pe_node_t *n = NULL;
out->begin_list(out, NULL, NULL, "=== Allowed Nodes");
g_hash_table_iter_init(&iter, rsc->allowed_nodes);
while (g_hash_table_iter_next(&iter, NULL, (void **)&n)) {
pe__output_node(n, FALSE, out);
}
out->end_list(out);
}
return pcmk_rc_ok;
}
void
common_print(pe_resource_t * rsc, const char *pre_text, const char *name, pe_node_t *node, long options, void *print_data)
{
const char *target_role = NULL;
CRM_ASSERT(rsc->variant == pe_native);
if (rsc->meta) {
const char *is_internal = g_hash_table_lookup(rsc->meta,
XML_RSC_ATTR_INTERNAL_RSC);
if (crm_is_true(is_internal) && is_not_set(options, pe_print_implicit)) {
crm_trace("skipping print of internal resource %s", rsc->id);
return;
}
target_role = g_hash_table_lookup(rsc->meta, XML_RSC_ATTR_TARGET_ROLE);
}
if (options & pe_print_xml) {
native_print_xml(rsc, pre_text, options, print_data);
return;
}
if ((pre_text == NULL) && (options & pe_print_printf)) {
pre_text = " ";
}
if (options & pe_print_html) {
if (is_not_set(rsc->flags, pe_rsc_managed)) {
status_print("");
} else if (is_set(rsc->flags, pe_rsc_failed)) {
status_print("");
} else if (rsc->running_on == NULL) {
status_print("");
} else if (pcmk__list_of_multiple(rsc->running_on)) {
status_print("");
} else if (is_set(rsc->flags, pe_rsc_failure_ignored)) {
status_print("");
} else {
status_print("");
}
}
{
gchar *resource_s = native_output_string(rsc, name, node, options,
target_role, false);
status_print("%s%s", (pre_text? pre_text : ""), resource_s);
g_free(resource_s);
}
#if CURSES_ENABLED
if (is_set(options, pe_print_ncurses)
&& is_not_set(options, pe_print_rsconly)
&& !pcmk__list_of_multiple(rsc->running_on)) {
/* coverity[negative_returns] False positive */
move(-1, 0);
}
#endif
if (is_set(options, pe_print_html)) {
status_print(" ");
}
if (is_not_set(options, pe_print_rsconly)
&& pcmk__list_of_multiple(rsc->running_on)) {
GListPtr gIter = rsc->running_on;
int counter = 0;
if (options & pe_print_html) {
status_print("\n");
} else if ((options & pe_print_printf)
|| (options & pe_print_ncurses)) {
status_print("[");
}
for (; gIter != NULL; gIter = gIter->next) {
pe_node_t *n = (pe_node_t *) gIter->data;
counter++;
if (options & pe_print_html) {
status_print("- \n%s", n->details->uname);
} else if ((options & pe_print_printf)
|| (options & pe_print_ncurses)) {
status_print(" %s", n->details->uname);
} else if ((options & pe_print_log)) {
status_print("\t%d : %s", counter, n->details->uname);
} else {
status_print("%s", n->details->uname);
}
if (options & pe_print_html) {
status_print("
\n");
}
}
if (options & pe_print_html) {
status_print("
\n");
} else if ((options & pe_print_printf)
|| (options & pe_print_ncurses)) {
status_print(" ]");
}
}
if (options & pe_print_html) {
status_print("
\n");
} else if (options & pe_print_suppres_nl) {
/* nothing */
} else if ((options & pe_print_printf) || (options & pe_print_ncurses)) {
status_print("\n");
}
if (options & pe_print_details) {
struct print_data_s pdata;
pdata.options = options;
pdata.print_data = print_data;
g_hash_table_foreach(rsc->parameters, native_print_attr, &pdata);
}
if (options & pe_print_dev) {
GHashTableIter iter;
pe_node_t *n = NULL;
status_print("%s\tAllowed Nodes", pre_text);
g_hash_table_iter_init(&iter, rsc->allowed_nodes);
while (g_hash_table_iter_next(&iter, NULL, (void **)&n)) {
status_print("%s\t * %s %d", pre_text, n->details->uname, n->weight);
}
}
if (options & pe_print_max_details) {
GHashTableIter iter;
pe_node_t *n = NULL;
status_print("%s\t=== Allowed Nodes\n", pre_text);
g_hash_table_iter_init(&iter, rsc->allowed_nodes);
while (g_hash_table_iter_next(&iter, NULL, (void **)&n)) {
print_node("\t", n, FALSE);
}
}
}
void
native_print(pe_resource_t * rsc, const char *pre_text, long options, void *print_data)
{
pe_node_t *node = NULL;
CRM_ASSERT(rsc->variant == pe_native);
if (options & pe_print_xml) {
native_print_xml(rsc, pre_text, options, print_data);
return;
}
node = pe__current_node(rsc);
if (node == NULL) {
// This is set only if a non-probe action is pending on this node
node = rsc->pending_node;
}
common_print(rsc, pre_text, rsc_printable_id(rsc), node, options, print_data);
}
int
pe__resource_xml(pcmk__output_t *out, va_list args)
{
unsigned int options = va_arg(args, unsigned int);
pe_resource_t *rsc = va_arg(args, pe_resource_t *);
const char *class = crm_element_value(rsc->xml, XML_AGENT_ATTR_CLASS);
const char *prov = crm_element_value(rsc->xml, XML_AGENT_ATTR_PROVIDER);
const char *rsc_state = native_displayable_state(rsc, options);
long is_print_pending = options & pe_print_pending;
long is_print_dev = options & pe_print_dev;
char ra_name[LINE_MAX];
char *nodes_running_on = NULL;
char *priority = NULL;
int rc = pcmk_rc_no_output;
CRM_ASSERT(rsc->variant == pe_native);
/* resource information. */
sprintf(ra_name, "%s%s%s:%s", class, prov ? "::" : "", prov ? prov : ""
, crm_element_value(rsc->xml, XML_ATTR_TYPE));
nodes_running_on = crm_itoa(g_list_length(rsc->running_on));
priority = crm_ftoa(rsc->priority);
rc = pe__name_and_nvpairs_xml(out, true, "resource", 16
, "id", rsc_printable_id(rsc)
, "resource_agent", ra_name
, "role", rsc_state
, "target_role", (rsc->meta ? g_hash_table_lookup(rsc->meta, XML_RSC_ATTR_TARGET_ROLE) : NULL)
, "active", BOOL2STR(rsc->fns->active(rsc, TRUE))
, "orphaned", BOOL2STR(is_set(rsc->flags, pe_rsc_orphan))
, "blocked", BOOL2STR(is_set(rsc->flags, pe_rsc_block))
, "managed", BOOL2STR(is_set(rsc->flags, pe_rsc_managed))
, "failed", BOOL2STR(is_set(rsc->flags, pe_rsc_failed))
, "failure_ignored", BOOL2STR(is_set(rsc->flags, pe_rsc_failure_ignored))
, "nodes_running_on", nodes_running_on
, "pending", (is_print_pending ? native_pending_task(rsc) : NULL)
, "provisional", (is_print_dev ? BOOL2STR(is_set(rsc->flags, pe_rsc_provisional)) : NULL)
, "runnable", (is_print_dev ? BOOL2STR(is_set(rsc->flags, pe_rsc_runnable)) : NULL)
, "priority", (is_print_dev ? priority : NULL)
, "variant", (is_print_dev ? crm_element_name(rsc->xml) : NULL));
free(priority);
free(nodes_running_on);
CRM_ASSERT(rc == pcmk_rc_ok);
if (rsc->running_on != NULL) {
GListPtr gIter = rsc->running_on;
for (; gIter != NULL; gIter = gIter->next) {
pe_node_t *node = (pe_node_t *) gIter->data;
rc = pe__name_and_nvpairs_xml(out, false, "node", 3
, "name", node->details->uname
, "id", node->details->id
, "cached", BOOL2STR(node->details->online));
CRM_ASSERT(rc == pcmk_rc_ok);
}
}
pcmk__output_xml_pop_parent(out);
return rc;
}
int
pe__resource_html(pcmk__output_t *out, va_list args)
{
unsigned int options = va_arg(args, unsigned int);
pe_resource_t *rsc = va_arg(args, pe_resource_t *);
pe_node_t *node = pe__current_node(rsc);
CRM_ASSERT(rsc->variant == pe_native);
if (node == NULL) {
// This is set only if a non-probe action is pending on this node
node = rsc->pending_node;
}
return pe__common_output_html(out, rsc, rsc_printable_id(rsc), node, options);
}
int
pe__resource_text(pcmk__output_t *out, va_list args)
{
unsigned int options = va_arg(args, unsigned int);
pe_resource_t *rsc = va_arg(args, pe_resource_t *);
pe_node_t *node = pe__current_node(rsc);
CRM_ASSERT(rsc->variant == pe_native);
if (node == NULL) {
// This is set only if a non-probe action is pending on this node
node = rsc->pending_node;
}
return pe__common_output_text(out, rsc, rsc_printable_id(rsc), node, options);
}
void
native_free(pe_resource_t * rsc)
{
pe_rsc_trace(rsc, "Freeing resource action list (not the data)");
common_free(rsc);
}
enum rsc_role_e
native_resource_state(const pe_resource_t * rsc, gboolean current)
{
enum rsc_role_e role = rsc->next_role;
if (current) {
role = rsc->role;
}
pe_rsc_trace(rsc, "%s state: %s", rsc->id, role2text(role));
return role;
}
/*!
* \internal
* \brief List nodes where a resource (or any of its children) is
*
* \param[in] rsc Resource to check
* \param[out] list List to add result to
* \param[in] current 0 = where known, 1 = running, 2 = running or pending
*
* \return If list contains only one node, that node
*/
pe_node_t *
native_location(const pe_resource_t *rsc, GList **list, int current)
{
pe_node_t *one = NULL;
GListPtr result = NULL;
if (rsc->children) {
GListPtr gIter = rsc->children;
for (; gIter != NULL; gIter = gIter->next) {
pe_resource_t *child = (pe_resource_t *) gIter->data;
child->fns->location(child, &result, current);
}
} else if (current) {
if (rsc->running_on) {
result = g_list_copy(rsc->running_on);
}
if ((current == 2) && rsc->pending_node
&& !pe_find_node_id(result, rsc->pending_node->details->id)) {
result = g_list_append(result, rsc->pending_node);
}
} else if (current == FALSE && rsc->allocated_to) {
result = g_list_append(NULL, rsc->allocated_to);
}
if (result && (result->next == NULL)) {
one = result->data;
}
if (list) {
GListPtr gIter = result;
for (; gIter != NULL; gIter = gIter->next) {
pe_node_t *node = (pe_node_t *) gIter->data;
if (*list == NULL || pe_find_node_id(*list, node->details->id) == NULL) {
*list = g_list_append(*list, node);
}
}
}
g_list_free(result);
return one;
}
static void
get_rscs_brief(GListPtr rsc_list, GHashTable * rsc_table, GHashTable * active_table)
{
GListPtr gIter = rsc_list;
for (; gIter != NULL; gIter = gIter->next) {
pe_resource_t *rsc = (pe_resource_t *) gIter->data;
const char *class = crm_element_value(rsc->xml, XML_AGENT_ATTR_CLASS);
const char *kind = crm_element_value(rsc->xml, XML_ATTR_TYPE);
int offset = 0;
char buffer[LINE_MAX];
int *rsc_counter = NULL;
int *active_counter = NULL;
if (rsc->variant != pe_native) {
continue;
}
offset += snprintf(buffer + offset, LINE_MAX - offset, "%s", class);
if (is_set(pcmk_get_ra_caps(class), pcmk_ra_cap_provider)) {
const char *prov = crm_element_value(rsc->xml, XML_AGENT_ATTR_PROVIDER);
offset += snprintf(buffer + offset, LINE_MAX - offset, "::%s", prov);
}
offset += snprintf(buffer + offset, LINE_MAX - offset, ":%s", kind);
CRM_LOG_ASSERT(offset > 0);
if (rsc_table) {
rsc_counter = g_hash_table_lookup(rsc_table, buffer);
if (rsc_counter == NULL) {
rsc_counter = calloc(1, sizeof(int));
*rsc_counter = 0;
g_hash_table_insert(rsc_table, strdup(buffer), rsc_counter);
}
(*rsc_counter)++;
}
if (active_table) {
GListPtr gIter2 = rsc->running_on;
for (; gIter2 != NULL; gIter2 = gIter2->next) {
pe_node_t *node = (pe_node_t *) gIter2->data;
GHashTable *node_table = NULL;
if (node->details->unclean == FALSE && node->details->online == FALSE) {
continue;
}
node_table = g_hash_table_lookup(active_table, node->details->uname);
if (node_table == NULL) {
node_table = crm_str_table_new();
g_hash_table_insert(active_table, strdup(node->details->uname), node_table);
}
active_counter = g_hash_table_lookup(node_table, buffer);
if (active_counter == NULL) {
active_counter = calloc(1, sizeof(int));
*active_counter = 0;
g_hash_table_insert(node_table, strdup(buffer), active_counter);
}
(*active_counter)++;
}
}
}
}
static void
destroy_node_table(gpointer data)
{
GHashTable *node_table = data;
if (node_table) {
g_hash_table_destroy(node_table);
}
}
void
print_rscs_brief(GListPtr rsc_list, const char *pre_text, long options,
void *print_data, gboolean print_all)
{
GHashTable *rsc_table = crm_str_table_new();
GHashTable *active_table = g_hash_table_new_full(crm_str_hash, g_str_equal,
free, destroy_node_table);
GHashTableIter hash_iter;
char *type = NULL;
int *rsc_counter = NULL;
get_rscs_brief(rsc_list, rsc_table, active_table);
g_hash_table_iter_init(&hash_iter, rsc_table);
while (g_hash_table_iter_next(&hash_iter, (gpointer *)&type, (gpointer *)&rsc_counter)) {
GHashTableIter hash_iter2;
char *node_name = NULL;
GHashTable *node_table = NULL;
int active_counter_all = 0;
g_hash_table_iter_init(&hash_iter2, active_table);
while (g_hash_table_iter_next(&hash_iter2, (gpointer *)&node_name, (gpointer *)&node_table)) {
int *active_counter = g_hash_table_lookup(node_table, type);
if (active_counter == NULL || *active_counter == 0) {
continue;
} else {
active_counter_all += *active_counter;
}
if (options & pe_print_rsconly) {
node_name = NULL;
}
if (options & pe_print_html) {
status_print("\n");
}
if (print_all) {
status_print("%s%d/%d\t(%s):\tActive %s\n", pre_text ? pre_text : "",
active_counter ? *active_counter : 0,
rsc_counter ? *rsc_counter : 0, type,
active_counter && (*active_counter > 0) && node_name ? node_name : "");
} else {
status_print("%s%d\t(%s):\tActive %s\n", pre_text ? pre_text : "",
active_counter ? *active_counter : 0, type,
active_counter && (*active_counter > 0) && node_name ? node_name : "");
}
if (options & pe_print_html) {
status_print("\n");
}
}
if (print_all && active_counter_all == 0) {
if (options & pe_print_html) {
status_print("\n");
}
status_print("%s%d/%d\t(%s):\tActive\n", pre_text ? pre_text : "",
active_counter_all,
rsc_counter ? *rsc_counter : 0, type);
if (options & pe_print_html) {
status_print("\n");
}
}
}
if (rsc_table) {
g_hash_table_destroy(rsc_table);
rsc_table = NULL;
}
if (active_table) {
g_hash_table_destroy(active_table);
active_table = NULL;
}
}
int
pe__rscs_brief_output(pcmk__output_t *out, GListPtr rsc_list, long options, gboolean print_all)
{
GHashTable *rsc_table = crm_str_table_new();
GHashTable *active_table = g_hash_table_new_full(crm_str_hash, g_str_equal,
free, destroy_node_table);
GHashTableIter hash_iter;
char *type = NULL;
int *rsc_counter = NULL;
int rc = pcmk_rc_no_output;
get_rscs_brief(rsc_list, rsc_table, active_table);
g_hash_table_iter_init(&hash_iter, rsc_table);
while (g_hash_table_iter_next(&hash_iter, (gpointer *)&type, (gpointer *)&rsc_counter)) {
GHashTableIter hash_iter2;
char *node_name = NULL;
GHashTable *node_table = NULL;
int active_counter_all = 0;
g_hash_table_iter_init(&hash_iter2, active_table);
while (g_hash_table_iter_next(&hash_iter2, (gpointer *)&node_name, (gpointer *)&node_table)) {
int *active_counter = g_hash_table_lookup(node_table, type);
if (active_counter == NULL || *active_counter == 0) {
continue;
} else {
active_counter_all += *active_counter;
}
if (options & pe_print_rsconly) {
node_name = NULL;
}
if (print_all) {
out->list_item(out, NULL, " %d/%d\t(%s):\tActive %s",
*active_counter,
rsc_counter ? *rsc_counter : 0, type,
(*active_counter > 0) && node_name ? node_name : "");
} else {
out->list_item(out, NULL, " %d\t(%s):\tActive %s",
*active_counter, type,
(*active_counter > 0) && node_name ? node_name : "");
}
rc = pcmk_rc_ok;
}
if (print_all && active_counter_all == 0) {
out->list_item(out, NULL, " %d/%d\t(%s):\tActive",
active_counter_all,
rsc_counter ? *rsc_counter : 0, type);
rc = pcmk_rc_ok;
}
}
if (rsc_table) {
g_hash_table_destroy(rsc_table);
rsc_table = NULL;
}
if (active_table) {
g_hash_table_destroy(active_table);
active_table = NULL;
}
return rc;
}
diff --git a/lib/pengine/unpack.c b/lib/pengine/unpack.c
index 11bbd08a00..fee6b9938c 100644
--- a/lib/pengine/unpack.c
+++ b/lib/pengine/unpack.c
@@ -1,3860 +1,3873 @@
/*
* Copyright 2004-2020 the Pacemaker project contributors
*
* The version control history for this file may have further details.
*
* This source code is licensed under the GNU Lesser General Public License
* version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY.
*/
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
CRM_TRACE_INIT_DATA(pe_status);
#define set_config_flag(data_set, option, flag) do { \
const char *tmp = pe_pref(data_set->config_hash, option); \
if(tmp) { \
if(crm_is_true(tmp)) { \
set_bit(data_set->flags, flag); \
} else { \
clear_bit(data_set->flags, flag); \
} \
} \
} while(0)
static void unpack_rsc_op(pe_resource_t *rsc, pe_node_t *node, xmlNode *xml_op,
xmlNode **last_failure,
enum action_fail_response *failed,
pe_working_set_t *data_set);
static gboolean determine_remote_online_status(pe_working_set_t * data_set, pe_node_t * this_node);
static void add_node_attrs(xmlNode *attrs, pe_node_t *node, bool overwrite,
pe_working_set_t *data_set);
// Bitmask for warnings we only want to print once
uint32_t pe_wo = 0;
static gboolean
is_dangling_guest_node(pe_node_t *node)
{
/* we are looking for a remote-node that was supposed to be mapped to a
* container resource, but all traces of that container have disappeared
* from both the config and the status section. */
if (pe__is_guest_or_remote_node(node) &&
node->details->remote_rsc &&
node->details->remote_rsc->container == NULL &&
is_set(node->details->remote_rsc->flags, pe_rsc_orphan_container_filler)) {
return TRUE;
}
return FALSE;
}
/*!
* \brief Schedule a fence action for a node
*
* \param[in,out] data_set Current working set of cluster
* \param[in,out] node Node to fence
* \param[in] reason Text description of why fencing is needed
+ * \param[in] priority_delay Whether to consider `priority-fencing-delay`
*/
void
-pe_fence_node(pe_working_set_t * data_set, pe_node_t * node, const char *reason)
+pe_fence_node(pe_working_set_t * data_set, pe_node_t * node,
+ const char *reason, bool priority_delay)
{
CRM_CHECK(node, return);
/* A guest node is fenced by marking its container as failed */
if (pe__is_guest_node(node)) {
pe_resource_t *rsc = node->details->remote_rsc->container;
if (is_set(rsc->flags, pe_rsc_failed) == FALSE) {
if (!is_set(rsc->flags, pe_rsc_managed)) {
crm_notice("Not fencing guest node %s "
"(otherwise would because %s): "
"its guest resource %s is unmanaged",
node->details->uname, reason, rsc->id);
} else {
crm_warn("Guest node %s will be fenced "
"(by recovering its guest resource %s): %s",
node->details->uname, rsc->id, reason);
/* We don't mark the node as unclean because that would prevent the
* node from running resources. We want to allow it to run resources
* in this transition if the recovery succeeds.
*/
node->details->remote_requires_reset = TRUE;
set_bit(rsc->flags, pe_rsc_failed);
}
}
} else if (is_dangling_guest_node(node)) {
crm_info("Cleaning up dangling connection for guest node %s: "
"fencing was already done because %s, "
"and guest resource no longer exists",
node->details->uname, reason);
set_bit(node->details->remote_rsc->flags, pe_rsc_failed);
} else if (pe__is_remote_node(node)) {
pe_resource_t *rsc = node->details->remote_rsc;
if (rsc && (!is_set(rsc->flags, pe_rsc_managed))) {
crm_notice("Not fencing remote node %s "
"(otherwise would because %s): connection is unmanaged",
node->details->uname, reason);
} else if(node->details->remote_requires_reset == FALSE) {
node->details->remote_requires_reset = TRUE;
crm_warn("Remote node %s %s: %s",
node->details->uname,
pe_can_fence(data_set, node)? "will be fenced" : "is unclean",
reason);
}
node->details->unclean = TRUE;
- pe_fence_op(node, NULL, TRUE, reason, data_set);
+ // No need to apply `priority-fencing-delay` for remote nodes
+ pe_fence_op(node, NULL, TRUE, reason, FALSE, data_set);
} else if (node->details->unclean) {
crm_trace("Cluster node %s %s because %s",
node->details->uname,
pe_can_fence(data_set, node)? "would also be fenced" : "also is unclean",
reason);
} else {
crm_warn("Cluster node %s %s: %s",
node->details->uname,
pe_can_fence(data_set, node)? "will be fenced" : "is unclean",
reason);
node->details->unclean = TRUE;
- pe_fence_op(node, NULL, TRUE, reason, data_set);
+ pe_fence_op(node, NULL, TRUE, reason, priority_delay, data_set);
}
}
// @TODO xpaths can't handle templates, rules, or id-refs
// nvpair with provides or requires set to unfencing
#define XPATH_UNFENCING_NVPAIR XML_CIB_TAG_NVPAIR \
"[(@" XML_NVPAIR_ATTR_NAME "='" XML_RSC_ATTR_PROVIDES "'" \
"or @" XML_NVPAIR_ATTR_NAME "='" XML_RSC_ATTR_REQUIRES "') " \
"and @" XML_NVPAIR_ATTR_VALUE "='unfencing']"
// unfencing in rsc_defaults or any resource
#define XPATH_ENABLE_UNFENCING \
"/" XML_TAG_CIB "/" XML_CIB_TAG_CONFIGURATION "/" XML_CIB_TAG_RESOURCES \
"//" XML_TAG_META_SETS "/" XPATH_UNFENCING_NVPAIR \
"|/" XML_TAG_CIB "/" XML_CIB_TAG_CONFIGURATION "/" XML_CIB_TAG_RSCCONFIG \
"/" XML_TAG_META_SETS "/" XPATH_UNFENCING_NVPAIR
static
void set_if_xpath(unsigned long long flag, const char *xpath,
pe_working_set_t *data_set)
{
xmlXPathObjectPtr result = NULL;
if (is_not_set(data_set->flags, flag)) {
result = xpath_search(data_set->input, xpath);
if (result && (numXpathResults(result) > 0)) {
set_bit(data_set->flags, flag);
}
freeXpathObject(result);
}
}
gboolean
unpack_config(xmlNode * config, pe_working_set_t * data_set)
{
const char *value = NULL;
GHashTable *config_hash = crm_str_table_new();
data_set->config_hash = config_hash;
pe__unpack_dataset_nvpairs(config, XML_CIB_TAG_PROPSET, NULL, config_hash,
CIB_OPTIONS_FIRST, FALSE, data_set);
verify_pe_options(data_set->config_hash);
set_config_flag(data_set, "enable-startup-probes", pe_flag_startup_probes);
if(is_not_set(data_set->flags, pe_flag_startup_probes)) {
crm_info("Startup probes: disabled (dangerous)");
}
value = pe_pref(data_set->config_hash, XML_ATTR_HAVE_WATCHDOG);
if (value && crm_is_true(value)) {
crm_notice("Watchdog will be used via SBD if fencing is required "
"and stonith-watchdog-timeout is nonzero");
set_bit(data_set->flags, pe_flag_have_stonith_resource);
}
/* Set certain flags via xpath here, so they can be used before the relevant
* configuration sections are unpacked.
*/
set_if_xpath(pe_flag_enable_unfencing, XPATH_ENABLE_UNFENCING, data_set);
value = pe_pref(data_set->config_hash, "stonith-timeout");
data_set->stonith_timeout = (int) crm_parse_interval_spec(value);
crm_debug("STONITH timeout: %d", data_set->stonith_timeout);
set_config_flag(data_set, "stonith-enabled", pe_flag_stonith_enabled);
crm_debug("STONITH of failed nodes is %s",
is_set(data_set->flags, pe_flag_stonith_enabled) ? "enabled" : "disabled");
data_set->stonith_action = pe_pref(data_set->config_hash, "stonith-action");
if (!strcmp(data_set->stonith_action, "poweroff")) {
pe_warn_once(pe_wo_poweroff,
"Support for stonith-action of 'poweroff' is deprecated "
"and will be removed in a future release (use 'off' instead)");
data_set->stonith_action = "off";
}
crm_trace("STONITH will %s nodes", data_set->stonith_action);
set_config_flag(data_set, "concurrent-fencing", pe_flag_concurrent_fencing);
crm_debug("Concurrent fencing is %s",
is_set(data_set->flags, pe_flag_concurrent_fencing) ? "enabled" : "disabled");
+ // Default value -1 means `priority-fencing-delay` is disabled
+ data_set->priority_fencing_delay = -1;
+ value = pe_pref(data_set->config_hash,
+ XML_CONFIG_ATTR_PRIORITY_FENCING_DELAY);
+ if (value) {
+ data_set->priority_fencing_delay = crm_parse_interval_spec(value) / 1000;
+ crm_trace("Priority fencing delay is %ds", data_set->priority_fencing_delay);
+ }
+
set_config_flag(data_set, "stop-all-resources", pe_flag_stop_everything);
crm_debug("Stop all active resources: %s",
is_set(data_set->flags, pe_flag_stop_everything) ? "true" : "false");
set_config_flag(data_set, "symmetric-cluster", pe_flag_symmetric_cluster);
if (is_set(data_set->flags, pe_flag_symmetric_cluster)) {
crm_debug("Cluster is symmetric" " - resources can run anywhere by default");
}
value = pe_pref(data_set->config_hash, "no-quorum-policy");
if (safe_str_eq(value, "ignore")) {
data_set->no_quorum_policy = no_quorum_ignore;
} else if (safe_str_eq(value, "freeze")) {
data_set->no_quorum_policy = no_quorum_freeze;
} else if (safe_str_eq(value, "suicide")) {
if (is_set(data_set->flags, pe_flag_stonith_enabled)) {
int do_panic = 0;
crm_element_value_int(data_set->input, XML_ATTR_QUORUM_PANIC,
&do_panic);
if (do_panic || is_set(data_set->flags, pe_flag_have_quorum)) {
data_set->no_quorum_policy = no_quorum_suicide;
} else {
crm_notice("Resetting no-quorum-policy to 'stop': cluster has never had quorum");
data_set->no_quorum_policy = no_quorum_stop;
}
} else {
pcmk__config_err("Resetting no-quorum-policy to 'stop' because "
"fencing is disabled");
data_set->no_quorum_policy = no_quorum_stop;
}
} else {
data_set->no_quorum_policy = no_quorum_stop;
}
switch (data_set->no_quorum_policy) {
case no_quorum_freeze:
crm_debug("On loss of quorum: Freeze resources");
break;
case no_quorum_stop:
crm_debug("On loss of quorum: Stop ALL resources");
break;
case no_quorum_suicide:
crm_notice("On loss of quorum: Fence all remaining nodes");
break;
case no_quorum_ignore:
crm_notice("On loss of quorum: Ignore");
break;
}
set_config_flag(data_set, "stop-orphan-resources", pe_flag_stop_rsc_orphans);
crm_trace("Orphan resources are %s",
is_set(data_set->flags, pe_flag_stop_rsc_orphans) ? "stopped" : "ignored");
set_config_flag(data_set, "stop-orphan-actions", pe_flag_stop_action_orphans);
crm_trace("Orphan resource actions are %s",
is_set(data_set->flags, pe_flag_stop_action_orphans) ? "stopped" : "ignored");
set_config_flag(data_set, "remove-after-stop", pe_flag_remove_after_stop);
crm_trace("Stopped resources are removed from the status section: %s",
is_set(data_set->flags, pe_flag_remove_after_stop) ? "true" : "false");
set_config_flag(data_set, "maintenance-mode", pe_flag_maintenance_mode);
crm_trace("Maintenance mode: %s",
is_set(data_set->flags, pe_flag_maintenance_mode) ? "true" : "false");
set_config_flag(data_set, "start-failure-is-fatal", pe_flag_start_failure_fatal);
crm_trace("Start failures are %s",
is_set(data_set->flags,
pe_flag_start_failure_fatal) ? "always fatal" : "handled by failcount");
if (is_set(data_set->flags, pe_flag_stonith_enabled)) {
set_config_flag(data_set, "startup-fencing", pe_flag_startup_fencing);
}
if (is_set(data_set->flags, pe_flag_startup_fencing)) {
crm_trace("Unseen nodes will be fenced");
} else {
pe_warn_once(pe_wo_blind, "Blind faith: not fencing unseen nodes");
}
node_score_red = char2score(pe_pref(data_set->config_hash, "node-health-red"));
node_score_green = char2score(pe_pref(data_set->config_hash, "node-health-green"));
node_score_yellow = char2score(pe_pref(data_set->config_hash, "node-health-yellow"));
crm_debug("Node scores: 'red' = %s, 'yellow' = %s, 'green' = %s",
pe_pref(data_set->config_hash, "node-health-red"),
pe_pref(data_set->config_hash, "node-health-yellow"),
pe_pref(data_set->config_hash, "node-health-green"));
data_set->placement_strategy = pe_pref(data_set->config_hash, "placement-strategy");
crm_trace("Placement strategy: %s", data_set->placement_strategy);
set_config_flag(data_set, "shutdown-lock", pe_flag_shutdown_lock);
crm_trace("Resources will%s be locked to cleanly shut down nodes",
(is_set(data_set->flags, pe_flag_shutdown_lock)? "" : " not"));
if (is_set(data_set->flags, pe_flag_shutdown_lock)) {
value = pe_pref(data_set->config_hash,
XML_CONFIG_ATTR_SHUTDOWN_LOCK_LIMIT);
data_set->shutdown_lock = crm_parse_interval_spec(value) / 1000;
crm_trace("Shutdown locks expire after %us", data_set->shutdown_lock);
}
return TRUE;
}
static void
destroy_digest_cache(gpointer ptr)
{
op_digest_cache_t *data = ptr;
free_xml(data->params_all);
free_xml(data->params_secure);
free_xml(data->params_restart);
free(data->digest_all_calc);
free(data->digest_restart_calc);
free(data->digest_secure_calc);
free(data);
}
pe_node_t *
pe_create_node(const char *id, const char *uname, const char *type,
const char *score, pe_working_set_t * data_set)
{
pe_node_t *new_node = NULL;
if (pe_find_node(data_set->nodes, uname) != NULL) {
pcmk__config_warn("More than one node entry has name '%s'", uname);
}
new_node = calloc(1, sizeof(pe_node_t));
if (new_node == NULL) {
return NULL;
}
new_node->weight = char2score(score);
new_node->fixed = FALSE;
new_node->details = calloc(1, sizeof(struct pe_node_shared_s));
if (new_node->details == NULL) {
free(new_node);
return NULL;
}
crm_trace("Creating node for entry %s/%s", uname, id);
new_node->details->id = id;
new_node->details->uname = uname;
new_node->details->online = FALSE;
new_node->details->shutdown = FALSE;
new_node->details->rsc_discovery_enabled = TRUE;
new_node->details->running_rsc = NULL;
new_node->details->type = node_ping;
if (safe_str_eq(type, "remote")) {
new_node->details->type = node_remote;
set_bit(data_set->flags, pe_flag_have_remote_nodes);
} else if ((type == NULL) || safe_str_eq(type, "member")) {
new_node->details->type = node_member;
}
new_node->details->attrs = crm_str_table_new();
if (pe__is_guest_or_remote_node(new_node)) {
g_hash_table_insert(new_node->details->attrs, strdup(CRM_ATTR_KIND),
strdup("remote"));
} else {
g_hash_table_insert(new_node->details->attrs, strdup(CRM_ATTR_KIND),
strdup("cluster"));
}
new_node->details->utilization = crm_str_table_new();
new_node->details->digest_cache = g_hash_table_new_full(crm_str_hash,
g_str_equal, free,
destroy_digest_cache);
data_set->nodes = g_list_insert_sorted(data_set->nodes, new_node, sort_node_uname);
return new_node;
}
static const char *
expand_remote_rsc_meta(xmlNode *xml_obj, xmlNode *parent, pe_working_set_t *data)
{
xmlNode *attr_set = NULL;
xmlNode *attr = NULL;
const char *container_id = ID(xml_obj);
const char *remote_name = NULL;
const char *remote_server = NULL;
const char *remote_port = NULL;
const char *connect_timeout = "60s";
const char *remote_allow_migrate=NULL;
const char *is_managed = NULL;
for (attr_set = __xml_first_child_element(xml_obj); attr_set != NULL;
attr_set = __xml_next_element(attr_set)) {
if (safe_str_neq((const char *)attr_set->name, XML_TAG_META_SETS)) {
continue;
}
for (attr = __xml_first_child_element(attr_set); attr != NULL;
attr = __xml_next_element(attr)) {
const char *value = crm_element_value(attr, XML_NVPAIR_ATTR_VALUE);
const char *name = crm_element_value(attr, XML_NVPAIR_ATTR_NAME);
if (safe_str_eq(name, XML_RSC_ATTR_REMOTE_NODE)) {
remote_name = value;
} else if (safe_str_eq(name, "remote-addr")) {
remote_server = value;
} else if (safe_str_eq(name, "remote-port")) {
remote_port = value;
} else if (safe_str_eq(name, "remote-connect-timeout")) {
connect_timeout = value;
} else if (safe_str_eq(name, "remote-allow-migrate")) {
remote_allow_migrate=value;
} else if (safe_str_eq(name, XML_RSC_ATTR_MANAGED)) {
is_managed = value;
}
}
}
if (remote_name == NULL) {
return NULL;
}
if (pe_find_resource(data->resources, remote_name) != NULL) {
return NULL;
}
pe_create_remote_xml(parent, remote_name, container_id,
remote_allow_migrate, is_managed,
connect_timeout, remote_server, remote_port);
return remote_name;
}
static void
handle_startup_fencing(pe_working_set_t *data_set, pe_node_t *new_node)
{
if ((new_node->details->type == node_remote) && (new_node->details->remote_rsc == NULL)) {
/* Ignore fencing for remote nodes that don't have a connection resource
* associated with them. This happens when remote node entries get left
* in the nodes section after the connection resource is removed.
*/
return;
}
if (is_set(data_set->flags, pe_flag_startup_fencing)) {
// All nodes are unclean until we've seen their status entry
new_node->details->unclean = TRUE;
} else {
// Blind faith ...
new_node->details->unclean = FALSE;
}
/* We need to be able to determine if a node's status section
* exists or not separate from whether the node is unclean. */
new_node->details->unseen = TRUE;
}
gboolean
unpack_nodes(xmlNode * xml_nodes, pe_working_set_t * data_set)
{
xmlNode *xml_obj = NULL;
pe_node_t *new_node = NULL;
const char *id = NULL;
const char *uname = NULL;
const char *type = NULL;
const char *score = NULL;
for (xml_obj = __xml_first_child_element(xml_nodes); xml_obj != NULL;
xml_obj = __xml_next_element(xml_obj)) {
if (crm_str_eq((const char *)xml_obj->name, XML_CIB_TAG_NODE, TRUE)) {
new_node = NULL;
id = crm_element_value(xml_obj, XML_ATTR_ID);
uname = crm_element_value(xml_obj, XML_ATTR_UNAME);
type = crm_element_value(xml_obj, XML_ATTR_TYPE);
score = crm_element_value(xml_obj, XML_RULE_ATTR_SCORE);
crm_trace("Processing node %s/%s", uname, id);
if (id == NULL) {
pcmk__config_err("Ignoring <" XML_CIB_TAG_NODE
"> entry in configuration without id");
continue;
}
new_node = pe_create_node(id, uname, type, score, data_set);
if (new_node == NULL) {
return FALSE;
}
/* if(data_set->have_quorum == FALSE */
/* && data_set->no_quorum_policy == no_quorum_stop) { */
/* /\* start shutting resources down *\/ */
/* new_node->weight = -INFINITY; */
/* } */
handle_startup_fencing(data_set, new_node);
add_node_attrs(xml_obj, new_node, FALSE, data_set);
pe__unpack_dataset_nvpairs(xml_obj, XML_TAG_UTILIZATION, NULL,
new_node->details->utilization, NULL,
FALSE, data_set);
crm_trace("Done with node %s", crm_element_value(xml_obj, XML_ATTR_UNAME));
}
}
if (data_set->localhost && pe_find_node(data_set->nodes, data_set->localhost) == NULL) {
crm_info("Creating a fake local node");
pe_create_node(data_set->localhost, data_set->localhost, NULL, 0,
data_set);
}
return TRUE;
}
static void
setup_container(pe_resource_t * rsc, pe_working_set_t * data_set)
{
const char *container_id = NULL;
if (rsc->children) {
GListPtr gIter = rsc->children;
for (; gIter != NULL; gIter = gIter->next) {
pe_resource_t *child_rsc = (pe_resource_t *) gIter->data;
setup_container(child_rsc, data_set);
}
return;
}
container_id = g_hash_table_lookup(rsc->meta, XML_RSC_ATTR_CONTAINER);
if (container_id && safe_str_neq(container_id, rsc->id)) {
pe_resource_t *container = pe_find_resource(data_set->resources, container_id);
if (container) {
rsc->container = container;
set_bit(container->flags, pe_rsc_is_container);
container->fillers = g_list_append(container->fillers, rsc);
pe_rsc_trace(rsc, "Resource %s's container is %s", rsc->id, container_id);
} else {
pe_err("Resource %s: Unknown resource container (%s)", rsc->id, container_id);
}
}
}
gboolean
unpack_remote_nodes(xmlNode * xml_resources, pe_working_set_t * data_set)
{
xmlNode *xml_obj = NULL;
/* Create remote nodes and guest nodes from the resource configuration
* before unpacking resources.
*/
for (xml_obj = __xml_first_child_element(xml_resources); xml_obj != NULL;
xml_obj = __xml_next_element(xml_obj)) {
const char *new_node_id = NULL;
/* Check for remote nodes, which are defined by ocf:pacemaker:remote
* primitives.
*/
if (xml_contains_remote_node(xml_obj)) {
new_node_id = ID(xml_obj);
/* The "pe_find_node" check is here to make sure we don't iterate over
* an expanded node that has already been added to the node list. */
if (new_node_id && pe_find_node(data_set->nodes, new_node_id) == NULL) {
crm_trace("Found remote node %s defined by resource %s",
new_node_id, ID(xml_obj));
pe_create_node(new_node_id, new_node_id, "remote", NULL,
data_set);
}
continue;
}
/* Check for guest nodes, which are defined by special meta-attributes
* of a primitive of any type (for example, VirtualDomain or Xen).
*/
if (crm_str_eq((const char *)xml_obj->name, XML_CIB_TAG_RESOURCE, TRUE)) {
/* This will add an ocf:pacemaker:remote primitive to the
* configuration for the guest node's connection, to be unpacked
* later.
*/
new_node_id = expand_remote_rsc_meta(xml_obj, xml_resources, data_set);
if (new_node_id && pe_find_node(data_set->nodes, new_node_id) == NULL) {
crm_trace("Found guest node %s in resource %s",
new_node_id, ID(xml_obj));
pe_create_node(new_node_id, new_node_id, "remote", NULL,
data_set);
}
continue;
}
/* Check for guest nodes inside a group. Clones are currently not
* supported as guest nodes.
*/
if (crm_str_eq((const char *)xml_obj->name, XML_CIB_TAG_GROUP, TRUE)) {
xmlNode *xml_obj2 = NULL;
for (xml_obj2 = __xml_first_child_element(xml_obj); xml_obj2 != NULL;
xml_obj2 = __xml_next_element(xml_obj2)) {
new_node_id = expand_remote_rsc_meta(xml_obj2, xml_resources, data_set);
if (new_node_id && pe_find_node(data_set->nodes, new_node_id) == NULL) {
crm_trace("Found guest node %s in resource %s inside group %s",
new_node_id, ID(xml_obj2), ID(xml_obj));
pe_create_node(new_node_id, new_node_id, "remote", NULL,
data_set);
}
}
}
}
return TRUE;
}
/* Call this after all the nodes and resources have been
* unpacked, but before the status section is read.
*
* A remote node's online status is reflected by the state
* of the remote node's connection resource. We need to link
* the remote node to this connection resource so we can have
* easy access to the connection resource during the scheduler calculations.
*/
static void
link_rsc2remotenode(pe_working_set_t *data_set, pe_resource_t *new_rsc)
{
pe_node_t *remote_node = NULL;
if (new_rsc->is_remote_node == FALSE) {
return;
}
if (is_set(data_set->flags, pe_flag_quick_location)) {
/* remote_nodes and remote_resources are not linked in quick location calculations */
return;
}
remote_node = pe_find_node(data_set->nodes, new_rsc->id);
CRM_CHECK(remote_node != NULL, return;);
pe_rsc_trace(new_rsc, "Linking remote connection resource %s to node %s",
new_rsc->id, remote_node->details->uname);
remote_node->details->remote_rsc = new_rsc;
if (new_rsc->container == NULL) {
/* Handle start-up fencing for remote nodes (as opposed to guest nodes)
* the same as is done for cluster nodes.
*/
handle_startup_fencing(data_set, remote_node);
} else {
/* pe_create_node() marks the new node as "remote" or "cluster"; now
* that we know the node is a guest node, update it correctly.
*/
g_hash_table_replace(remote_node->details->attrs, strdup(CRM_ATTR_KIND),
strdup("container"));
}
}
static void
destroy_tag(gpointer data)
{
pe_tag_t *tag = data;
if (tag) {
free(tag->id);
g_list_free_full(tag->refs, free);
free(tag);
}
}
/*!
* \internal
* \brief Parse configuration XML for resource information
*
* \param[in] xml_resources Top of resource configuration XML
* \param[in,out] data_set Where to put resource information
*
* \return TRUE
*
* \note unpack_remote_nodes() MUST be called before this, so that the nodes can
* be used when common_unpack() calls resource_location()
*/
gboolean
unpack_resources(xmlNode * xml_resources, pe_working_set_t * data_set)
{
xmlNode *xml_obj = NULL;
GListPtr gIter = NULL;
data_set->template_rsc_sets = g_hash_table_new_full(crm_str_hash,
g_str_equal, free,
destroy_tag);
for (xml_obj = __xml_first_child_element(xml_resources); xml_obj != NULL;
xml_obj = __xml_next_element(xml_obj)) {
pe_resource_t *new_rsc = NULL;
if (crm_str_eq((const char *)xml_obj->name, XML_CIB_TAG_RSC_TEMPLATE, TRUE)) {
const char *template_id = ID(xml_obj);
if (template_id && g_hash_table_lookup_extended(data_set->template_rsc_sets,
template_id, NULL, NULL) == FALSE) {
/* Record the template's ID for the knowledge of its existence anyway. */
g_hash_table_insert(data_set->template_rsc_sets, strdup(template_id), NULL);
}
continue;
}
crm_trace("Beginning unpack... <%s id=%s... >", crm_element_name(xml_obj), ID(xml_obj));
if (common_unpack(xml_obj, &new_rsc, NULL, data_set)) {
data_set->resources = g_list_append(data_set->resources, new_rsc);
pe_rsc_trace(new_rsc, "Added resource %s", new_rsc->id);
} else {
pcmk__config_err("Ignoring <%s> resource '%s' "
"because configuration is invalid",
crm_element_name(xml_obj), crm_str(ID(xml_obj)));
if (new_rsc != NULL && new_rsc->fns != NULL) {
new_rsc->fns->free(new_rsc);
}
}
}
for (gIter = data_set->resources; gIter != NULL; gIter = gIter->next) {
pe_resource_t *rsc = (pe_resource_t *) gIter->data;
setup_container(rsc, data_set);
link_rsc2remotenode(data_set, rsc);
}
data_set->resources = g_list_sort(data_set->resources, sort_rsc_priority);
if (is_set(data_set->flags, pe_flag_quick_location)) {
/* Ignore */
} else if (is_set(data_set->flags, pe_flag_stonith_enabled)
&& is_set(data_set->flags, pe_flag_have_stonith_resource) == FALSE) {
pcmk__config_err("Resource start-up disabled since no STONITH resources have been defined");
pcmk__config_err("Either configure some or disable STONITH with the stonith-enabled option");
pcmk__config_err("NOTE: Clusters with shared data need STONITH to ensure data integrity");
}
return TRUE;
}
gboolean
unpack_tags(xmlNode * xml_tags, pe_working_set_t * data_set)
{
xmlNode *xml_tag = NULL;
data_set->tags = g_hash_table_new_full(crm_str_hash, g_str_equal, free,
destroy_tag);
for (xml_tag = __xml_first_child_element(xml_tags); xml_tag != NULL;
xml_tag = __xml_next_element(xml_tag)) {
xmlNode *xml_obj_ref = NULL;
const char *tag_id = ID(xml_tag);
if (crm_str_eq((const char *)xml_tag->name, XML_CIB_TAG_TAG, TRUE) == FALSE) {
continue;
}
if (tag_id == NULL) {
pcmk__config_err("Ignoring <%s> without " XML_ATTR_ID,
crm_element_name(xml_tag));
continue;
}
for (xml_obj_ref = __xml_first_child_element(xml_tag); xml_obj_ref != NULL;
xml_obj_ref = __xml_next_element(xml_obj_ref)) {
const char *obj_ref = ID(xml_obj_ref);
if (crm_str_eq((const char *)xml_obj_ref->name, XML_CIB_TAG_OBJ_REF, TRUE) == FALSE) {
continue;
}
if (obj_ref == NULL) {
pcmk__config_err("Ignoring <%s> for tag '%s' without " XML_ATTR_ID,
crm_element_name(xml_obj_ref), tag_id);
continue;
}
if (add_tag_ref(data_set->tags, tag_id, obj_ref) == FALSE) {
return FALSE;
}
}
}
return TRUE;
}
/* The ticket state section:
* "/cib/status/tickets/ticket_state" */
static gboolean
unpack_ticket_state(xmlNode * xml_ticket, pe_working_set_t * data_set)
{
const char *ticket_id = NULL;
const char *granted = NULL;
const char *last_granted = NULL;
const char *standby = NULL;
xmlAttrPtr xIter = NULL;
pe_ticket_t *ticket = NULL;
ticket_id = ID(xml_ticket);
if (ticket_id == NULL || strlen(ticket_id) == 0) {
return FALSE;
}
crm_trace("Processing ticket state for %s", ticket_id);
ticket = g_hash_table_lookup(data_set->tickets, ticket_id);
if (ticket == NULL) {
ticket = ticket_new(ticket_id, data_set);
if (ticket == NULL) {
return FALSE;
}
}
for (xIter = xml_ticket->properties; xIter; xIter = xIter->next) {
const char *prop_name = (const char *)xIter->name;
const char *prop_value = crm_element_value(xml_ticket, prop_name);
if (crm_str_eq(prop_name, XML_ATTR_ID, TRUE)) {
continue;
}
g_hash_table_replace(ticket->state, strdup(prop_name), strdup(prop_value));
}
granted = g_hash_table_lookup(ticket->state, "granted");
if (granted && crm_is_true(granted)) {
ticket->granted = TRUE;
crm_info("We have ticket '%s'", ticket->id);
} else {
ticket->granted = FALSE;
crm_info("We do not have ticket '%s'", ticket->id);
}
last_granted = g_hash_table_lookup(ticket->state, "last-granted");
if (last_granted) {
ticket->last_granted = crm_parse_int(last_granted, 0);
}
standby = g_hash_table_lookup(ticket->state, "standby");
if (standby && crm_is_true(standby)) {
ticket->standby = TRUE;
if (ticket->granted) {
crm_info("Granted ticket '%s' is in standby-mode", ticket->id);
}
} else {
ticket->standby = FALSE;
}
crm_trace("Done with ticket state for %s", ticket_id);
return TRUE;
}
static gboolean
unpack_tickets_state(xmlNode * xml_tickets, pe_working_set_t * data_set)
{
xmlNode *xml_obj = NULL;
for (xml_obj = __xml_first_child_element(xml_tickets); xml_obj != NULL;
xml_obj = __xml_next_element(xml_obj)) {
if (crm_str_eq((const char *)xml_obj->name, XML_CIB_TAG_TICKET_STATE, TRUE) == FALSE) {
continue;
}
unpack_ticket_state(xml_obj, data_set);
}
return TRUE;
}
static void
unpack_handle_remote_attrs(pe_node_t *this_node, xmlNode *state, pe_working_set_t * data_set)
{
const char *resource_discovery_enabled = NULL;
xmlNode *attrs = NULL;
pe_resource_t *rsc = NULL;
if (crm_str_eq((const char *)state->name, XML_CIB_TAG_STATE, TRUE) == FALSE) {
return;
}
if ((this_node == NULL) || !pe__is_guest_or_remote_node(this_node)) {
return;
}
crm_trace("Processing remote node id=%s, uname=%s", this_node->details->id, this_node->details->uname);
this_node->details->remote_maintenance =
crm_atoi(crm_element_value(state, XML_NODE_IS_MAINTENANCE), "0");
rsc = this_node->details->remote_rsc;
if (this_node->details->remote_requires_reset == FALSE) {
this_node->details->unclean = FALSE;
this_node->details->unseen = FALSE;
}
attrs = find_xml_node(state, XML_TAG_TRANSIENT_NODEATTRS, FALSE);
add_node_attrs(attrs, this_node, TRUE, data_set);
if (pe__shutdown_requested(this_node)) {
crm_info("Node %s is shutting down", this_node->details->uname);
this_node->details->shutdown = TRUE;
if (rsc) {
rsc->next_role = RSC_ROLE_STOPPED;
}
}
if (crm_is_true(pe_node_attribute_raw(this_node, "standby"))) {
crm_info("Node %s is in standby-mode", this_node->details->uname);
this_node->details->standby = TRUE;
}
if (crm_is_true(pe_node_attribute_raw(this_node, "maintenance")) ||
(rsc && !is_set(rsc->flags, pe_rsc_managed))) {
crm_info("Node %s is in maintenance-mode", this_node->details->uname);
this_node->details->maintenance = TRUE;
}
resource_discovery_enabled = pe_node_attribute_raw(this_node, XML_NODE_ATTR_RSC_DISCOVERY);
if (resource_discovery_enabled && !crm_is_true(resource_discovery_enabled)) {
if (pe__is_remote_node(this_node)
&& is_not_set(data_set->flags, pe_flag_stonith_enabled)) {
crm_warn("Ignoring %s attribute on remote node %s because stonith is disabled",
XML_NODE_ATTR_RSC_DISCOVERY, this_node->details->uname);
} else {
/* This is either a remote node with fencing enabled, or a guest
* node. We don't care whether fencing is enabled when fencing guest
* nodes, because they are "fenced" by recovering their containing
* resource.
*/
crm_info("Node %s has resource discovery disabled", this_node->details->uname);
this_node->details->rsc_discovery_enabled = FALSE;
}
}
}
static bool
unpack_node_loop(xmlNode * status, bool fence, pe_working_set_t * data_set)
{
bool changed = false;
xmlNode *lrm_rsc = NULL;
for (xmlNode *state = __xml_first_child_element(status); state != NULL;
state = __xml_next_element(state)) {
const char *id = NULL;
const char *uname = NULL;
pe_node_t *this_node = NULL;
bool process = FALSE;
if (crm_str_eq((const char *)state->name, XML_CIB_TAG_STATE, TRUE) == FALSE) {
continue;
}
id = crm_element_value(state, XML_ATTR_ID);
uname = crm_element_value(state, XML_ATTR_UNAME);
this_node = pe_find_node_any(data_set->nodes, id, uname);
if (this_node == NULL) {
crm_info("Node %s is unknown", id);
continue;
} else if (this_node->details->unpacked) {
crm_info("Node %s is already processed", id);
continue;
} else if (!pe__is_guest_or_remote_node(this_node)
&& is_set(data_set->flags, pe_flag_stonith_enabled)) {
// A redundant test, but preserves the order for regression tests
process = TRUE;
} else if (pe__is_guest_or_remote_node(this_node)) {
bool check = FALSE;
pe_resource_t *rsc = this_node->details->remote_rsc;
if(fence) {
check = TRUE;
} else if(rsc == NULL) {
/* Not ready yet */
} else if (pe__is_guest_node(this_node)
&& rsc->role == RSC_ROLE_STARTED
&& rsc->container->role == RSC_ROLE_STARTED) {
/* Both the connection and its containing resource need to be
* known to be up before we process resources running in it.
*/
check = TRUE;
crm_trace("Checking node %s/%s/%s status %d/%d/%d", id, rsc->id, rsc->container->id, fence, rsc->role, RSC_ROLE_STARTED);
} else if (!pe__is_guest_node(this_node)
&& ((rsc->role == RSC_ROLE_STARTED)
|| is_set(data_set->flags, pe_flag_shutdown_lock))) {
check = TRUE;
crm_trace("Checking node %s/%s status %d/%d/%d", id, rsc->id, fence, rsc->role, RSC_ROLE_STARTED);
}
if (check) {
determine_remote_online_status(data_set, this_node);
unpack_handle_remote_attrs(this_node, state, data_set);
process = TRUE;
}
} else if (this_node->details->online) {
process = TRUE;
} else if (fence) {
process = TRUE;
} else if (is_set(data_set->flags, pe_flag_shutdown_lock)) {
process = TRUE;
}
if(process) {
crm_trace("Processing lrm resource entries on %shealthy%s node: %s",
fence?"un":"",
(pe__is_guest_or_remote_node(this_node)? " remote" : ""),
this_node->details->uname);
changed = TRUE;
this_node->details->unpacked = TRUE;
lrm_rsc = find_xml_node(state, XML_CIB_TAG_LRM, FALSE);
lrm_rsc = find_xml_node(lrm_rsc, XML_LRM_TAG_RESOURCES, FALSE);
unpack_lrm_resources(this_node, lrm_rsc, data_set);
}
}
return changed;
}
/* remove nodes that are down, stopping */
/* create positive rsc_to_node constraints between resources and the nodes they are running on */
/* anything else? */
gboolean
unpack_status(xmlNode * status, pe_working_set_t * data_set)
{
const char *id = NULL;
const char *uname = NULL;
xmlNode *state = NULL;
pe_node_t *this_node = NULL;
crm_trace("Beginning unpack");
if (data_set->tickets == NULL) {
data_set->tickets = g_hash_table_new_full(crm_str_hash, g_str_equal,
free, destroy_ticket);
}
for (state = __xml_first_child_element(status); state != NULL;
state = __xml_next_element(state)) {
if (crm_str_eq((const char *)state->name, XML_CIB_TAG_TICKETS, TRUE)) {
unpack_tickets_state((xmlNode *) state, data_set);
} else if (crm_str_eq((const char *)state->name, XML_CIB_TAG_STATE, TRUE)) {
xmlNode *attrs = NULL;
const char *resource_discovery_enabled = NULL;
id = crm_element_value(state, XML_ATTR_ID);
uname = crm_element_value(state, XML_ATTR_UNAME);
this_node = pe_find_node_any(data_set->nodes, id, uname);
if (uname == NULL) {
/* error */
continue;
} else if (this_node == NULL) {
pcmk__config_warn("Ignoring recorded node status for '%s' "
"because no longer in configuration", uname);
continue;
} else if (pe__is_guest_or_remote_node(this_node)) {
/* online state for remote nodes is determined by the
* rsc state after all the unpacking is done. we do however
* need to mark whether or not the node has been fenced as this plays
* a role during unpacking cluster node resource state */
this_node->details->remote_was_fenced =
crm_atoi(crm_element_value(state, XML_NODE_IS_FENCED), "0");
continue;
}
crm_trace("Processing node id=%s, uname=%s", id, uname);
/* Mark the node as provisionally clean
* - at least we have seen it in the current cluster's lifetime
*/
this_node->details->unclean = FALSE;
this_node->details->unseen = FALSE;
attrs = find_xml_node(state, XML_TAG_TRANSIENT_NODEATTRS, FALSE);
add_node_attrs(attrs, this_node, TRUE, data_set);
if (crm_is_true(pe_node_attribute_raw(this_node, "standby"))) {
crm_info("Node %s is in standby-mode", this_node->details->uname);
this_node->details->standby = TRUE;
}
if (crm_is_true(pe_node_attribute_raw(this_node, "maintenance"))) {
crm_info("Node %s is in maintenance-mode", this_node->details->uname);
this_node->details->maintenance = TRUE;
}
resource_discovery_enabled = pe_node_attribute_raw(this_node, XML_NODE_ATTR_RSC_DISCOVERY);
if (resource_discovery_enabled && !crm_is_true(resource_discovery_enabled)) {
crm_warn("ignoring %s attribute on node %s, disabling resource discovery is not allowed on cluster nodes",
XML_NODE_ATTR_RSC_DISCOVERY, this_node->details->uname);
}
crm_trace("determining node state");
determine_online_status(state, this_node, data_set);
if (is_not_set(data_set->flags, pe_flag_have_quorum)
&& this_node->details->online
&& (data_set->no_quorum_policy == no_quorum_suicide)) {
/* Everything else should flow from this automatically
* (at least until the scheduler becomes able to migrate off
* healthy resources)
*/
- pe_fence_node(data_set, this_node, "cluster does not have quorum");
+ pe_fence_node(data_set, this_node, "cluster does not have quorum", FALSE);
}
}
}
while(unpack_node_loop(status, FALSE, data_set)) {
crm_trace("Start another loop");
}
// Now catch any nodes we didn't see
unpack_node_loop(status, is_set(data_set->flags, pe_flag_stonith_enabled), data_set);
/* Now that we know where resources are, we can schedule stops of containers
* with failed bundle connections
*/
if (data_set->stop_needed != NULL) {
for (GList *item = data_set->stop_needed; item; item = item->next) {
pe_resource_t *container = item->data;
pe_node_t *node = pe__current_node(container);
if (node) {
stop_action(container, node, FALSE);
}
}
g_list_free(data_set->stop_needed);
data_set->stop_needed = NULL;
}
for (GListPtr gIter = data_set->nodes; gIter != NULL; gIter = gIter->next) {
pe_node_t *this_node = gIter->data;
if (this_node == NULL) {
continue;
} else if (!pe__is_guest_or_remote_node(this_node)) {
continue;
} else if(this_node->details->unpacked) {
continue;
}
determine_remote_online_status(data_set, this_node);
}
return TRUE;
}
static gboolean
determine_online_status_no_fencing(pe_working_set_t * data_set, xmlNode * node_state,
pe_node_t * this_node)
{
gboolean online = FALSE;
const char *join = crm_element_value(node_state, XML_NODE_JOIN_STATE);
const char *is_peer = crm_element_value(node_state, XML_NODE_IS_PEER);
const char *in_cluster = crm_element_value(node_state, XML_NODE_IN_CLUSTER);
const char *exp_state = crm_element_value(node_state, XML_NODE_EXPECTED);
if (!crm_is_true(in_cluster)) {
crm_trace("Node is down: in_cluster=%s", crm_str(in_cluster));
} else if (safe_str_eq(is_peer, ONLINESTATUS)) {
if (safe_str_eq(join, CRMD_JOINSTATE_MEMBER)) {
online = TRUE;
} else {
crm_debug("Node is not ready to run resources: %s", join);
}
} else if (this_node->details->expected_up == FALSE) {
crm_trace("Controller is down: in_cluster=%s", crm_str(in_cluster));
crm_trace("\tis_peer=%s, join=%s, expected=%s",
crm_str(is_peer), crm_str(join), crm_str(exp_state));
} else {
/* mark it unclean */
- pe_fence_node(data_set, this_node, "peer is unexpectedly down");
+ pe_fence_node(data_set, this_node, "peer is unexpectedly down", FALSE);
crm_info("\tin_cluster=%s, is_peer=%s, join=%s, expected=%s",
crm_str(in_cluster), crm_str(is_peer), crm_str(join), crm_str(exp_state));
}
return online;
}
static gboolean
determine_online_status_fencing(pe_working_set_t * data_set, xmlNode * node_state,
pe_node_t * this_node)
{
gboolean online = FALSE;
gboolean do_terminate = FALSE;
bool crmd_online = FALSE;
const char *join = crm_element_value(node_state, XML_NODE_JOIN_STATE);
const char *is_peer = crm_element_value(node_state, XML_NODE_IS_PEER);
const char *in_cluster = crm_element_value(node_state, XML_NODE_IN_CLUSTER);
const char *exp_state = crm_element_value(node_state, XML_NODE_EXPECTED);
const char *terminate = pe_node_attribute_raw(this_node, "terminate");
/*
- XML_NODE_IN_CLUSTER ::= true|false
- XML_NODE_IS_PEER ::= online|offline
- XML_NODE_JOIN_STATE ::= member|down|pending|banned
- XML_NODE_EXPECTED ::= member|down
*/
if (crm_is_true(terminate)) {
do_terminate = TRUE;
} else if (terminate != NULL && strlen(terminate) > 0) {
/* could be a time() value */
char t = terminate[0];
if (t != '0' && isdigit(t)) {
do_terminate = TRUE;
}
}
crm_trace("%s: in_cluster=%s, is_peer=%s, join=%s, expected=%s, term=%d",
this_node->details->uname, crm_str(in_cluster), crm_str(is_peer),
crm_str(join), crm_str(exp_state), do_terminate);
online = crm_is_true(in_cluster);
crmd_online = safe_str_eq(is_peer, ONLINESTATUS);
if (exp_state == NULL) {
exp_state = CRMD_JOINSTATE_DOWN;
}
if (this_node->details->shutdown) {
crm_debug("%s is shutting down", this_node->details->uname);
/* Slightly different criteria since we can't shut down a dead peer */
online = crmd_online;
} else if (in_cluster == NULL) {
- pe_fence_node(data_set, this_node, "peer has not been seen by the cluster");
+ pe_fence_node(data_set, this_node, "peer has not been seen by the cluster", FALSE);
} else if (safe_str_eq(join, CRMD_JOINSTATE_NACK)) {
- pe_fence_node(data_set, this_node, "peer failed the pacemaker membership criteria");
+ pe_fence_node(data_set, this_node, "peer failed the pacemaker membership criteria", FALSE);
} else if (do_terminate == FALSE && safe_str_eq(exp_state, CRMD_JOINSTATE_DOWN)) {
if (crm_is_true(in_cluster) || crmd_online) {
crm_info("- Node %s is not ready to run resources", this_node->details->uname);
this_node->details->standby = TRUE;
this_node->details->pending = TRUE;
} else {
crm_trace("%s is down or still coming up", this_node->details->uname);
}
} else if (do_terminate && safe_str_eq(join, CRMD_JOINSTATE_DOWN)
&& crm_is_true(in_cluster) == FALSE && !crmd_online) {
crm_info("Node %s was just shot", this_node->details->uname);
online = FALSE;
} else if (crm_is_true(in_cluster) == FALSE) {
- pe_fence_node(data_set, this_node, "peer is no longer part of the cluster");
+ // Consider `priority-fencing-delay` for lost nodes
+ pe_fence_node(data_set, this_node, "peer is no longer part of the cluster", TRUE);
} else if (!crmd_online) {
- pe_fence_node(data_set, this_node, "peer process is no longer available");
+ pe_fence_node(data_set, this_node, "peer process is no longer available", FALSE);
/* Everything is running at this point, now check join state */
} else if (do_terminate) {
- pe_fence_node(data_set, this_node, "termination was requested");
+ pe_fence_node(data_set, this_node, "termination was requested", FALSE);
} else if (safe_str_eq(join, CRMD_JOINSTATE_MEMBER)) {
crm_info("Node %s is active", this_node->details->uname);
} else if (safe_str_eq(join, CRMD_JOINSTATE_PENDING)
|| safe_str_eq(join, CRMD_JOINSTATE_DOWN)) {
crm_info("Node %s is not ready to run resources", this_node->details->uname);
this_node->details->standby = TRUE;
this_node->details->pending = TRUE;
} else {
- pe_fence_node(data_set, this_node, "peer was in an unknown state");
+ pe_fence_node(data_set, this_node, "peer was in an unknown state", FALSE);
crm_warn("%s: in-cluster=%s, is-peer=%s, join=%s, expected=%s, term=%d, shutdown=%d",
this_node->details->uname, crm_str(in_cluster), crm_str(is_peer),
crm_str(join), crm_str(exp_state), do_terminate, this_node->details->shutdown);
}
return online;
}
static gboolean
determine_remote_online_status(pe_working_set_t * data_set, pe_node_t * this_node)
{
pe_resource_t *rsc = this_node->details->remote_rsc;
pe_resource_t *container = NULL;
pe_node_t *host = NULL;
/* If there is a node state entry for a (former) Pacemaker Remote node
* but no resource creating that node, the node's connection resource will
* be NULL. Consider it an offline remote node in that case.
*/
if (rsc == NULL) {
this_node->details->online = FALSE;
goto remote_online_done;
}
container = rsc->container;
if (container && pcmk__list_of_1(rsc->running_on)) {
host = rsc->running_on->data;
}
/* If the resource is currently started, mark it online. */
if (rsc->role == RSC_ROLE_STARTED) {
crm_trace("%s node %s presumed ONLINE because connection resource is started",
(container? "Guest" : "Remote"), this_node->details->id);
this_node->details->online = TRUE;
}
/* consider this node shutting down if transitioning start->stop */
if (rsc->role == RSC_ROLE_STARTED && rsc->next_role == RSC_ROLE_STOPPED) {
crm_trace("%s node %s shutting down because connection resource is stopping",
(container? "Guest" : "Remote"), this_node->details->id);
this_node->details->shutdown = TRUE;
}
/* Now check all the failure conditions. */
if(container && is_set(container->flags, pe_rsc_failed)) {
crm_trace("Guest node %s UNCLEAN because guest resource failed",
this_node->details->id);
this_node->details->online = FALSE;
this_node->details->remote_requires_reset = TRUE;
} else if(is_set(rsc->flags, pe_rsc_failed)) {
crm_trace("%s node %s OFFLINE because connection resource failed",
(container? "Guest" : "Remote"), this_node->details->id);
this_node->details->online = FALSE;
} else if (rsc->role == RSC_ROLE_STOPPED
|| (container && container->role == RSC_ROLE_STOPPED)) {
crm_trace("%s node %s OFFLINE because its resource is stopped",
(container? "Guest" : "Remote"), this_node->details->id);
this_node->details->online = FALSE;
this_node->details->remote_requires_reset = FALSE;
} else if (host && (host->details->online == FALSE)
&& host->details->unclean) {
crm_trace("Guest node %s UNCLEAN because host is unclean",
this_node->details->id);
this_node->details->online = FALSE;
this_node->details->remote_requires_reset = TRUE;
}
remote_online_done:
crm_trace("Remote node %s online=%s",
this_node->details->id, this_node->details->online ? "TRUE" : "FALSE");
return this_node->details->online;
}
gboolean
determine_online_status(xmlNode * node_state, pe_node_t * this_node, pe_working_set_t * data_set)
{
gboolean online = FALSE;
const char *exp_state = crm_element_value(node_state, XML_NODE_EXPECTED);
CRM_CHECK(this_node != NULL, return FALSE);
this_node->details->shutdown = FALSE;
this_node->details->expected_up = FALSE;
if (pe__shutdown_requested(this_node)) {
this_node->details->shutdown = TRUE;
} else if (safe_str_eq(exp_state, CRMD_JOINSTATE_MEMBER)) {
this_node->details->expected_up = TRUE;
}
if (this_node->details->type == node_ping) {
this_node->details->unclean = FALSE;
online = FALSE; /* As far as resource management is concerned,
* the node is safely offline.
* Anyone caught abusing this logic will be shot
*/
} else if (is_set(data_set->flags, pe_flag_stonith_enabled) == FALSE) {
online = determine_online_status_no_fencing(data_set, node_state, this_node);
} else {
online = determine_online_status_fencing(data_set, node_state, this_node);
}
if (online) {
this_node->details->online = TRUE;
} else {
/* remove node from contention */
this_node->fixed = TRUE;
this_node->weight = -INFINITY;
}
if (online && this_node->details->shutdown) {
/* don't run resources here */
this_node->fixed = TRUE;
this_node->weight = -INFINITY;
}
if (this_node->details->type == node_ping) {
crm_info("Node %s is not a pacemaker node", this_node->details->uname);
} else if (this_node->details->unclean) {
pe_proc_warn("Node %s is unclean", this_node->details->uname);
} else if (this_node->details->online) {
crm_info("Node %s is %s", this_node->details->uname,
this_node->details->shutdown ? "shutting down" :
this_node->details->pending ? "pending" :
this_node->details->standby ? "standby" :
this_node->details->maintenance ? "maintenance" : "online");
} else {
crm_trace("Node %s is offline", this_node->details->uname);
}
return online;
}
/*!
* \internal
* \brief Find the end of a resource's name, excluding any clone suffix
*
* \param[in] id Resource ID to check
*
* \return Pointer to last character of resource's base name
*/
const char *
pe_base_name_end(const char *id)
{
if (!pcmk__str_empty(id)) {
const char *end = id + strlen(id) - 1;
for (const char *s = end; s > id; --s) {
switch (*s) {
case '0':
case '1':
case '2':
case '3':
case '4':
case '5':
case '6':
case '7':
case '8':
case '9':
break;
case ':':
return (s == end)? s : (s - 1);
default:
return end;
}
}
return end;
}
return NULL;
}
/*!
* \internal
* \brief Get a resource name excluding any clone suffix
*
* \param[in] last_rsc_id Resource ID to check
*
* \return Pointer to newly allocated string with resource's base name
* \note It is the caller's responsibility to free() the result.
* This asserts on error, so callers can assume result is not NULL.
*/
char *
clone_strip(const char *last_rsc_id)
{
const char *end = pe_base_name_end(last_rsc_id);
char *basename = NULL;
CRM_ASSERT(end);
basename = strndup(last_rsc_id, end - last_rsc_id + 1);
CRM_ASSERT(basename);
return basename;
}
/*!
* \internal
* \brief Get the name of the first instance of a cloned resource
*
* \param[in] last_rsc_id Resource ID to check
*
* \return Pointer to newly allocated string with resource's base name plus :0
* \note It is the caller's responsibility to free() the result.
* This asserts on error, so callers can assume result is not NULL.
*/
char *
clone_zero(const char *last_rsc_id)
{
const char *end = pe_base_name_end(last_rsc_id);
size_t base_name_len = end - last_rsc_id + 1;
char *zero = NULL;
CRM_ASSERT(end);
zero = calloc(base_name_len + 3, sizeof(char));
CRM_ASSERT(zero);
memcpy(zero, last_rsc_id, base_name_len);
zero[base_name_len] = ':';
zero[base_name_len + 1] = '0';
return zero;
}
static pe_resource_t *
create_fake_resource(const char *rsc_id, xmlNode * rsc_entry, pe_working_set_t * data_set)
{
pe_resource_t *rsc = NULL;
xmlNode *xml_rsc = create_xml_node(NULL, XML_CIB_TAG_RESOURCE);
copy_in_properties(xml_rsc, rsc_entry);
crm_xml_add(xml_rsc, XML_ATTR_ID, rsc_id);
crm_log_xml_debug(xml_rsc, "Orphan resource");
if (!common_unpack(xml_rsc, &rsc, NULL, data_set)) {
return NULL;
}
if (xml_contains_remote_node(xml_rsc)) {
pe_node_t *node;
crm_debug("Detected orphaned remote node %s", rsc_id);
node = pe_find_node(data_set->nodes, rsc_id);
if (node == NULL) {
node = pe_create_node(rsc_id, rsc_id, "remote", NULL, data_set);
}
link_rsc2remotenode(data_set, rsc);
if (node) {
crm_trace("Setting node %s as shutting down due to orphaned connection resource", rsc_id);
node->details->shutdown = TRUE;
}
}
if (crm_element_value(rsc_entry, XML_RSC_ATTR_CONTAINER)) {
/* This orphaned rsc needs to be mapped to a container. */
crm_trace("Detected orphaned container filler %s", rsc_id);
set_bit(rsc->flags, pe_rsc_orphan_container_filler);
}
set_bit(rsc->flags, pe_rsc_orphan);
data_set->resources = g_list_append(data_set->resources, rsc);
return rsc;
}
/*!
* \internal
* \brief Create orphan instance for anonymous clone resource history
*/
static pe_resource_t *
create_anonymous_orphan(pe_resource_t *parent, const char *rsc_id,
pe_node_t *node, pe_working_set_t *data_set)
{
pe_resource_t *top = pe__create_clone_child(parent, data_set);
// find_rsc() because we might be a cloned group
pe_resource_t *orphan = top->fns->find_rsc(top, rsc_id, NULL, pe_find_clone);
pe_rsc_debug(parent, "Created orphan %s for %s: %s on %s",
top->id, parent->id, rsc_id, node->details->uname);
return orphan;
}
/*!
* \internal
* \brief Check a node for an instance of an anonymous clone
*
* Return a child instance of the specified anonymous clone, in order of
* preference: (1) the instance running on the specified node, if any;
* (2) an inactive instance (i.e. within the total of clone-max instances);
* (3) a newly created orphan (i.e. clone-max instances are already active).
*
* \param[in] data_set Cluster information
* \param[in] node Node on which to check for instance
* \param[in] parent Clone to check
* \param[in] rsc_id Name of cloned resource in history (without instance)
*/
static pe_resource_t *
find_anonymous_clone(pe_working_set_t * data_set, pe_node_t * node, pe_resource_t * parent,
const char *rsc_id)
{
GListPtr rIter = NULL;
pe_resource_t *rsc = NULL;
pe_resource_t *inactive_instance = NULL;
gboolean skip_inactive = FALSE;
CRM_ASSERT(parent != NULL);
CRM_ASSERT(pe_rsc_is_clone(parent));
CRM_ASSERT(is_not_set(parent->flags, pe_rsc_unique));
// Check for active (or partially active, for cloned groups) instance
pe_rsc_trace(parent, "Looking for %s on %s in %s", rsc_id, node->details->uname, parent->id);
for (rIter = parent->children; rsc == NULL && rIter; rIter = rIter->next) {
GListPtr locations = NULL;
pe_resource_t *child = rIter->data;
/* Check whether this instance is already known to be active or pending
* anywhere, at this stage of unpacking. Because this function is called
* for a resource before the resource's individual operation history
* entries are unpacked, locations will generally not contain the
* desired node.
*
* However, there are three exceptions:
* (1) when child is a cloned group and we have already unpacked the
* history of another member of the group on the same node;
* (2) when we've already unpacked the history of another numbered
* instance on the same node (which can happen if globally-unique
* was flipped from true to false); and
* (3) when we re-run calculations on the same data set as part of a
* simulation.
*/
child->fns->location(child, &locations, 2);
if (locations) {
/* We should never associate the same numbered anonymous clone
* instance with multiple nodes, and clone instances can't migrate,
* so there must be only one location, regardless of history.
*/
CRM_LOG_ASSERT(locations->next == NULL);
if (((pe_node_t *)locations->data)->details == node->details) {
/* This child instance is active on the requested node, so check
* for a corresponding configured resource. We use find_rsc()
* instead of child because child may be a cloned group, and we
* need the particular member corresponding to rsc_id.
*
* If the history entry is orphaned, rsc will be NULL.
*/
rsc = parent->fns->find_rsc(child, rsc_id, NULL, pe_find_clone);
if (rsc) {
/* If there are multiple instance history entries for an
* anonymous clone in a single node's history (which can
* happen if globally-unique is switched from true to
* false), we want to consider the instances beyond the
* first as orphans, even if there are inactive instance
* numbers available.
*/
if (rsc->running_on) {
crm_notice("Active (now-)anonymous clone %s has "
"multiple (orphan) instance histories on %s",
parent->id, node->details->uname);
skip_inactive = TRUE;
rsc = NULL;
} else {
pe_rsc_trace(parent, "Resource %s, active", rsc->id);
}
}
}
g_list_free(locations);
} else {
pe_rsc_trace(parent, "Resource %s, skip inactive", child->id);
if (!skip_inactive && !inactive_instance
&& is_not_set(child->flags, pe_rsc_block)) {
// Remember one inactive instance in case we don't find active
inactive_instance = parent->fns->find_rsc(child, rsc_id, NULL,
pe_find_clone);
/* ... but don't use it if it was already associated with a
* pending action on another node
*/
if (inactive_instance && inactive_instance->pending_node
&& (inactive_instance->pending_node->details != node->details)) {
inactive_instance = NULL;
}
}
}
}
if ((rsc == NULL) && !skip_inactive && (inactive_instance != NULL)) {
pe_rsc_trace(parent, "Resource %s, empty slot", inactive_instance->id);
rsc = inactive_instance;
}
/* If the resource has "requires" set to "quorum" or "nothing", and we don't
* have a clone instance for every node, we don't want to consume a valid
* instance number for unclean nodes. Such instances may appear to be active
* according to the history, but should be considered inactive, so we can
* start an instance elsewhere. Treat such instances as orphans.
*
* An exception is instances running on guest nodes -- since guest node
* "fencing" is actually just a resource stop, requires shouldn't apply.
*
* @TODO Ideally, we'd use an inactive instance number if it is not needed
* for any clean instances. However, we don't know that at this point.
*/
if ((rsc != NULL) && is_not_set(rsc->flags, pe_rsc_needs_fencing)
&& (!node->details->online || node->details->unclean)
&& !pe__is_guest_node(node)
&& !pe__is_universal_clone(parent, data_set)) {
rsc = NULL;
}
if (rsc == NULL) {
rsc = create_anonymous_orphan(parent, rsc_id, node, data_set);
pe_rsc_trace(parent, "Resource %s, orphan", rsc->id);
}
return rsc;
}
static pe_resource_t *
unpack_find_resource(pe_working_set_t * data_set, pe_node_t * node, const char *rsc_id,
xmlNode * rsc_entry)
{
pe_resource_t *rsc = NULL;
pe_resource_t *parent = NULL;
crm_trace("looking for %s", rsc_id);
rsc = pe_find_resource(data_set->resources, rsc_id);
if (rsc == NULL) {
/* If we didn't find the resource by its name in the operation history,
* check it again as a clone instance. Even when clone-max=0, we create
* a single :0 orphan to match against here.
*/
char *clone0_id = clone_zero(rsc_id);
pe_resource_t *clone0 = pe_find_resource(data_set->resources, clone0_id);
if (clone0 && is_not_set(clone0->flags, pe_rsc_unique)) {
rsc = clone0;
parent = uber_parent(clone0);
crm_trace("%s found as %s (%s)", rsc_id, clone0_id, parent->id);
} else {
crm_trace("%s is not known as %s either (orphan)",
rsc_id, clone0_id);
}
free(clone0_id);
} else if (rsc->variant > pe_native) {
crm_trace("Resource history for %s is orphaned because it is no longer primitive",
rsc_id);
return NULL;
} else {
parent = uber_parent(rsc);
}
if (pe_rsc_is_anon_clone(parent)) {
if (pe_rsc_is_bundled(parent)) {
rsc = pe__find_bundle_replica(parent->parent, node);
} else {
char *base = clone_strip(rsc_id);
rsc = find_anonymous_clone(data_set, node, parent, base);
free(base);
CRM_ASSERT(rsc != NULL);
}
}
if (rsc && safe_str_neq(rsc_id, rsc->id)
&& safe_str_neq(rsc_id, rsc->clone_name)) {
free(rsc->clone_name);
rsc->clone_name = strdup(rsc_id);
pe_rsc_debug(rsc, "Internally renamed %s on %s to %s%s",
rsc_id, node->details->uname, rsc->id,
(is_set(rsc->flags, pe_rsc_orphan)? " (ORPHAN)" : ""));
}
return rsc;
}
static pe_resource_t *
process_orphan_resource(xmlNode * rsc_entry, pe_node_t * node, pe_working_set_t * data_set)
{
pe_resource_t *rsc = NULL;
const char *rsc_id = crm_element_value(rsc_entry, XML_ATTR_ID);
crm_debug("Detected orphan resource %s on %s", rsc_id, node->details->uname);
rsc = create_fake_resource(rsc_id, rsc_entry, data_set);
if (is_set(data_set->flags, pe_flag_stop_rsc_orphans) == FALSE) {
clear_bit(rsc->flags, pe_rsc_managed);
} else {
CRM_CHECK(rsc != NULL, return NULL);
pe_rsc_trace(rsc, "Added orphan %s", rsc->id);
resource_location(rsc, NULL, -INFINITY, "__orphan_do_not_run__", data_set);
}
return rsc;
}
static void
process_rsc_state(pe_resource_t * rsc, pe_node_t * node,
enum action_fail_response on_fail,
xmlNode * migrate_op, pe_working_set_t * data_set)
{
pe_node_t *tmpnode = NULL;
char *reason = NULL;
CRM_ASSERT(rsc);
pe_rsc_trace(rsc, "Resource %s is %s on %s: on_fail=%s",
rsc->id, role2text(rsc->role), node->details->uname, fail2text(on_fail));
/* process current state */
if (rsc->role != RSC_ROLE_UNKNOWN) {
pe_resource_t *iter = rsc;
while (iter) {
if (g_hash_table_lookup(iter->known_on, node->details->id) == NULL) {
pe_node_t *n = node_copy(node);
pe_rsc_trace(rsc, "%s (aka. %s) known on %s", rsc->id, rsc->clone_name,
n->details->uname);
g_hash_table_insert(iter->known_on, (gpointer) n->details->id, n);
}
if (is_set(iter->flags, pe_rsc_unique)) {
break;
}
iter = iter->parent;
}
}
/* If a managed resource is believed to be running, but node is down ... */
if (rsc->role > RSC_ROLE_STOPPED
&& node->details->online == FALSE
&& node->details->maintenance == FALSE
&& is_set(rsc->flags, pe_rsc_managed)) {
gboolean should_fence = FALSE;
/* If this is a guest node, fence it (regardless of whether fencing is
* enabled, because guest node fencing is done by recovery of the
* container resource rather than by the fencer). Mark the resource
* we're processing as failed. When the guest comes back up, its
* operation history in the CIB will be cleared, freeing the affected
* resource to run again once we are sure we know its state.
*/
if (pe__is_guest_node(node)) {
set_bit(rsc->flags, pe_rsc_failed);
should_fence = TRUE;
} else if (is_set(data_set->flags, pe_flag_stonith_enabled)) {
if (pe__is_remote_node(node) && node->details->remote_rsc
&& is_not_set(node->details->remote_rsc->flags, pe_rsc_failed)) {
/* Setting unseen means that fencing of the remote node will
* occur only if the connection resource is not going to start
* somewhere. This allows connection resources on a failed
* cluster node to move to another node without requiring the
* remote nodes to be fenced as well.
*/
node->details->unseen = TRUE;
reason = crm_strdup_printf("%s is active there (fencing will be"
" revoked if remote connection can "
"be re-established elsewhere)",
rsc->id);
}
should_fence = TRUE;
}
if (should_fence) {
if (reason == NULL) {
reason = crm_strdup_printf("%s is thought to be active there", rsc->id);
}
- pe_fence_node(data_set, node, reason);
+ pe_fence_node(data_set, node, reason, FALSE);
}
free(reason);
}
if (node->details->unclean) {
/* No extra processing needed
* Also allows resources to be started again after a node is shot
*/
on_fail = action_fail_ignore;
}
switch (on_fail) {
case action_fail_ignore:
/* nothing to do */
break;
case action_fail_fence:
/* treat it as if it is still running
* but also mark the node as unclean
*/
reason = crm_strdup_printf("%s failed there", rsc->id);
- pe_fence_node(data_set, node, reason);
+ pe_fence_node(data_set, node, reason, FALSE);
free(reason);
break;
case action_fail_standby:
node->details->standby = TRUE;
node->details->standby_onfail = TRUE;
break;
case action_fail_block:
/* is_managed == FALSE will prevent any
* actions being sent for the resource
*/
clear_bit(rsc->flags, pe_rsc_managed);
set_bit(rsc->flags, pe_rsc_block);
break;
case action_fail_migrate:
/* make sure it comes up somewhere else
* or not at all
*/
resource_location(rsc, node, -INFINITY, "__action_migration_auto__", data_set);
break;
case action_fail_stop:
rsc->next_role = RSC_ROLE_STOPPED;
break;
case action_fail_recover:
if (rsc->role != RSC_ROLE_STOPPED && rsc->role != RSC_ROLE_UNKNOWN) {
set_bit(rsc->flags, pe_rsc_failed);
stop_action(rsc, node, FALSE);
}
break;
case action_fail_restart_container:
set_bit(rsc->flags, pe_rsc_failed);
if (rsc->container && pe_rsc_is_bundled(rsc)) {
/* A bundle's remote connection can run on a different node than
* the bundle's container. We don't necessarily know where the
* container is running yet, so remember it and add a stop
* action for it later.
*/
data_set->stop_needed = g_list_prepend(data_set->stop_needed,
rsc->container);
} else if (rsc->container) {
stop_action(rsc->container, node, FALSE);
} else if (rsc->role != RSC_ROLE_STOPPED && rsc->role != RSC_ROLE_UNKNOWN) {
stop_action(rsc, node, FALSE);
}
break;
case action_fail_reset_remote:
set_bit(rsc->flags, pe_rsc_failed);
if (is_set(data_set->flags, pe_flag_stonith_enabled)) {
tmpnode = NULL;
if (rsc->is_remote_node) {
tmpnode = pe_find_node(data_set->nodes, rsc->id);
}
if (tmpnode &&
pe__is_remote_node(tmpnode) &&
tmpnode->details->remote_was_fenced == 0) {
/* The remote connection resource failed in a way that
* should result in fencing the remote node.
*/
pe_fence_node(data_set, tmpnode,
- "remote connection is unrecoverable");
+ "remote connection is unrecoverable", FALSE);
}
}
/* require the stop action regardless if fencing is occurring or not. */
if (rsc->role > RSC_ROLE_STOPPED) {
stop_action(rsc, node, FALSE);
}
/* if reconnect delay is in use, prevent the connection from exiting the
* "STOPPED" role until the failure is cleared by the delay timeout. */
if (rsc->remote_reconnect_ms) {
rsc->next_role = RSC_ROLE_STOPPED;
}
break;
}
/* ensure a remote-node connection failure forces an unclean remote-node
* to be fenced. By setting unseen = FALSE, the remote-node failure will
* result in a fencing operation regardless if we're going to attempt to
* reconnect to the remote-node in this transition or not. */
if (is_set(rsc->flags, pe_rsc_failed) && rsc->is_remote_node) {
tmpnode = pe_find_node(data_set->nodes, rsc->id);
if (tmpnode && tmpnode->details->unclean) {
tmpnode->details->unseen = FALSE;
}
}
if (rsc->role != RSC_ROLE_STOPPED && rsc->role != RSC_ROLE_UNKNOWN) {
if (is_set(rsc->flags, pe_rsc_orphan)) {
if (is_set(rsc->flags, pe_rsc_managed)) {
pcmk__config_warn("Detected active orphan %s running on %s",
rsc->id, node->details->uname);
} else {
pcmk__config_warn("Resource '%s' must be stopped manually on "
"%s because cluster is configured not to "
"stop active orphans",
rsc->id, node->details->uname);
}
}
native_add_running(rsc, node, data_set);
if (on_fail != action_fail_ignore) {
set_bit(rsc->flags, pe_rsc_failed);
}
} else if (rsc->clone_name && strchr(rsc->clone_name, ':') != NULL) {
/* Only do this for older status sections that included instance numbers
* Otherwise stopped instances will appear as orphans
*/
pe_rsc_trace(rsc, "Resetting clone_name %s for %s (stopped)", rsc->clone_name, rsc->id);
free(rsc->clone_name);
rsc->clone_name = NULL;
} else {
GList *possible_matches = pe__resource_actions(rsc, node, RSC_STOP,
FALSE);
GListPtr gIter = possible_matches;
for (; gIter != NULL; gIter = gIter->next) {
pe_action_t *stop = (pe_action_t *) gIter->data;
stop->flags |= pe_action_optional;
}
g_list_free(possible_matches);
}
}
/* create active recurring operations as optional */
static void
process_recurring(pe_node_t * node, pe_resource_t * rsc,
int start_index, int stop_index,
GListPtr sorted_op_list, pe_working_set_t * data_set)
{
int counter = -1;
const char *task = NULL;
const char *status = NULL;
GListPtr gIter = sorted_op_list;
CRM_ASSERT(rsc);
pe_rsc_trace(rsc, "%s: Start index %d, stop index = %d", rsc->id, start_index, stop_index);
for (; gIter != NULL; gIter = gIter->next) {
xmlNode *rsc_op = (xmlNode *) gIter->data;
guint interval_ms = 0;
char *key = NULL;
const char *id = ID(rsc_op);
counter++;
if (node->details->online == FALSE) {
pe_rsc_trace(rsc, "Skipping %s/%s: node is offline", rsc->id, node->details->uname);
break;
/* Need to check if there's a monitor for role="Stopped" */
} else if (start_index < stop_index && counter <= stop_index) {
pe_rsc_trace(rsc, "Skipping %s/%s: resource is not active", id, node->details->uname);
continue;
} else if (counter < start_index) {
pe_rsc_trace(rsc, "Skipping %s/%s: old %d", id, node->details->uname, counter);
continue;
}
crm_element_value_ms(rsc_op, XML_LRM_ATTR_INTERVAL_MS, &interval_ms);
if (interval_ms == 0) {
pe_rsc_trace(rsc, "Skipping %s/%s: non-recurring", id, node->details->uname);
continue;
}
status = crm_element_value(rsc_op, XML_LRM_ATTR_OPSTATUS);
if (safe_str_eq(status, "-1")) {
pe_rsc_trace(rsc, "Skipping %s/%s: status", id, node->details->uname);
continue;
}
task = crm_element_value(rsc_op, XML_LRM_ATTR_TASK);
/* create the action */
key = pcmk__op_key(rsc->id, task, interval_ms);
pe_rsc_trace(rsc, "Creating %s/%s", key, node->details->uname);
custom_action(rsc, key, task, node, TRUE, TRUE, data_set);
}
}
void
calculate_active_ops(GListPtr sorted_op_list, int *start_index, int *stop_index)
{
int counter = -1;
int implied_monitor_start = -1;
int implied_clone_start = -1;
const char *task = NULL;
const char *status = NULL;
GListPtr gIter = sorted_op_list;
*stop_index = -1;
*start_index = -1;
for (; gIter != NULL; gIter = gIter->next) {
xmlNode *rsc_op = (xmlNode *) gIter->data;
counter++;
task = crm_element_value(rsc_op, XML_LRM_ATTR_TASK);
status = crm_element_value(rsc_op, XML_LRM_ATTR_OPSTATUS);
if (safe_str_eq(task, CRMD_ACTION_STOP)
&& safe_str_eq(status, "0")) {
*stop_index = counter;
} else if (safe_str_eq(task, CRMD_ACTION_START) || safe_str_eq(task, CRMD_ACTION_MIGRATED)) {
*start_index = counter;
} else if ((implied_monitor_start <= *stop_index) && safe_str_eq(task, CRMD_ACTION_STATUS)) {
const char *rc = crm_element_value(rsc_op, XML_LRM_ATTR_RC);
if (safe_str_eq(rc, "0") || safe_str_eq(rc, "8")) {
implied_monitor_start = counter;
}
} else if (safe_str_eq(task, CRMD_ACTION_PROMOTE) || safe_str_eq(task, CRMD_ACTION_DEMOTE)) {
implied_clone_start = counter;
}
}
if (*start_index == -1) {
if (implied_clone_start != -1) {
*start_index = implied_clone_start;
} else if (implied_monitor_start != -1) {
*start_index = implied_monitor_start;
}
}
}
// If resource history entry has shutdown lock, remember lock node and time
static void
unpack_shutdown_lock(xmlNode *rsc_entry, pe_resource_t *rsc, pe_node_t *node,
pe_working_set_t *data_set)
{
time_t lock_time = 0; // When lock started (i.e. node shutdown time)
if ((crm_element_value_epoch(rsc_entry, XML_CONFIG_ATTR_SHUTDOWN_LOCK,
&lock_time) == pcmk_ok) && (lock_time != 0)) {
if ((data_set->shutdown_lock > 0)
&& (get_effective_time(data_set)
> (lock_time + data_set->shutdown_lock))) {
pe_rsc_info(rsc, "Shutdown lock for %s on %s expired",
rsc->id, node->details->uname);
pe__clear_resource_history(rsc, node, data_set);
} else {
rsc->lock_node = node;
rsc->lock_time = lock_time;
}
}
}
static pe_resource_t *
unpack_lrm_rsc_state(pe_node_t * node, xmlNode * rsc_entry, pe_working_set_t * data_set)
{
GListPtr gIter = NULL;
int stop_index = -1;
int start_index = -1;
enum rsc_role_e req_role = RSC_ROLE_UNKNOWN;
const char *task = NULL;
const char *rsc_id = crm_element_value(rsc_entry, XML_ATTR_ID);
pe_resource_t *rsc = NULL;
GListPtr op_list = NULL;
GListPtr sorted_op_list = NULL;
xmlNode *migrate_op = NULL;
xmlNode *rsc_op = NULL;
xmlNode *last_failure = NULL;
enum action_fail_response on_fail = FALSE;
enum rsc_role_e saved_role = RSC_ROLE_UNKNOWN;
crm_trace("[%s] Processing %s on %s",
crm_element_name(rsc_entry), rsc_id, node->details->uname);
/* extract operations */
op_list = NULL;
sorted_op_list = NULL;
for (rsc_op = __xml_first_child_element(rsc_entry); rsc_op != NULL;
rsc_op = __xml_next_element(rsc_op)) {
if (crm_str_eq((const char *)rsc_op->name, XML_LRM_TAG_RSC_OP, TRUE)) {
op_list = g_list_prepend(op_list, rsc_op);
}
}
if (is_not_set(data_set->flags, pe_flag_shutdown_lock)) {
if (op_list == NULL) {
// If there are no operations, there is nothing to do
return NULL;
}
}
/* find the resource */
rsc = unpack_find_resource(data_set, node, rsc_id, rsc_entry);
if (rsc == NULL) {
if (op_list == NULL) {
// If there are no operations, there is nothing to do
return NULL;
} else {
rsc = process_orphan_resource(rsc_entry, node, data_set);
}
}
CRM_ASSERT(rsc != NULL);
// Check whether the resource is "shutdown-locked" to this node
if (is_set(data_set->flags, pe_flag_shutdown_lock)) {
unpack_shutdown_lock(rsc_entry, rsc, node, data_set);
}
/* process operations */
saved_role = rsc->role;
on_fail = action_fail_ignore;
rsc->role = RSC_ROLE_UNKNOWN;
sorted_op_list = g_list_sort(op_list, sort_op_by_callid);
for (gIter = sorted_op_list; gIter != NULL; gIter = gIter->next) {
xmlNode *rsc_op = (xmlNode *) gIter->data;
task = crm_element_value(rsc_op, XML_LRM_ATTR_TASK);
if (safe_str_eq(task, CRMD_ACTION_MIGRATED)) {
migrate_op = rsc_op;
}
unpack_rsc_op(rsc, node, rsc_op, &last_failure, &on_fail, data_set);
}
/* create active recurring operations as optional */
calculate_active_ops(sorted_op_list, &start_index, &stop_index);
process_recurring(node, rsc, start_index, stop_index, sorted_op_list, data_set);
/* no need to free the contents */
g_list_free(sorted_op_list);
process_rsc_state(rsc, node, on_fail, migrate_op, data_set);
if (get_target_role(rsc, &req_role)) {
if (rsc->next_role == RSC_ROLE_UNKNOWN || req_role < rsc->next_role) {
pe_rsc_debug(rsc, "%s: Overwriting calculated next role %s"
" with requested next role %s",
rsc->id, role2text(rsc->next_role), role2text(req_role));
rsc->next_role = req_role;
} else if (req_role > rsc->next_role) {
pe_rsc_info(rsc, "%s: Not overwriting calculated next role %s"
" with requested next role %s",
rsc->id, role2text(rsc->next_role), role2text(req_role));
}
}
if (saved_role > rsc->role) {
rsc->role = saved_role;
}
return rsc;
}
static void
handle_orphaned_container_fillers(xmlNode * lrm_rsc_list, pe_working_set_t * data_set)
{
xmlNode *rsc_entry = NULL;
for (rsc_entry = __xml_first_child_element(lrm_rsc_list); rsc_entry != NULL;
rsc_entry = __xml_next_element(rsc_entry)) {
pe_resource_t *rsc;
pe_resource_t *container;
const char *rsc_id;
const char *container_id;
if (safe_str_neq((const char *)rsc_entry->name, XML_LRM_TAG_RESOURCE)) {
continue;
}
container_id = crm_element_value(rsc_entry, XML_RSC_ATTR_CONTAINER);
rsc_id = crm_element_value(rsc_entry, XML_ATTR_ID);
if (container_id == NULL || rsc_id == NULL) {
continue;
}
container = pe_find_resource(data_set->resources, container_id);
if (container == NULL) {
continue;
}
rsc = pe_find_resource(data_set->resources, rsc_id);
if (rsc == NULL ||
is_set(rsc->flags, pe_rsc_orphan_container_filler) == FALSE ||
rsc->container != NULL) {
continue;
}
pe_rsc_trace(rsc, "Mapped container of orphaned resource %s to %s",
rsc->id, container_id);
rsc->container = container;
container->fillers = g_list_append(container->fillers, rsc);
}
}
gboolean
unpack_lrm_resources(pe_node_t * node, xmlNode * lrm_rsc_list, pe_working_set_t * data_set)
{
xmlNode *rsc_entry = NULL;
gboolean found_orphaned_container_filler = FALSE;
CRM_CHECK(node != NULL, return FALSE);
crm_trace("Unpacking resources on %s", node->details->uname);
for (rsc_entry = __xml_first_child_element(lrm_rsc_list); rsc_entry != NULL;
rsc_entry = __xml_next_element(rsc_entry)) {
if (crm_str_eq((const char *)rsc_entry->name, XML_LRM_TAG_RESOURCE, TRUE)) {
pe_resource_t *rsc = unpack_lrm_rsc_state(node, rsc_entry, data_set);
if (!rsc) {
continue;
}
if (is_set(rsc->flags, pe_rsc_orphan_container_filler)) {
found_orphaned_container_filler = TRUE;
}
}
}
/* now that all the resource state has been unpacked for this node
* we have to go back and map any orphaned container fillers to their
* container resource */
if (found_orphaned_container_filler) {
handle_orphaned_container_fillers(lrm_rsc_list, data_set);
}
return TRUE;
}
static void
set_active(pe_resource_t * rsc)
{
pe_resource_t *top = uber_parent(rsc);
if (top && is_set(top->flags, pe_rsc_promotable)) {
rsc->role = RSC_ROLE_SLAVE;
} else {
rsc->role = RSC_ROLE_STARTED;
}
}
static void
set_node_score(gpointer key, gpointer value, gpointer user_data)
{
pe_node_t *node = value;
int *score = user_data;
node->weight = *score;
}
#define STATUS_PATH_MAX 1024
static xmlNode *
find_lrm_op(const char *resource, const char *op, const char *node, const char *source,
bool success_only, pe_working_set_t *data_set)
{
int offset = 0;
char xpath[STATUS_PATH_MAX];
xmlNode *xml = NULL;
offset += snprintf(xpath + offset, STATUS_PATH_MAX - offset, "//node_state[@uname='%s']", node);
offset +=
snprintf(xpath + offset, STATUS_PATH_MAX - offset, "//" XML_LRM_TAG_RESOURCE "[@id='%s']",
resource);
/* Need to check against transition_magic too? */
if (source && safe_str_eq(op, CRMD_ACTION_MIGRATE)) {
offset +=
snprintf(xpath + offset, STATUS_PATH_MAX - offset,
"/" XML_LRM_TAG_RSC_OP "[@operation='%s' and @migrate_target='%s']", op,
source);
} else if (source && safe_str_eq(op, CRMD_ACTION_MIGRATED)) {
offset +=
snprintf(xpath + offset, STATUS_PATH_MAX - offset,
"/" XML_LRM_TAG_RSC_OP "[@operation='%s' and @migrate_source='%s']", op,
source);
} else {
offset +=
snprintf(xpath + offset, STATUS_PATH_MAX - offset,
"/" XML_LRM_TAG_RSC_OP "[@operation='%s']", op);
}
CRM_LOG_ASSERT(offset > 0);
xml = get_xpath_object(xpath, data_set->input, LOG_DEBUG);
if (xml && success_only) {
int rc = PCMK_OCF_UNKNOWN_ERROR;
int status = PCMK_LRM_OP_ERROR;
crm_element_value_int(xml, XML_LRM_ATTR_RC, &rc);
crm_element_value_int(xml, XML_LRM_ATTR_OPSTATUS, &status);
if ((rc != PCMK_OCF_OK) || (status != PCMK_LRM_OP_DONE)) {
return NULL;
}
}
return xml;
}
static int
pe__call_id(xmlNode *op_xml)
{
int id = 0;
if (op_xml) {
crm_element_value_int(op_xml, XML_LRM_ATTR_CALLID, &id);
}
return id;
}
/*!
* \brief Check whether a stop happened on the same node after some event
*
* \param[in] rsc Resource being checked
* \param[in] node Node being checked
* \param[in] xml_op Event that stop is being compared to
* \param[in] data_set Cluster working set
*
* \return TRUE if stop happened after event, FALSE otherwise
*
* \note This is really unnecessary, but kept as a safety mechanism. We
* currently don't save more than one successful event in history, so this
* only matters when processing really old CIB files that we don't
* technically support anymore, or as preparation for logging an extended
* history in the future.
*/
static bool
stop_happened_after(pe_resource_t *rsc, pe_node_t *node, xmlNode *xml_op,
pe_working_set_t *data_set)
{
xmlNode *stop_op = find_lrm_op(rsc->id, CRMD_ACTION_STOP,
node->details->uname, NULL, TRUE, data_set);
return (stop_op && (pe__call_id(stop_op) > pe__call_id(xml_op)));
}
static void
unpack_migrate_to_success(pe_resource_t *rsc, pe_node_t *node, xmlNode *xml_op,
pe_working_set_t *data_set)
{
/* A successful migration sequence is:
* migrate_to on source node
* migrate_from on target node
* stop on source node
*
* If a migrate_to is followed by a stop, the entire migration (successful
* or failed) is complete, and we don't care what happened on the target.
*
* If no migrate_from has happened, the migration is considered to be
* "partial". If the migrate_from failed, make sure the resource gets
* stopped on both source and target (if up).
*
* If the migrate_to and migrate_from both succeeded (which also implies the
* resource is no longer running on the source), but there is no stop, the
* migration is considered to be "dangling". Schedule a stop on the source
* in this case.
*/
int from_rc = 0;
int from_status = 0;
pe_node_t *target_node = NULL;
pe_node_t *source_node = NULL;
xmlNode *migrate_from = NULL;
const char *source = crm_element_value(xml_op, XML_LRM_ATTR_MIGRATE_SOURCE);
const char *target = crm_element_value(xml_op, XML_LRM_ATTR_MIGRATE_TARGET);
// Sanity check
CRM_CHECK(source && target && !strcmp(source, node->details->uname), return);
if (stop_happened_after(rsc, node, xml_op, data_set)) {
return;
}
// Clones are not allowed to migrate, so role can't be master
rsc->role = RSC_ROLE_STARTED;
target_node = pe_find_node(data_set->nodes, target);
source_node = pe_find_node(data_set->nodes, source);
// Check whether there was a migrate_from action on the target
migrate_from = find_lrm_op(rsc->id, CRMD_ACTION_MIGRATED, target,
source, FALSE, data_set);
if (migrate_from) {
crm_element_value_int(migrate_from, XML_LRM_ATTR_RC, &from_rc);
crm_element_value_int(migrate_from, XML_LRM_ATTR_OPSTATUS, &from_status);
pe_rsc_trace(rsc, "%s op on %s exited with status=%d, rc=%d",
ID(migrate_from), target, from_status, from_rc);
}
if (migrate_from && from_rc == PCMK_OCF_OK
&& from_status == PCMK_LRM_OP_DONE) {
/* The migrate_to and migrate_from both succeeded, so mark the migration
* as "dangling". This will be used to schedule a stop action on the
* source without affecting the target.
*/
pe_rsc_trace(rsc, "Detected dangling migration op: %s on %s", ID(xml_op),
source);
rsc->role = RSC_ROLE_STOPPED;
rsc->dangling_migrations = g_list_prepend(rsc->dangling_migrations, node);
} else if (migrate_from && (from_status != PCMK_LRM_OP_PENDING)) { // Failed
if (target_node && target_node->details->online) {
pe_rsc_trace(rsc, "Marking active on %s %p %d", target, target_node,
target_node->details->online);
native_add_running(rsc, target_node, data_set);
}
} else { // Pending, or complete but erased
if (target_node && target_node->details->online) {
pe_rsc_trace(rsc, "Marking active on %s %p %d", target, target_node,
target_node->details->online);
native_add_running(rsc, target_node, data_set);
if (source_node && source_node->details->online) {
/* This is a partial migration: the migrate_to completed
* successfully on the source, but the migrate_from has not
* completed. Remember the source and target; if the newly
* chosen target remains the same when we schedule actions
* later, we may continue with the migration.
*/
rsc->partial_migration_target = target_node;
rsc->partial_migration_source = source_node;
}
} else {
/* Consider it failed here - forces a restart, prevents migration */
set_bit(rsc->flags, pe_rsc_failed);
clear_bit(rsc->flags, pe_rsc_allow_migrate);
}
}
}
static void
unpack_migrate_to_failure(pe_resource_t *rsc, pe_node_t *node, xmlNode *xml_op,
pe_working_set_t *data_set)
{
int target_stop_id = 0;
int target_migrate_from_id = 0;
xmlNode *target_stop = NULL;
xmlNode *target_migrate_from = NULL;
const char *source = crm_element_value(xml_op, XML_LRM_ATTR_MIGRATE_SOURCE);
const char *target = crm_element_value(xml_op, XML_LRM_ATTR_MIGRATE_TARGET);
// Sanity check
CRM_CHECK(source && target && !strcmp(source, node->details->uname), return);
/* If a migration failed, we have to assume the resource is active. Clones
* are not allowed to migrate, so role can't be master.
*/
rsc->role = RSC_ROLE_STARTED;
// Check for stop on the target
target_stop = find_lrm_op(rsc->id, CRMD_ACTION_STOP, target, NULL,
TRUE, data_set);
target_stop_id = pe__call_id(target_stop);
// Check for migrate_from on the target
target_migrate_from = find_lrm_op(rsc->id, CRMD_ACTION_MIGRATED, target,
source, TRUE, data_set);
target_migrate_from_id = pe__call_id(target_migrate_from);
if ((target_stop == NULL) || (target_stop_id < target_migrate_from_id)) {
/* There was no stop on the source, or a stop that happened before a
* migrate_from, so assume the resource is still active on the target
* (if it is up).
*/
pe_node_t *target_node = pe_find_node(data_set->nodes, target);
pe_rsc_trace(rsc, "stop (%d) + migrate_from (%d)",
target_stop_id, target_migrate_from_id);
if (target_node && target_node->details->online) {
native_add_running(rsc, target_node, data_set);
}
} else if (target_migrate_from == NULL) {
/* We know there was a stop on the target, but there may not have been a
* migrate_from (the stop could have happened before migrate_from was
* scheduled or attempted).
*
* That means this could be a "dangling" migration. But first, check
* whether there is a newer migrate_from or start on the source node --
* it's possible the failed migration was followed by a successful
* full restart or migration in the reverse direction, in which case we
* don't want to force it to stop.
*/
xmlNode *source_migrate_from = NULL;
xmlNode *source_start = NULL;
int source_migrate_to_id = pe__call_id(xml_op);
source_migrate_from = find_lrm_op(rsc->id, CRMD_ACTION_MIGRATED, source,
NULL, TRUE, data_set);
if (pe__call_id(source_migrate_from) > source_migrate_to_id) {
return;
}
source_start = find_lrm_op(rsc->id, CRMD_ACTION_START, source, NULL,
TRUE, data_set);
if (pe__call_id(source_start) > source_migrate_to_id) {
return;
}
// Mark node as having dangling migration so we can force a stop later
rsc->dangling_migrations = g_list_prepend(rsc->dangling_migrations, node);
}
}
static void
unpack_migrate_from_failure(pe_resource_t *rsc, pe_node_t *node,
xmlNode *xml_op, pe_working_set_t *data_set)
{
xmlNode *source_stop = NULL;
xmlNode *source_migrate_to = NULL;
const char *source = crm_element_value(xml_op, XML_LRM_ATTR_MIGRATE_SOURCE);
const char *target = crm_element_value(xml_op, XML_LRM_ATTR_MIGRATE_TARGET);
// Sanity check
CRM_CHECK(source && target && !strcmp(target, node->details->uname), return);
/* If a migration failed, we have to assume the resource is active. Clones
* are not allowed to migrate, so role can't be master.
*/
rsc->role = RSC_ROLE_STARTED;
// Check for a stop on the source
source_stop = find_lrm_op(rsc->id, CRMD_ACTION_STOP, source, NULL,
TRUE, data_set);
// Check for a migrate_to on the source
source_migrate_to = find_lrm_op(rsc->id, CRMD_ACTION_MIGRATE,
source, target, TRUE, data_set);
if ((source_stop == NULL)
|| (pe__call_id(source_stop) < pe__call_id(source_migrate_to))) {
/* There was no stop on the source, or a stop that happened before
* migrate_to, so assume the resource is still active on the source (if
* it is up).
*/
pe_node_t *source_node = pe_find_node(data_set->nodes, source);
if (source_node && source_node->details->online) {
native_add_running(rsc, source_node, data_set);
}
}
}
static void
record_failed_op(xmlNode *op, const pe_node_t *node,
const pe_resource_t *rsc, pe_working_set_t *data_set)
{
xmlNode *xIter = NULL;
const char *op_key = crm_element_value(op, XML_LRM_ATTR_TASK_KEY);
if (node->details->online == FALSE) {
return;
}
for (xIter = data_set->failed->children; xIter; xIter = xIter->next) {
const char *key = crm_element_value(xIter, XML_LRM_ATTR_TASK_KEY);
const char *uname = crm_element_value(xIter, XML_ATTR_UNAME);
if(safe_str_eq(op_key, key) && safe_str_eq(uname, node->details->uname)) {
crm_trace("Skipping duplicate entry %s on %s", op_key, node->details->uname);
return;
}
}
crm_trace("Adding entry %s on %s", op_key, node->details->uname);
crm_xml_add(op, XML_ATTR_UNAME, node->details->uname);
crm_xml_add(op, XML_LRM_ATTR_RSCID, rsc->id);
add_node_copy(data_set->failed, op);
}
static const char *get_op_key(xmlNode *xml_op)
{
const char *key = crm_element_value(xml_op, XML_LRM_ATTR_TASK_KEY);
if(key == NULL) {
key = ID(xml_op);
}
return key;
}
static const char *
last_change_str(xmlNode *xml_op)
{
time_t when;
const char *when_s = NULL;
if (crm_element_value_epoch(xml_op, XML_RSC_OP_LAST_CHANGE,
&when) == pcmk_ok) {
when_s = pcmk__epoch2str(&when);
if (when_s) {
// Skip day of week to make message shorter
when_s = strchr(when_s, ' ');
if (when_s) {
++when_s;
}
}
}
return ((when_s && *when_s)? when_s : "unknown time");
}
static void
unpack_rsc_op_failure(pe_resource_t * rsc, pe_node_t * node, int rc, xmlNode * xml_op, xmlNode ** last_failure,
enum action_fail_response * on_fail, pe_working_set_t * data_set)
{
guint interval_ms = 0;
bool is_probe = false;
pe_action_t *action = NULL;
const char *key = get_op_key(xml_op);
const char *task = crm_element_value(xml_op, XML_LRM_ATTR_TASK);
const char *exit_reason = crm_element_value(xml_op,
XML_LRM_ATTR_EXIT_REASON);
CRM_ASSERT(rsc);
CRM_CHECK(task != NULL, return);
*last_failure = xml_op;
crm_element_value_ms(xml_op, XML_LRM_ATTR_INTERVAL_MS, &interval_ms);
if ((interval_ms == 0) && !strcmp(task, CRMD_ACTION_STATUS)) {
is_probe = true;
}
if (exit_reason == NULL) {
exit_reason = "";
}
if (is_not_set(data_set->flags, pe_flag_symmetric_cluster)
&& (rc == PCMK_OCF_NOT_INSTALLED)) {
crm_trace("Unexpected result (%s%s%s) was recorded for "
"%s of %s on %s at %s " CRM_XS " rc=%d id=%s",
services_ocf_exitcode_str(rc),
(*exit_reason? ": " : ""), exit_reason,
(is_probe? "probe" : task), rsc->id, node->details->uname,
last_change_str(xml_op), rc, ID(xml_op));
} else {
crm_warn("Unexpected result (%s%s%s) was recorded for "
"%s of %s on %s at %s " CRM_XS " rc=%d id=%s",
services_ocf_exitcode_str(rc),
(*exit_reason? ": " : ""), exit_reason,
(is_probe? "probe" : task), rsc->id, node->details->uname,
last_change_str(xml_op), rc, ID(xml_op));
if (is_probe && (rc != PCMK_OCF_OK)
&& (rc != PCMK_OCF_NOT_RUNNING)
&& (rc != PCMK_OCF_RUNNING_MASTER)) {
/* A failed (not just unexpected) probe result could mean the user
* didn't know resources will be probed even where they can't run.
*/
crm_notice("If it is not possible for %s to run on %s, see "
"the resource-discovery option for location constraints",
rsc->id, node->details->uname);
}
record_failed_op(xml_op, node, rsc, data_set);
}
action = custom_action(rsc, strdup(key), task, NULL, TRUE, FALSE, data_set);
if ((action->on_fail <= action_fail_fence && *on_fail < action->on_fail) ||
(action->on_fail == action_fail_reset_remote && *on_fail <= action_fail_recover) ||
(action->on_fail == action_fail_restart_container && *on_fail <= action_fail_recover) ||
(*on_fail == action_fail_restart_container && action->on_fail >= action_fail_migrate)) {
pe_rsc_trace(rsc, "on-fail %s -> %s for %s (%s)", fail2text(*on_fail),
fail2text(action->on_fail), action->uuid, key);
*on_fail = action->on_fail;
}
if (!strcmp(task, CRMD_ACTION_STOP)) {
resource_location(rsc, node, -INFINITY, "__stop_fail__", data_set);
} else if (!strcmp(task, CRMD_ACTION_MIGRATE)) {
unpack_migrate_to_failure(rsc, node, xml_op, data_set);
} else if (!strcmp(task, CRMD_ACTION_MIGRATED)) {
unpack_migrate_from_failure(rsc, node, xml_op, data_set);
} else if (!strcmp(task, CRMD_ACTION_PROMOTE)) {
rsc->role = RSC_ROLE_MASTER;
} else if (!strcmp(task, CRMD_ACTION_DEMOTE)) {
if (action->on_fail == action_fail_block) {
rsc->role = RSC_ROLE_MASTER;
rsc->next_role = RSC_ROLE_STOPPED;
} else if(rc == PCMK_OCF_NOT_RUNNING) {
rsc->role = RSC_ROLE_STOPPED;
} else {
/* Staying in master role would put the scheduler and controller
* into a loop. Setting slave role is not dangerous because the
* resource will be stopped as part of recovery, and any master
* promotion will be ordered after that stop.
*/
rsc->role = RSC_ROLE_SLAVE;
}
}
if(is_probe && rc == PCMK_OCF_NOT_INSTALLED) {
/* leave stopped */
pe_rsc_trace(rsc, "Leaving %s stopped", rsc->id);
rsc->role = RSC_ROLE_STOPPED;
} else if (rsc->role < RSC_ROLE_STARTED) {
pe_rsc_trace(rsc, "Setting %s active", rsc->id);
set_active(rsc);
}
pe_rsc_trace(rsc, "Resource %s: role=%s, unclean=%s, on_fail=%s, fail_role=%s",
rsc->id, role2text(rsc->role),
node->details->unclean ? "true" : "false",
fail2text(action->on_fail), role2text(action->fail_role));
if (action->fail_role != RSC_ROLE_STARTED && rsc->next_role < action->fail_role) {
rsc->next_role = action->fail_role;
}
if (action->fail_role == RSC_ROLE_STOPPED) {
int score = -INFINITY;
pe_resource_t *fail_rsc = rsc;
if (fail_rsc->parent) {
pe_resource_t *parent = uber_parent(fail_rsc);
if (pe_rsc_is_clone(parent)
&& is_not_set(parent->flags, pe_rsc_unique)) {
/* For clone resources, if a child fails on an operation
* with on-fail = stop, all the resources fail. Do this by preventing
* the parent from coming up again. */
fail_rsc = parent;
}
}
crm_notice("%s will not be started under current conditions",
fail_rsc->id);
/* make sure it doesn't come up again */
if (fail_rsc->allowed_nodes != NULL) {
g_hash_table_destroy(fail_rsc->allowed_nodes);
}
fail_rsc->allowed_nodes = node_hash_from_list(data_set->nodes);
g_hash_table_foreach(fail_rsc->allowed_nodes, set_node_score, &score);
}
pe_free_action(action);
}
/*!
* \internal
* \brief Remap operation status based on action result
*
* Given an action result, determine an appropriate operation status for the
* purposes of responding to the action (the status provided by the executor is
* not directly usable since the executor does not know what was expected).
*
* \param[in,out] rsc Resource that operation history entry is for
* \param[in] rc Actual return code of operation
* \param[in] target_rc Expected return code of operation
* \param[in] node Node where operation was executed
* \param[in] xml_op Operation history entry XML from CIB status
* \param[in,out] on_fail What should be done about the result
* \param[in] data_set Current cluster working set
*
* \return Operation status based on return code and action info
* \note This may update the resource's current and next role.
*/
static int
determine_op_status(
pe_resource_t *rsc, int rc, int target_rc, pe_node_t * node, xmlNode * xml_op, enum action_fail_response * on_fail, pe_working_set_t * data_set)
{
guint interval_ms = 0;
bool is_probe = false;
int result = PCMK_LRM_OP_DONE;
const char *key = get_op_key(xml_op);
const char *task = crm_element_value(xml_op, XML_LRM_ATTR_TASK);
const char *exit_reason = crm_element_value(xml_op,
XML_LRM_ATTR_EXIT_REASON);
CRM_ASSERT(rsc);
CRM_CHECK(task != NULL, return PCMK_LRM_OP_ERROR);
if (exit_reason == NULL) {
exit_reason = "";
}
crm_element_value_ms(xml_op, XML_LRM_ATTR_INTERVAL_MS, &interval_ms);
if ((interval_ms == 0) && !strcmp(task, CRMD_ACTION_STATUS)) {
is_probe = true;
task = "probe";
}
if (target_rc < 0) {
/* Pre-1.0 Pacemaker versions, and Pacemaker 1.1.6 or earlier with
* Heartbeat 2.0.7 or earlier as the cluster layer, did not include the
* target_rc in the transition key, which (along with the similar case
* of a corrupted transition key in the CIB) will be reported to this
* function as -1. Pacemaker 2.0+ does not support rolling upgrades from
* those versions or processing of saved CIB files from those versions,
* so we do not need to care much about this case.
*/
result = PCMK_LRM_OP_ERROR;
crm_warn("Expected result not found for %s on %s (corrupt or obsolete CIB?)",
key, node->details->uname);
} else if (target_rc != rc) {
result = PCMK_LRM_OP_ERROR;
pe_rsc_debug(rsc, "%s on %s: expected %d (%s), got %d (%s%s%s)",
key, node->details->uname,
target_rc, services_ocf_exitcode_str(target_rc),
rc, services_ocf_exitcode_str(rc),
(*exit_reason? ": " : ""), exit_reason);
}
switch (rc) {
case PCMK_OCF_OK:
if (is_probe && (target_rc == PCMK_OCF_NOT_RUNNING)) {
result = PCMK_LRM_OP_DONE;
pe_rsc_info(rsc, "Probe found %s active on %s at %s",
rsc->id, node->details->uname,
last_change_str(xml_op));
}
break;
case PCMK_OCF_NOT_RUNNING:
if (is_probe || target_rc == rc || is_not_set(rsc->flags, pe_rsc_managed)) {
result = PCMK_LRM_OP_DONE;
rsc->role = RSC_ROLE_STOPPED;
/* clear any previous failure actions */
*on_fail = action_fail_ignore;
rsc->next_role = RSC_ROLE_UNKNOWN;
}
break;
case PCMK_OCF_RUNNING_MASTER:
if (is_probe && (rc != target_rc)) {
result = PCMK_LRM_OP_DONE;
pe_rsc_info(rsc,
"Probe found %s active and promoted on %s at %s",
rsc->id, node->details->uname,
last_change_str(xml_op));
}
rsc->role = RSC_ROLE_MASTER;
break;
case PCMK_OCF_DEGRADED_MASTER:
case PCMK_OCF_FAILED_MASTER:
rsc->role = RSC_ROLE_MASTER;
result = PCMK_LRM_OP_ERROR;
break;
case PCMK_OCF_NOT_CONFIGURED:
result = PCMK_LRM_OP_ERROR_FATAL;
break;
case PCMK_OCF_UNIMPLEMENT_FEATURE:
if (interval_ms > 0) {
result = PCMK_LRM_OP_NOTSUPPORTED;
break;
}
// fall through
case PCMK_OCF_NOT_INSTALLED:
case PCMK_OCF_INVALID_PARAM:
case PCMK_OCF_INSUFFICIENT_PRIV:
if (!pe_can_fence(data_set, node)
&& !strcmp(task, CRMD_ACTION_STOP)) {
/* If a stop fails and we can't fence, there's nothing else we can do */
pe_proc_err("No further recovery can be attempted for %s "
"because %s on %s failed (%s%s%s) at %s "
CRM_XS " rc=%d id=%s", rsc->id, task,
node->details->uname, services_ocf_exitcode_str(rc),
(*exit_reason? ": " : ""), exit_reason,
last_change_str(xml_op), rc, ID(xml_op));
clear_bit(rsc->flags, pe_rsc_managed);
set_bit(rsc->flags, pe_rsc_block);
}
result = PCMK_LRM_OP_ERROR_HARD;
break;
default:
if (result == PCMK_LRM_OP_DONE) {
crm_info("Treating unknown exit status %d from %s of %s "
"on %s at %s as failure",
rc, task, rsc->id, node->details->uname,
last_change_str(xml_op));
result = PCMK_LRM_OP_ERROR;
}
break;
}
return result;
}
// return TRUE if start or monitor last failure but parameters changed
static bool
should_clear_for_param_change(xmlNode *xml_op, const char *task,
pe_resource_t *rsc, pe_node_t *node,
pe_working_set_t *data_set)
{
if (!strcmp(task, "start") || !strcmp(task, "monitor")) {
if (pe__bundle_needs_remote_name(rsc)) {
/* We haven't allocated resources yet, so we can't reliably
* substitute addr parameters for the REMOTE_CONTAINER_HACK.
* When that's needed, defer the check until later.
*/
pe__add_param_check(xml_op, rsc, node, pe_check_last_failure,
data_set);
} else {
op_digest_cache_t *digest_data = NULL;
digest_data = rsc_action_digest_cmp(rsc, xml_op, node, data_set);
switch (digest_data->rc) {
case RSC_DIGEST_UNKNOWN:
crm_trace("Resource %s history entry %s on %s"
" has no digest to compare",
rsc->id, get_op_key(xml_op), node->details->id);
break;
case RSC_DIGEST_MATCH:
break;
default:
return TRUE;
}
}
}
return FALSE;
}
// Order action after fencing of remote node, given connection rsc
static void
order_after_remote_fencing(pe_action_t *action, pe_resource_t *remote_conn,
pe_working_set_t *data_set)
{
pe_node_t *remote_node = pe_find_node(data_set->nodes, remote_conn->id);
if (remote_node) {
pe_action_t *fence = pe_fence_op(remote_node, NULL, TRUE, NULL,
- data_set);
+ FALSE, data_set);
order_actions(fence, action, pe_order_implies_then);
}
}
static bool
should_ignore_failure_timeout(pe_resource_t *rsc, xmlNode *xml_op,
const char *task, guint interval_ms,
bool is_last_failure, pe_working_set_t *data_set)
{
/* Clearing failures of recurring monitors has special concerns. The
* executor reports only changes in the monitor result, so if the
* monitor is still active and still getting the same failure result,
* that will go undetected after the failure is cleared.
*
* Also, the operation history will have the time when the recurring
* monitor result changed to the given code, not the time when the
* result last happened.
*
* @TODO We probably should clear such failures only when the failure
* timeout has passed since the last occurrence of the failed result.
* However we don't record that information. We could maybe approximate
* that by clearing only if there is a more recent successful monitor or
* stop result, but we don't even have that information at this point
* since we are still unpacking the resource's operation history.
*
* This is especially important for remote connection resources with a
* reconnect interval, so in that case, we skip clearing failures
* if the remote node hasn't been fenced.
*/
if (rsc->remote_reconnect_ms
&& is_set(data_set->flags, pe_flag_stonith_enabled)
&& (interval_ms != 0) && safe_str_eq(task, CRMD_ACTION_STATUS)) {
pe_node_t *remote_node = pe_find_node(data_set->nodes, rsc->id);
if (remote_node && !remote_node->details->remote_was_fenced) {
if (is_last_failure) {
crm_info("Waiting to clear monitor failure for remote node %s"
" until fencing has occurred", rsc->id);
}
return TRUE;
}
}
return FALSE;
}
/*!
* \internal
* \brief Check operation age and schedule failure clearing when appropriate
*
* This function has two distinct purposes. The first is to check whether an
* operation history entry is expired (i.e. the resource has a failure timeout,
* the entry is older than the timeout, and the resource either has no fail
* count or its fail count is entirely older than the timeout). The second is to
* schedule fail count clearing when appropriate (i.e. the operation is expired
* and either the resource has an expired fail count or the operation is a
* last_failure for a remote connection resource with a reconnect interval,
* or the operation is a last_failure for a start or monitor operation and the
* resource's parameters have changed since the operation).
*
* \param[in] rsc Resource that operation happened to
* \param[in] node Node that operation happened on
* \param[in] rc Actual result of operation
* \param[in] xml_op Operation history entry XML
* \param[in] data_set Current working set
*
* \return TRUE if operation history entry is expired, FALSE otherwise
*/
static bool
check_operation_expiry(pe_resource_t *rsc, pe_node_t *node, int rc,
xmlNode *xml_op, pe_working_set_t *data_set)
{
bool expired = FALSE;
bool is_last_failure = pcmk__ends_with(ID(xml_op), "_last_failure_0");
time_t last_run = 0;
guint interval_ms = 0;
int unexpired_fail_count = 0;
const char *task = crm_element_value(xml_op, XML_LRM_ATTR_TASK);
const char *clear_reason = NULL;
crm_element_value_ms(xml_op, XML_LRM_ATTR_INTERVAL_MS, &interval_ms);
if ((rsc->failure_timeout > 0)
&& (crm_element_value_epoch(xml_op, XML_RSC_OP_LAST_CHANGE,
&last_run) == 0)) {
// Resource has a failure-timeout, and history entry has a timestamp
time_t now = get_effective_time(data_set);
time_t last_failure = 0;
// Is this particular operation history older than the failure timeout?
if ((now >= (last_run + rsc->failure_timeout))
&& !should_ignore_failure_timeout(rsc, xml_op, task, interval_ms,
is_last_failure, data_set)) {
expired = TRUE;
}
// Does the resource as a whole have an unexpired fail count?
unexpired_fail_count = pe_get_failcount(node, rsc, &last_failure,
pe_fc_effective, xml_op,
data_set);
// Update scheduler recheck time according to *last* failure
crm_trace("%s@%lld is %sexpired @%lld with unexpired_failures=%d timeout=%ds"
" last-failure@%lld",
ID(xml_op), (long long) last_run, (expired? "" : "not "),
(long long) now, unexpired_fail_count, rsc->failure_timeout,
(long long) last_failure);
last_failure += rsc->failure_timeout + 1;
if (unexpired_fail_count && (now < last_failure)) {
pe__update_recheck_time(last_failure, data_set);
}
}
if (expired) {
if (pe_get_failcount(node, rsc, NULL, pe_fc_default, xml_op, data_set)) {
// There is a fail count ignoring timeout
if (unexpired_fail_count == 0) {
// There is no fail count considering timeout
clear_reason = "it expired";
} else {
/* This operation is old, but there is an unexpired fail count.
* In a properly functioning cluster, this should only be
* possible if this operation is not a failure (otherwise the
* fail count should be expired too), so this is really just a
* failsafe.
*/
expired = FALSE;
}
} else if (is_last_failure && rsc->remote_reconnect_ms) {
/* Clear any expired last failure when reconnect interval is set,
* even if there is no fail count.
*/
clear_reason = "reconnect interval is set";
}
}
if (!expired && is_last_failure
&& should_clear_for_param_change(xml_op, task, rsc, node, data_set)) {
clear_reason = "resource parameters have changed";
}
if (clear_reason != NULL) {
// Schedule clearing of the fail count
pe_action_t *clear_op = pe__clear_failcount(rsc, node, clear_reason,
data_set);
if (is_set(data_set->flags, pe_flag_stonith_enabled)
&& rsc->remote_reconnect_ms) {
/* If we're clearing a remote connection due to a reconnect
* interval, we want to wait until any scheduled fencing
* completes.
*
* We could limit this to remote_node->details->unclean, but at
* this point, that's always true (it won't be reliable until
* after unpack_node_loop() is done).
*/
crm_info("Clearing %s failure will wait until any scheduled "
"fencing of %s completes", task, rsc->id);
order_after_remote_fencing(clear_op, rsc, data_set);
}
}
if (expired && (interval_ms == 0) && safe_str_eq(task, CRMD_ACTION_STATUS)) {
switch(rc) {
case PCMK_OCF_OK:
case PCMK_OCF_NOT_RUNNING:
case PCMK_OCF_RUNNING_MASTER:
case PCMK_OCF_DEGRADED:
case PCMK_OCF_DEGRADED_MASTER:
// Don't expire probes that return these values
expired = FALSE;
break;
}
}
return expired;
}
int pe__target_rc_from_xml(xmlNode *xml_op)
{
int target_rc = 0;
const char *key = crm_element_value(xml_op, XML_ATTR_TRANSITION_KEY);
if (key == NULL) {
return -1;
}
decode_transition_key(key, NULL, NULL, NULL, &target_rc);
return target_rc;
}
static enum action_fail_response
get_action_on_fail(pe_resource_t *rsc, const char *key, const char *task, pe_working_set_t * data_set)
{
int result = action_fail_recover;
pe_action_t *action = custom_action(rsc, strdup(key), task, NULL, TRUE, FALSE, data_set);
result = action->on_fail;
pe_free_action(action);
return result;
}
static void
update_resource_state(pe_resource_t * rsc, pe_node_t * node, xmlNode * xml_op, const char * task, int rc,
xmlNode * last_failure, enum action_fail_response * on_fail, pe_working_set_t * data_set)
{
gboolean clear_past_failure = FALSE;
CRM_ASSERT(rsc);
CRM_ASSERT(xml_op);
if (rc == PCMK_OCF_NOT_RUNNING) {
clear_past_failure = TRUE;
} else if (rc == PCMK_OCF_NOT_INSTALLED) {
rsc->role = RSC_ROLE_STOPPED;
} else if (safe_str_eq(task, CRMD_ACTION_STATUS)) {
if (last_failure) {
const char *op_key = get_op_key(xml_op);
const char *last_failure_key = get_op_key(last_failure);
if (safe_str_eq(op_key, last_failure_key)) {
clear_past_failure = TRUE;
}
}
if (rsc->role < RSC_ROLE_STARTED) {
set_active(rsc);
}
} else if (safe_str_eq(task, CRMD_ACTION_START)) {
rsc->role = RSC_ROLE_STARTED;
clear_past_failure = TRUE;
} else if (safe_str_eq(task, CRMD_ACTION_STOP)) {
rsc->role = RSC_ROLE_STOPPED;
clear_past_failure = TRUE;
} else if (safe_str_eq(task, CRMD_ACTION_PROMOTE)) {
rsc->role = RSC_ROLE_MASTER;
clear_past_failure = TRUE;
} else if (safe_str_eq(task, CRMD_ACTION_DEMOTE)) {
/* Demote from Master does not clear an error */
rsc->role = RSC_ROLE_SLAVE;
} else if (safe_str_eq(task, CRMD_ACTION_MIGRATED)) {
rsc->role = RSC_ROLE_STARTED;
clear_past_failure = TRUE;
} else if (safe_str_eq(task, CRMD_ACTION_MIGRATE)) {
unpack_migrate_to_success(rsc, node, xml_op, data_set);
} else if (rsc->role < RSC_ROLE_STARTED) {
pe_rsc_trace(rsc, "%s active on %s", rsc->id, node->details->uname);
set_active(rsc);
}
/* clear any previous failure actions */
if (clear_past_failure) {
switch (*on_fail) {
case action_fail_stop:
case action_fail_fence:
case action_fail_migrate:
case action_fail_standby:
pe_rsc_trace(rsc, "%s.%s is not cleared by a completed stop",
rsc->id, fail2text(*on_fail));
break;
case action_fail_block:
case action_fail_ignore:
case action_fail_recover:
case action_fail_restart_container:
*on_fail = action_fail_ignore;
rsc->next_role = RSC_ROLE_UNKNOWN;
break;
case action_fail_reset_remote:
if (rsc->remote_reconnect_ms == 0) {
/* With no reconnect interval, the connection is allowed to
* start again after the remote node is fenced and
* completely stopped. (With a reconnect interval, we wait
* for the failure to be cleared entirely before attempting
* to reconnect.)
*/
*on_fail = action_fail_ignore;
rsc->next_role = RSC_ROLE_UNKNOWN;
}
break;
}
}
}
/*!
* \internal
* \brief Remap informational monitor results to usual values
*
* Certain OCF result codes are for providing extended information to the
* user about services that aren't yet failed but not entirely healthy either.
* These must be treated as the "normal" result by pacemaker.
*
* \param[in] rc Actual result of a monitor action
* \param[in] xml_op Operation history XML
* \param[in] node Node that operation happened on
* \param[in] rsc Resource that operation happened to
* \param[in] data_set Cluster working set
*
* \return Result code that pacemaker should use
*
* \note If the result is remapped, and the node is not shutting down or failed,
* the operation will be recorded in the data set's list of failed
* operations, to highlight it for the user.
*/
static int
remap_monitor_rc(int rc, xmlNode *xml_op, const pe_node_t *node,
const pe_resource_t *rsc, pe_working_set_t *data_set)
{
int remapped_rc = rc;
switch (rc) {
case PCMK_OCF_DEGRADED:
remapped_rc = PCMK_OCF_OK;
break;
case PCMK_OCF_DEGRADED_MASTER:
remapped_rc = PCMK_OCF_RUNNING_MASTER;
break;
default:
break;
}
if (rc != remapped_rc) {
crm_trace("Remapping monitor result %d to %d", rc, remapped_rc);
if (!node->details->shutdown || node->details->online) {
record_failed_op(xml_op, node, rsc, data_set);
}
}
return remapped_rc;
}
static void
unpack_rsc_op(pe_resource_t *rsc, pe_node_t *node, xmlNode *xml_op,
xmlNode **last_failure, enum action_fail_response *on_fail,
pe_working_set_t *data_set)
{
int rc = 0;
int task_id = 0;
int target_rc = 0;
int status = PCMK_LRM_OP_UNKNOWN;
guint interval_ms = 0;
const char *task = NULL;
const char *task_key = NULL;
const char *exit_reason = NULL;
bool expired = FALSE;
pe_resource_t *parent = rsc;
enum action_fail_response failure_strategy = action_fail_recover;
CRM_CHECK(rsc && node && xml_op, return);
target_rc = pe__target_rc_from_xml(xml_op);
task_key = get_op_key(xml_op);
task = crm_element_value(xml_op, XML_LRM_ATTR_TASK);
exit_reason = crm_element_value(xml_op, XML_LRM_ATTR_EXIT_REASON);
if (exit_reason == NULL) {
exit_reason = "";
}
crm_element_value_int(xml_op, XML_LRM_ATTR_RC, &rc);
crm_element_value_int(xml_op, XML_LRM_ATTR_CALLID, &task_id);
crm_element_value_int(xml_op, XML_LRM_ATTR_OPSTATUS, &status);
crm_element_value_ms(xml_op, XML_LRM_ATTR_INTERVAL_MS, &interval_ms);
CRM_CHECK(task != NULL, return);
CRM_CHECK(status <= PCMK_LRM_OP_INVALID, return);
CRM_CHECK(status >= PCMK_LRM_OP_PENDING, return);
if (!strcmp(task, CRMD_ACTION_NOTIFY) ||
!strcmp(task, CRMD_ACTION_METADATA)) {
/* safe to ignore these */
return;
}
if (is_not_set(rsc->flags, pe_rsc_unique)) {
parent = uber_parent(rsc);
}
pe_rsc_trace(rsc, "Unpacking task %s/%s (call_id=%d, status=%d, rc=%d) on %s (role=%s)",
task_key, task, task_id, status, rc, node->details->uname, role2text(rsc->role));
if (node->details->unclean) {
pe_rsc_trace(rsc, "Node %s (where %s is running) is unclean."
" Further action depends on the value of the stop's on-fail attribute",
node->details->uname, rsc->id);
}
/* It should be possible to call remap_monitor_rc() first then call
* check_operation_expiry() only if rc != target_rc, because there should
* never be a fail count without at least one unexpected result in the
* resource history. That would be more efficient by avoiding having to call
* check_operation_expiry() for expected results.
*
* However, we do have such configurations in the scheduler regression
* tests, even if it shouldn't be possible with the current code. It's
* probably a good idea anyway, but that would require updating the test
* inputs to something currently possible.
*/
if ((status != PCMK_LRM_OP_NOT_INSTALLED)
&& check_operation_expiry(rsc, node, rc, xml_op, data_set)) {
expired = TRUE;
}
if (!strcmp(task, CRMD_ACTION_STATUS)) {
rc = remap_monitor_rc(rc, xml_op, node, rsc, data_set);
}
if (expired && (rc != target_rc)) {
const char *magic = crm_element_value(xml_op, XML_ATTR_TRANSITION_MAGIC);
if (interval_ms == 0) {
crm_notice("Ignoring expired %s failure on %s "
CRM_XS " actual=%d expected=%d magic=%s",
task_key, node->details->uname, rc, target_rc, magic);
goto done;
} else if(node->details->online && node->details->unclean == FALSE) {
/* Reschedule the recurring monitor. CancelXmlOp() won't work at
* this stage, so as a hacky workaround, forcibly change the restart
* digest so check_action_definition() does what we want later.
*
* @TODO We should skip this if there is a newer successful monitor.
* Also, this causes rescheduling only if the history entry
* has an op-digest (which the expire-non-blocked-failure
* scheduler regression test doesn't, but that may not be a
* realistic scenario in production).
*/
crm_notice("Rescheduling %s after failure expired on %s "
CRM_XS " actual=%d expected=%d magic=%s",
task_key, node->details->uname, rc, target_rc, magic);
crm_xml_add(xml_op, XML_LRM_ATTR_RESTART_DIGEST, "calculated-failure-timeout");
goto done;
}
}
/* If the executor reported an operation status of anything but done or
* error, consider that final. But for done or error, we know better whether
* it should be treated as a failure or not, because we know the expected
* result.
*/
if(status == PCMK_LRM_OP_DONE || status == PCMK_LRM_OP_ERROR) {
status = determine_op_status(rsc, rc, target_rc, node, xml_op, on_fail, data_set);
pe_rsc_trace(rsc, "Remapped %s status to %d", task_key, status);
}
switch (status) {
case PCMK_LRM_OP_CANCELLED:
// Should never happen
pe_err("Resource history contains cancellation '%s' "
"(%s of %s on %s at %s)",
ID(xml_op), task, rsc->id, node->details->uname,
last_change_str(xml_op));
break;
case PCMK_LRM_OP_PENDING:
if (!strcmp(task, CRMD_ACTION_START)) {
set_bit(rsc->flags, pe_rsc_start_pending);
set_active(rsc);
} else if (!strcmp(task, CRMD_ACTION_PROMOTE)) {
rsc->role = RSC_ROLE_MASTER;
} else if (!strcmp(task, CRMD_ACTION_MIGRATE) && node->details->unclean) {
/* If a pending migrate_to action is out on a unclean node,
* we have to force the stop action on the target. */
const char *migrate_target = crm_element_value(xml_op, XML_LRM_ATTR_MIGRATE_TARGET);
pe_node_t *target = pe_find_node(data_set->nodes, migrate_target);
if (target) {
stop_action(rsc, target, FALSE);
}
}
if (rsc->pending_task == NULL) {
if ((interval_ms != 0) || strcmp(task, CRMD_ACTION_STATUS)) {
rsc->pending_task = strdup(task);
rsc->pending_node = node;
} else {
/* Pending probes are not printed, even if pending
* operations are requested. If someone ever requests that
* behavior, enable the below and the corresponding part of
* native.c:native_pending_task().
*/
#if 0
rsc->pending_task = strdup("probe");
rsc->pending_node = node;
#endif
}
}
break;
case PCMK_LRM_OP_DONE:
pe_rsc_trace(rsc, "%s of %s on %s completed at %s " CRM_XS " id=%s",
task, rsc->id, node->details->uname,
last_change_str(xml_op), ID(xml_op));
update_resource_state(rsc, node, xml_op, task, rc, *last_failure, on_fail, data_set);
break;
case PCMK_LRM_OP_NOT_INSTALLED:
failure_strategy = get_action_on_fail(rsc, task_key, task, data_set);
if (failure_strategy == action_fail_ignore) {
crm_warn("Cannot ignore failed %s of %s on %s: "
"Resource agent doesn't exist "
CRM_XS " status=%d rc=%d id=%s",
task, rsc->id, node->details->uname, status, rc,
ID(xml_op));
/* Also for printing it as "FAILED" by marking it as pe_rsc_failed later */
*on_fail = action_fail_migrate;
}
resource_location(parent, node, -INFINITY, "hard-error", data_set);
unpack_rsc_op_failure(rsc, node, rc, xml_op, last_failure, on_fail, data_set);
break;
case PCMK_LRM_OP_NOT_CONNECTED:
if (pe__is_guest_or_remote_node(node)
&& is_set(node->details->remote_rsc->flags, pe_rsc_managed)) {
/* We should never get into a situation where a managed remote
* connection resource is considered OK but a resource action
* behind the connection gets a "not connected" status. But as a
* fail-safe in case a bug or unusual circumstances do lead to
* that, ensure the remote connection is considered failed.
*/
set_bit(node->details->remote_rsc->flags, pe_rsc_failed);
}
// fall through
case PCMK_LRM_OP_ERROR:
case PCMK_LRM_OP_ERROR_HARD:
case PCMK_LRM_OP_ERROR_FATAL:
case PCMK_LRM_OP_TIMEOUT:
case PCMK_LRM_OP_NOTSUPPORTED:
case PCMK_LRM_OP_INVALID:
failure_strategy = get_action_on_fail(rsc, task_key, task, data_set);
if ((failure_strategy == action_fail_ignore)
|| (failure_strategy == action_fail_restart_container
&& !strcmp(task, CRMD_ACTION_STOP))) {
crm_warn("Pretending failed %s (%s%s%s) of %s on %s at %s "
"succeeded " CRM_XS " rc=%d id=%s",
task, services_ocf_exitcode_str(rc),
(*exit_reason? ": " : ""), exit_reason, rsc->id,
node->details->uname, last_change_str(xml_op), rc,
ID(xml_op));
update_resource_state(rsc, node, xml_op, task, target_rc, *last_failure, on_fail, data_set);
crm_xml_add(xml_op, XML_ATTR_UNAME, node->details->uname);
set_bit(rsc->flags, pe_rsc_failure_ignored);
record_failed_op(xml_op, node, rsc, data_set);
if (failure_strategy == action_fail_restart_container && *on_fail <= action_fail_recover) {
*on_fail = failure_strategy;
}
} else {
unpack_rsc_op_failure(rsc, node, rc, xml_op, last_failure, on_fail, data_set);
if(status == PCMK_LRM_OP_ERROR_HARD) {
do_crm_log(rc != PCMK_OCF_NOT_INSTALLED?LOG_ERR:LOG_NOTICE,
"Preventing %s from restarting on %s because "
"of hard failure (%s%s%s)" CRM_XS " rc=%d id=%s",
parent->id, node->details->uname,
services_ocf_exitcode_str(rc),
(*exit_reason? ": " : ""), exit_reason,
rc, ID(xml_op));
resource_location(parent, node, -INFINITY, "hard-error", data_set);
} else if(status == PCMK_LRM_OP_ERROR_FATAL) {
crm_err("Preventing %s from restarting anywhere because "
"of fatal failure (%s%s%s) " CRM_XS " rc=%d id=%s",
parent->id, services_ocf_exitcode_str(rc),
(*exit_reason? ": " : ""), exit_reason,
rc, ID(xml_op));
resource_location(parent, NULL, -INFINITY, "fatal-error", data_set);
}
}
break;
}
done:
pe_rsc_trace(rsc, "Resource %s after %s: role=%s, next=%s",
rsc->id, task, role2text(rsc->role),
role2text(rsc->next_role));
}
static void
add_node_attrs(xmlNode *xml_obj, pe_node_t *node, bool overwrite,
pe_working_set_t *data_set)
{
const char *cluster_name = NULL;
g_hash_table_insert(node->details->attrs,
strdup(CRM_ATTR_UNAME), strdup(node->details->uname));
g_hash_table_insert(node->details->attrs, strdup(CRM_ATTR_ID),
strdup(node->details->id));
if (safe_str_eq(node->details->id, data_set->dc_uuid)) {
data_set->dc_node = node;
node->details->is_dc = TRUE;
g_hash_table_insert(node->details->attrs,
strdup(CRM_ATTR_IS_DC), strdup(XML_BOOLEAN_TRUE));
} else {
g_hash_table_insert(node->details->attrs,
strdup(CRM_ATTR_IS_DC), strdup(XML_BOOLEAN_FALSE));
}
cluster_name = g_hash_table_lookup(data_set->config_hash, "cluster-name");
if (cluster_name) {
g_hash_table_insert(node->details->attrs, strdup(CRM_ATTR_CLUSTER_NAME),
strdup(cluster_name));
}
pe__unpack_dataset_nvpairs(xml_obj, XML_TAG_ATTR_SETS, NULL,
node->details->attrs, NULL, overwrite, data_set);
if (pe_node_attribute_raw(node, CRM_ATTR_SITE_NAME) == NULL) {
const char *site_name = pe_node_attribute_raw(node, "site-name");
if (site_name) {
g_hash_table_insert(node->details->attrs,
strdup(CRM_ATTR_SITE_NAME),
strdup(site_name));
} else if (cluster_name) {
/* Default to cluster-name if unset */
g_hash_table_insert(node->details->attrs,
strdup(CRM_ATTR_SITE_NAME),
strdup(cluster_name));
}
}
}
static GListPtr
extract_operations(const char *node, const char *rsc, xmlNode * rsc_entry, gboolean active_filter)
{
int counter = -1;
int stop_index = -1;
int start_index = -1;
xmlNode *rsc_op = NULL;
GListPtr gIter = NULL;
GListPtr op_list = NULL;
GListPtr sorted_op_list = NULL;
/* extract operations */
op_list = NULL;
sorted_op_list = NULL;
for (rsc_op = __xml_first_child_element(rsc_entry);
rsc_op != NULL; rsc_op = __xml_next_element(rsc_op)) {
if (crm_str_eq((const char *)rsc_op->name, XML_LRM_TAG_RSC_OP, TRUE)) {
crm_xml_add(rsc_op, "resource", rsc);
crm_xml_add(rsc_op, XML_ATTR_UNAME, node);
op_list = g_list_prepend(op_list, rsc_op);
}
}
if (op_list == NULL) {
/* if there are no operations, there is nothing to do */
return NULL;
}
sorted_op_list = g_list_sort(op_list, sort_op_by_callid);
/* create active recurring operations as optional */
if (active_filter == FALSE) {
return sorted_op_list;
}
op_list = NULL;
calculate_active_ops(sorted_op_list, &start_index, &stop_index);
for (gIter = sorted_op_list; gIter != NULL; gIter = gIter->next) {
xmlNode *rsc_op = (xmlNode *) gIter->data;
counter++;
if (start_index < stop_index) {
crm_trace("Skipping %s: not active", ID(rsc_entry));
break;
} else if (counter < start_index) {
crm_trace("Skipping %s: old", ID(rsc_op));
continue;
}
op_list = g_list_append(op_list, rsc_op);
}
g_list_free(sorted_op_list);
return op_list;
}
GListPtr
find_operations(const char *rsc, const char *node, gboolean active_filter,
pe_working_set_t * data_set)
{
GListPtr output = NULL;
GListPtr intermediate = NULL;
xmlNode *tmp = NULL;
xmlNode *status = find_xml_node(data_set->input, XML_CIB_TAG_STATUS, TRUE);
pe_node_t *this_node = NULL;
xmlNode *node_state = NULL;
for (node_state = __xml_first_child_element(status); node_state != NULL;
node_state = __xml_next_element(node_state)) {
if (crm_str_eq((const char *)node_state->name, XML_CIB_TAG_STATE, TRUE)) {
const char *uname = crm_element_value(node_state, XML_ATTR_UNAME);
if (node != NULL && safe_str_neq(uname, node)) {
continue;
}
this_node = pe_find_node(data_set->nodes, uname);
if(this_node == NULL) {
CRM_LOG_ASSERT(this_node != NULL);
continue;
} else if (pe__is_guest_or_remote_node(this_node)) {
determine_remote_online_status(data_set, this_node);
} else {
determine_online_status(node_state, this_node, data_set);
}
if (this_node->details->online || is_set(data_set->flags, pe_flag_stonith_enabled)) {
/* offline nodes run no resources...
* unless stonith is enabled in which case we need to
* make sure rsc start events happen after the stonith
*/
xmlNode *lrm_rsc = NULL;
tmp = find_xml_node(node_state, XML_CIB_TAG_LRM, FALSE);
tmp = find_xml_node(tmp, XML_LRM_TAG_RESOURCES, FALSE);
for (lrm_rsc = __xml_first_child_element(tmp); lrm_rsc != NULL;
lrm_rsc = __xml_next_element(lrm_rsc)) {
if (crm_str_eq((const char *)lrm_rsc->name, XML_LRM_TAG_RESOURCE, TRUE)) {
const char *rsc_id = crm_element_value(lrm_rsc, XML_ATTR_ID);
if (rsc != NULL && safe_str_neq(rsc_id, rsc)) {
continue;
}
intermediate = extract_operations(uname, rsc_id, lrm_rsc, active_filter);
output = g_list_concat(output, intermediate);
}
}
}
}
}
return output;
}
diff --git a/lib/pengine/utils.c b/lib/pengine/utils.c
index 2e72dcddb8..9a8c2f0ac7 100644
--- a/lib/pengine/utils.c
+++ b/lib/pengine/utils.c
@@ -1,2676 +1,2766 @@
/*
* Copyright 2004-2020 the Pacemaker project contributors
*
* The version control history for this file may have further details.
*
* This source code is licensed under the GNU Lesser General Public License
* version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY.
*/
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
extern xmlNode *get_object_root(const char *object_type, xmlNode * the_root);
void print_str_str(gpointer key, gpointer value, gpointer user_data);
gboolean ghash_free_str_str(gpointer key, gpointer value, gpointer user_data);
void unpack_operation(pe_action_t * action, xmlNode * xml_obj, pe_resource_t * container,
pe_working_set_t * data_set);
static xmlNode *find_rsc_op_entry_helper(pe_resource_t * rsc, const char *key,
gboolean include_disabled);
#if ENABLE_VERSIONED_ATTRS
pe_rsc_action_details_t *
pe_rsc_action_details(pe_action_t *action)
{
pe_rsc_action_details_t *details;
CRM_CHECK(action != NULL, return NULL);
if (action->action_details == NULL) {
action->action_details = calloc(1, sizeof(pe_rsc_action_details_t));
CRM_CHECK(action->action_details != NULL, return NULL);
}
details = (pe_rsc_action_details_t *) action->action_details;
if (details->versioned_parameters == NULL) {
details->versioned_parameters = create_xml_node(NULL,
XML_TAG_OP_VER_ATTRS);
}
if (details->versioned_meta == NULL) {
details->versioned_meta = create_xml_node(NULL, XML_TAG_OP_VER_META);
}
return details;
}
static void
pe_free_rsc_action_details(pe_action_t *action)
{
pe_rsc_action_details_t *details;
if ((action == NULL) || (action->action_details == NULL)) {
return;
}
details = (pe_rsc_action_details_t *) action->action_details;
if (details->versioned_parameters) {
free_xml(details->versioned_parameters);
}
if (details->versioned_meta) {
free_xml(details->versioned_meta);
}
action->action_details = NULL;
}
#endif
/*!
* \internal
* \brief Check whether we can fence a particular node
*
* \param[in] data_set Working set for cluster
* \param[in] node Name of node to check
*
* \return true if node can be fenced, false otherwise
*/
bool
pe_can_fence(pe_working_set_t *data_set, pe_node_t *node)
{
if (pe__is_guest_node(node)) {
/* Guest nodes are fenced by stopping their container resource. We can
* do that if the container's host is either online or fenceable.
*/
pe_resource_t *rsc = node->details->remote_rsc->container;
for (GList *n = rsc->running_on; n != NULL; n = n->next) {
pe_node_t *container_node = n->data;
if (!container_node->details->online
&& !pe_can_fence(data_set, container_node)) {
return false;
}
}
return true;
} else if(is_not_set(data_set->flags, pe_flag_stonith_enabled)) {
return false; /* Turned off */
} else if (is_not_set(data_set->flags, pe_flag_have_stonith_resource)) {
return false; /* No devices */
} else if (is_set(data_set->flags, pe_flag_have_quorum)) {
return true;
} else if (data_set->no_quorum_policy == no_quorum_ignore) {
return true;
} else if(node == NULL) {
return false;
} else if(node->details->online) {
crm_notice("We can fence %s without quorum because they're in our membership", node->details->uname);
return true;
}
crm_trace("Cannot fence %s", node->details->uname);
return false;
}
pe_node_t *
node_copy(const pe_node_t *this_node)
{
pe_node_t *new_node = NULL;
CRM_CHECK(this_node != NULL, return NULL);
new_node = calloc(1, sizeof(pe_node_t));
CRM_ASSERT(new_node != NULL);
crm_trace("Copying %p (%s) to %p", this_node, this_node->details->uname, new_node);
new_node->rsc_discover_mode = this_node->rsc_discover_mode;
new_node->weight = this_node->weight;
new_node->fixed = this_node->fixed;
new_node->details = this_node->details;
return new_node;
}
/* any node in list1 or list2 and not in the other gets a score of -INFINITY */
void
node_list_exclude(GHashTable * hash, GListPtr list, gboolean merge_scores)
{
GHashTable *result = hash;
pe_node_t *other_node = NULL;
GListPtr gIter = list;
GHashTableIter iter;
pe_node_t *node = NULL;
g_hash_table_iter_init(&iter, hash);
while (g_hash_table_iter_next(&iter, NULL, (void **)&node)) {
other_node = pe_find_node_id(list, node->details->id);
if (other_node == NULL) {
node->weight = -INFINITY;
} else if (merge_scores) {
node->weight = merge_weights(node->weight, other_node->weight);
}
}
for (; gIter != NULL; gIter = gIter->next) {
pe_node_t *node = (pe_node_t *) gIter->data;
other_node = pe_hash_table_lookup(result, node->details->id);
if (other_node == NULL) {
pe_node_t *new_node = node_copy(node);
new_node->weight = -INFINITY;
g_hash_table_insert(result, (gpointer) new_node->details->id, new_node);
}
}
}
GHashTable *
node_hash_from_list(GListPtr list)
{
GListPtr gIter = list;
GHashTable *result = g_hash_table_new_full(crm_str_hash, g_str_equal, NULL,
free);
for (; gIter != NULL; gIter = gIter->next) {
pe_node_t *node = (pe_node_t *) gIter->data;
pe_node_t *n = node_copy(node);
g_hash_table_insert(result, (gpointer) n->details->id, n);
}
return result;
}
GListPtr
node_list_dup(GListPtr list1, gboolean reset, gboolean filter)
{
GListPtr result = NULL;
GListPtr gIter = list1;
for (; gIter != NULL; gIter = gIter->next) {
pe_node_t *new_node = NULL;
pe_node_t *this_node = (pe_node_t *) gIter->data;
if (filter && this_node->weight < 0) {
continue;
}
new_node = node_copy(this_node);
if (reset) {
new_node->weight = 0;
}
if (new_node != NULL) {
result = g_list_prepend(result, new_node);
}
}
return result;
}
gint
sort_node_uname(gconstpointer a, gconstpointer b)
{
const char *name_a = ((const pe_node_t *) a)->details->uname;
const char *name_b = ((const pe_node_t *) b)->details->uname;
while (*name_a && *name_b) {
if (isdigit(*name_a) && isdigit(*name_b)) {
// If node names contain a number, sort numerically
char *end_a = NULL;
char *end_b = NULL;
long num_a = strtol(name_a, &end_a, 10);
long num_b = strtol(name_b, &end_b, 10);
// allow ordering e.g. 007 > 7
size_t len_a = end_a - name_a;
size_t len_b = end_b - name_b;
if (num_a < num_b) {
return -1;
} else if (num_a > num_b) {
return 1;
} else if (len_a < len_b) {
return -1;
} else if (len_a > len_b) {
return 1;
}
name_a = end_a;
name_b = end_b;
} else {
// Compare non-digits case-insensitively
int lower_a = tolower(*name_a);
int lower_b = tolower(*name_b);
if (lower_a < lower_b) {
return -1;
} else if (lower_a > lower_b) {
return 1;
}
++name_a;
++name_b;
}
}
if (!*name_a && *name_b) {
return -1;
} else if (*name_a && !*name_b) {
return 1;
}
return 0;
}
/*!
* \internal
* \brief Output node weights to stdout
*
* \param[in] rsc Use allowed nodes for this resource
* \param[in] comment Text description to prefix lines with
* \param[in] nodes If rsc is not specified, use these nodes
*/
static void
pe__output_node_weights(pe_resource_t *rsc, const char *comment,
GHashTable *nodes)
{
char score[128]; // Stack-allocated since this is called frequently
// Sort the nodes so the output is consistent for regression tests
GList *list = g_list_sort(g_hash_table_get_values(nodes), sort_node_uname);
for (GList *gIter = list; gIter != NULL; gIter = gIter->next) {
pe_node_t *node = (pe_node_t *) gIter->data;
score2char_stack(node->weight, score, sizeof(score));
if (rsc) {
printf("%s: %s allocation score on %s: %s\n",
comment, rsc->id, node->details->uname, score);
} else {
printf("%s: %s = %s\n", comment, node->details->uname, score);
}
}
g_list_free(list);
}
/*!
* \internal
* \brief Log node weights at trace level
*
* \param[in] file Caller's filename
* \param[in] function Caller's function name
* \param[in] line Caller's line number
* \param[in] rsc Use allowed nodes for this resource
* \param[in] comment Text description to prefix lines with
* \param[in] nodes If rsc is not specified, use these nodes
*/
static void
pe__log_node_weights(const char *file, const char *function, int line,
pe_resource_t *rsc, const char *comment, GHashTable *nodes)
{
GHashTableIter iter;
pe_node_t *node = NULL;
char score[128]; // Stack-allocated since this is called frequently
// Don't waste time if we're not tracing at this point
pcmk__log_else(LOG_TRACE, return);
g_hash_table_iter_init(&iter, nodes);
while (g_hash_table_iter_next(&iter, NULL, (void **) &node)) {
score2char_stack(node->weight, score, sizeof(score));
if (rsc) {
qb_log_from_external_source(function, file,
"%s: %s allocation score on %s: %s",
LOG_TRACE, line, 0,
comment, rsc->id,
node->details->uname, score);
} else {
qb_log_from_external_source(function, file, "%s: %s = %s",
LOG_TRACE, line, 0,
comment, node->details->uname,
score);
}
}
}
/*!
* \internal
* \brief Log or output node weights
*
* \param[in] file Caller's filename
* \param[in] function Caller's function name
* \param[in] line Caller's line number
* \param[in] to_log Log if true, otherwise output
* \param[in] rsc Use allowed nodes for this resource
* \param[in] comment Text description to prefix lines with
* \param[in] nodes If rsc is not specified, use these nodes
*/
void
pe__show_node_weights_as(const char *file, const char *function, int line,
bool to_log, pe_resource_t *rsc, const char *comment,
GHashTable *nodes)
{
if (rsc != NULL) {
if (is_set(rsc->flags, pe_rsc_orphan)) {
// Don't show allocation scores for orphans
return;
}
nodes = rsc->allowed_nodes;
}
if (nodes == NULL) {
// Nothing to show
return;
}
if (to_log) {
pe__log_node_weights(file, function, line, rsc, comment, nodes);
} else {
pe__output_node_weights(rsc, comment, nodes);
}
// If this resource has children, repeat recursively for each
if (rsc && rsc->children) {
for (GList *gIter = rsc->children; gIter != NULL; gIter = gIter->next) {
pe_resource_t *child = (pe_resource_t *) gIter->data;
pe__show_node_weights_as(file, function, line, to_log, child,
comment, nodes);
}
}
}
static void
append_dump_text(gpointer key, gpointer value, gpointer user_data)
{
char **dump_text = user_data;
char *new_text = crm_strdup_printf("%s %s=%s",
*dump_text, (char *)key, (char *)value);
free(*dump_text);
*dump_text = new_text;
}
void
dump_node_capacity(int level, const char *comment, pe_node_t * node)
{
char *dump_text = crm_strdup_printf("%s: %s capacity:",
comment, node->details->uname);
g_hash_table_foreach(node->details->utilization, append_dump_text, &dump_text);
if (level == LOG_STDOUT) {
fprintf(stdout, "%s\n", dump_text);
} else {
crm_trace("%s", dump_text);
}
free(dump_text);
}
void
dump_rsc_utilization(int level, const char *comment, pe_resource_t * rsc, pe_node_t * node)
{
char *dump_text = crm_strdup_printf("%s: %s utilization on %s:",
comment, rsc->id, node->details->uname);
g_hash_table_foreach(rsc->utilization, append_dump_text, &dump_text);
switch (level) {
case LOG_STDOUT:
fprintf(stdout, "%s\n", dump_text);
break;
case LOG_NEVER:
break;
default:
crm_trace("%s", dump_text);
}
free(dump_text);
}
gint
sort_rsc_index(gconstpointer a, gconstpointer b)
{
const pe_resource_t *resource1 = (const pe_resource_t *)a;
const pe_resource_t *resource2 = (const pe_resource_t *)b;
if (a == NULL && b == NULL) {
return 0;
}
if (a == NULL) {
return 1;
}
if (b == NULL) {
return -1;
}
if (resource1->sort_index > resource2->sort_index) {
return -1;
}
if (resource1->sort_index < resource2->sort_index) {
return 1;
}
return 0;
}
gint
sort_rsc_priority(gconstpointer a, gconstpointer b)
{
const pe_resource_t *resource1 = (const pe_resource_t *)a;
const pe_resource_t *resource2 = (const pe_resource_t *)b;
if (a == NULL && b == NULL) {
return 0;
}
if (a == NULL) {
return 1;
}
if (b == NULL) {
return -1;
}
if (resource1->priority > resource2->priority) {
return -1;
}
if (resource1->priority < resource2->priority) {
return 1;
}
return 0;
}
pe_action_t *
custom_action(pe_resource_t * rsc, char *key, const char *task,
pe_node_t * on_node, gboolean optional, gboolean save_action,
pe_working_set_t * data_set)
{
pe_action_t *action = NULL;
GListPtr possible_matches = NULL;
CRM_CHECK(key != NULL, return NULL);
CRM_CHECK(task != NULL, free(key); return NULL);
if (save_action && rsc != NULL) {
possible_matches = find_actions(rsc->actions, key, on_node);
} else if(save_action) {
#if 0
action = g_hash_table_lookup(data_set->singletons, key);
#else
/* More expensive but takes 'node' into account */
possible_matches = find_actions(data_set->actions, key, on_node);
#endif
}
if(data_set->singletons == NULL) {
data_set->singletons = g_hash_table_new_full(crm_str_hash, g_str_equal, NULL, NULL);
}
if (possible_matches != NULL) {
if (pcmk__list_of_multiple(possible_matches)) {
pe_warn("Action %s for %s on %s exists %d times",
task, rsc ? rsc->id : "",
on_node ? on_node->details->uname : "", g_list_length(possible_matches));
}
action = g_list_nth_data(possible_matches, 0);
pe_rsc_trace(rsc, "Found existing action %d (%s) for %s (%s) on %s",
action->id, action->uuid,
(rsc? rsc->id : "no resource"), task,
(on_node? on_node->details->uname : "no node"));
g_list_free(possible_matches);
}
if (action == NULL) {
if (save_action) {
pe_rsc_trace(rsc, "Creating %s action %d: %s for %s (%s) on %s",
(optional? "optional" : " mandatory"),
data_set->action_id, key,
(rsc? rsc->id : "no resource"), task,
(on_node? on_node->details->uname : "no node"));
}
action = calloc(1, sizeof(pe_action_t));
if (save_action) {
action->id = data_set->action_id++;
} else {
action->id = 0;
}
action->rsc = rsc;
CRM_ASSERT(task != NULL);
action->task = strdup(task);
if (on_node) {
action->node = node_copy(on_node);
}
action->uuid = strdup(key);
if (safe_str_eq(task, CRM_OP_LRM_DELETE)) {
// Resource history deletion for a node can be done on the DC
pe_set_action_bit(action, pe_action_dc);
}
pe_set_action_bit(action, pe_action_runnable);
if (optional) {
pe_set_action_bit(action, pe_action_optional);
} else {
pe_clear_action_bit(action, pe_action_optional);
}
action->extra = crm_str_table_new();
action->meta = crm_str_table_new();
if (save_action) {
data_set->actions = g_list_prepend(data_set->actions, action);
if(rsc == NULL) {
g_hash_table_insert(data_set->singletons, action->uuid, action);
}
}
if (rsc != NULL) {
action->op_entry = find_rsc_op_entry_helper(rsc, key, TRUE);
unpack_operation(action, action->op_entry, rsc->container, data_set);
if (save_action) {
rsc->actions = g_list_prepend(rsc->actions, action);
}
}
if (save_action) {
pe_rsc_trace(rsc, "Action %d created", action->id);
}
}
if (!optional && is_set(action->flags, pe_action_optional)) {
pe_rsc_trace(rsc, "Unset optional on action %d", action->id);
pe_clear_action_bit(action, pe_action_optional);
}
if (rsc != NULL) {
enum action_tasks a_task = text2task(action->task);
int warn_level = LOG_TRACE;
if (save_action) {
warn_level = LOG_WARNING;
}
if (is_set(action->flags, pe_action_have_node_attrs) == FALSE
&& action->node != NULL && action->op_entry != NULL) {
pe_set_action_bit(action, pe_action_have_node_attrs);
pe__unpack_dataset_nvpairs(action->op_entry, XML_TAG_ATTR_SETS,
action->node->details->attrs,
action->extra, NULL, FALSE, data_set);
}
if (is_set(action->flags, pe_action_pseudo)) {
/* leave untouched */
} else if (action->node == NULL) {
pe_rsc_trace(rsc, "Unset runnable on %s", action->uuid);
pe_clear_action_bit(action, pe_action_runnable);
} else if (is_not_set(rsc->flags, pe_rsc_managed)
&& g_hash_table_lookup(action->meta,
XML_LRM_ATTR_INTERVAL_MS) == NULL) {
crm_debug("Action %s (unmanaged)", action->uuid);
pe_rsc_trace(rsc, "Set optional on %s", action->uuid);
pe_set_action_bit(action, pe_action_optional);
/* action->runnable = FALSE; */
} else if (is_not_set(action->flags, pe_action_dc)
&& !(action->node->details->online)
&& (!pe__is_guest_node(action->node)
|| action->node->details->remote_requires_reset)) {
pe_clear_action_bit(action, pe_action_runnable);
do_crm_log(warn_level, "Action %s on %s is unrunnable (offline)",
action->uuid, action->node->details->uname);
if (is_set(action->rsc->flags, pe_rsc_managed)
&& save_action && a_task == stop_rsc
&& action->node->details->unclean == FALSE) {
- pe_fence_node(data_set, action->node, "resource actions are unrunnable");
+ pe_fence_node(data_set, action->node, "resource actions are unrunnable", FALSE);
}
} else if (is_not_set(action->flags, pe_action_dc)
&& action->node->details->pending) {
pe_clear_action_bit(action, pe_action_runnable);
do_crm_log(warn_level, "Action %s on %s is unrunnable (pending)",
action->uuid, action->node->details->uname);
} else if (action->needs == rsc_req_nothing) {
pe_rsc_trace(rsc, "Action %s does not require anything", action->uuid);
pe_action_set_reason(action, NULL, TRUE);
if (pe__is_guest_node(action->node)
&& !pe_can_fence(data_set, action->node)) {
/* An action that requires nothing usually does not require any
* fencing in order to be runnable. However, there is an
* exception: an action cannot be completed if it is on a guest
* node whose host is unclean and cannot be fenced.
*/
pe_clear_action_bit(action, pe_action_runnable);
crm_debug("%s\t%s (cancelled : host cannot be fenced)",
action->node->details->uname, action->uuid);
} else {
pe_set_action_bit(action, pe_action_runnable);
}
#if 0
/*
* No point checking this
* - if we don't have quorum we can't stonith anyway
*/
} else if (action->needs == rsc_req_stonith) {
crm_trace("Action %s requires only stonith", action->uuid);
action->runnable = TRUE;
#endif
} else if (is_set(data_set->flags, pe_flag_have_quorum) == FALSE
&& data_set->no_quorum_policy == no_quorum_stop) {
pe_action_set_flag_reason(__FUNCTION__, __LINE__, action, NULL, "no quorum", pe_action_runnable, TRUE);
crm_debug("%s\t%s (cancelled : quorum)", action->node->details->uname, action->uuid);
} else if (is_set(data_set->flags, pe_flag_have_quorum) == FALSE
&& data_set->no_quorum_policy == no_quorum_freeze) {
pe_rsc_trace(rsc, "Check resource is already active: %s %s %s %s", rsc->id, action->uuid, role2text(rsc->next_role), role2text(rsc->role));
if (rsc->fns->active(rsc, TRUE) == FALSE || rsc->next_role > rsc->role) {
pe_action_set_flag_reason(__FUNCTION__, __LINE__, action, NULL, "quorum freeze", pe_action_runnable, TRUE);
pe_rsc_debug(rsc, "%s\t%s (cancelled : quorum freeze)",
action->node->details->uname, action->uuid);
}
} else if(is_not_set(action->flags, pe_action_runnable)) {
pe_rsc_trace(rsc, "Action %s is runnable", action->uuid);
//pe_action_set_reason(action, NULL, TRUE);
pe_set_action_bit(action, pe_action_runnable);
}
if (save_action) {
switch (a_task) {
case stop_rsc:
set_bit(rsc->flags, pe_rsc_stopping);
break;
case start_rsc:
clear_bit(rsc->flags, pe_rsc_starting);
if (is_set(action->flags, pe_action_runnable)) {
set_bit(rsc->flags, pe_rsc_starting);
}
break;
default:
break;
}
}
}
free(key);
return action;
}
static const char *
unpack_operation_on_fail(pe_action_t * action)
{
const char *value = g_hash_table_lookup(action->meta, XML_OP_ATTR_ON_FAIL);
if (safe_str_eq(action->task, CRMD_ACTION_STOP) && safe_str_eq(value, "standby")) {
pcmk__config_err("Resetting '" XML_OP_ATTR_ON_FAIL "' for %s stop "
"action to default value because 'standby' is not "
"allowed for stop", action->rsc->id);
return NULL;
} else if (safe_str_eq(action->task, CRMD_ACTION_DEMOTE) && !value) {
/* demote on_fail defaults to master monitor value if present */
xmlNode *operation = NULL;
const char *name = NULL;
const char *role = NULL;
const char *on_fail = NULL;
const char *interval_spec = NULL;
const char *enabled = NULL;
CRM_CHECK(action->rsc != NULL, return NULL);
for (operation = __xml_first_child_element(action->rsc->ops_xml);
operation && !value; operation = __xml_next_element(operation)) {
if (!crm_str_eq((const char *)operation->name, "op", TRUE)) {
continue;
}
name = crm_element_value(operation, "name");
role = crm_element_value(operation, "role");
on_fail = crm_element_value(operation, XML_OP_ATTR_ON_FAIL);
enabled = crm_element_value(operation, "enabled");
interval_spec = crm_element_value(operation, XML_LRM_ATTR_INTERVAL);
if (!on_fail) {
continue;
} else if (enabled && !crm_is_true(enabled)) {
continue;
} else if (safe_str_neq(name, "monitor") || safe_str_neq(role, "Master")) {
continue;
} else if (crm_parse_interval_spec(interval_spec) == 0) {
continue;
}
value = on_fail;
}
} else if (safe_str_eq(action->task, CRM_OP_LRM_DELETE)) {
value = "ignore";
}
return value;
}
static xmlNode *
find_min_interval_mon(pe_resource_t * rsc, gboolean include_disabled)
{
guint interval_ms = 0;
guint min_interval_ms = G_MAXUINT;
const char *name = NULL;
const char *value = NULL;
const char *interval_spec = NULL;
xmlNode *op = NULL;
xmlNode *operation = NULL;
for (operation = __xml_first_child_element(rsc->ops_xml); operation != NULL;
operation = __xml_next_element(operation)) {
if (crm_str_eq((const char *)operation->name, "op", TRUE)) {
name = crm_element_value(operation, "name");
interval_spec = crm_element_value(operation, XML_LRM_ATTR_INTERVAL);
value = crm_element_value(operation, "enabled");
if (!include_disabled && value && crm_is_true(value) == FALSE) {
continue;
}
if (safe_str_neq(name, RSC_STATUS)) {
continue;
}
interval_ms = crm_parse_interval_spec(interval_spec);
if (interval_ms && (interval_ms < min_interval_ms)) {
min_interval_ms = interval_ms;
op = operation;
}
}
}
return op;
}
static int
unpack_start_delay(const char *value, GHashTable *meta)
{
int start_delay = 0;
if (value != NULL) {
start_delay = crm_get_msec(value);
if (start_delay < 0) {
start_delay = 0;
}
if (meta) {
g_hash_table_replace(meta, strdup(XML_OP_ATTR_START_DELAY), crm_itoa(start_delay));
}
}
return start_delay;
}
// true if value contains valid, non-NULL interval origin for recurring op
static bool
unpack_interval_origin(const char *value, xmlNode *xml_obj, guint interval_ms,
crm_time_t *now, long long *start_delay)
{
long long result = 0;
guint interval_sec = interval_ms / 1000;
crm_time_t *origin = NULL;
// Ignore unspecified values and non-recurring operations
if ((value == NULL) || (interval_ms == 0) || (now == NULL)) {
return false;
}
// Parse interval origin from text
origin = crm_time_new(value);
if (origin == NULL) {
pcmk__config_err("Ignoring '" XML_OP_ATTR_ORIGIN "' for operation "
"'%s' because '%s' is not valid",
(ID(xml_obj)? ID(xml_obj) : "(missing ID)"), value);
return false;
}
// Get seconds since origin (negative if origin is in the future)
result = crm_time_get_seconds(now) - crm_time_get_seconds(origin);
crm_time_free(origin);
// Calculate seconds from closest interval to now
result = result % interval_sec;
// Calculate seconds remaining until next interval
result = ((result <= 0)? 0 : interval_sec) - result;
crm_info("Calculated a start delay of %llds for operation '%s'",
result,
(ID(xml_obj)? ID(xml_obj) : "(unspecified)"));
if (start_delay != NULL) {
*start_delay = result * 1000; // milliseconds
}
return true;
}
static int
unpack_timeout(const char *value)
{
int timeout = crm_get_msec(value);
if (timeout < 0) {
timeout = crm_get_msec(CRM_DEFAULT_OP_TIMEOUT_S);
}
return timeout;
}
int
pe_get_configured_timeout(pe_resource_t *rsc, const char *action, pe_working_set_t *data_set)
{
xmlNode *child = NULL;
const char *timeout = NULL;
int timeout_ms = 0;
for (child = first_named_child(rsc->ops_xml, XML_ATTR_OP);
child != NULL; child = crm_next_same_xml(child)) {
if (safe_str_eq(action, crm_element_value(child, XML_NVPAIR_ATTR_NAME))) {
timeout = crm_element_value(child, XML_ATTR_TIMEOUT);
break;
}
}
if (timeout == NULL && data_set->op_defaults) {
GHashTable *action_meta = crm_str_table_new();
pe__unpack_dataset_nvpairs(data_set->op_defaults, XML_TAG_META_SETS,
NULL, action_meta, NULL, FALSE, data_set);
timeout = g_hash_table_lookup(action_meta, XML_ATTR_TIMEOUT);
}
// @TODO check meta-attributes (including versioned meta-attributes)
// @TODO maybe use min-interval monitor timeout as default for monitors
timeout_ms = crm_get_msec(timeout);
if (timeout_ms < 0) {
timeout_ms = crm_get_msec(CRM_DEFAULT_OP_TIMEOUT_S);
}
return timeout_ms;
}
#if ENABLE_VERSIONED_ATTRS
static void
unpack_versioned_meta(xmlNode *versioned_meta, xmlNode *xml_obj,
guint interval_ms, crm_time_t *now)
{
xmlNode *attrs = NULL;
xmlNode *attr = NULL;
for (attrs = __xml_first_child_element(versioned_meta); attrs != NULL;
attrs = __xml_next_element(attrs)) {
for (attr = __xml_first_child_element(attrs); attr != NULL;
attr = __xml_next_element(attr)) {
const char *name = crm_element_value(attr, XML_NVPAIR_ATTR_NAME);
const char *value = crm_element_value(attr, XML_NVPAIR_ATTR_VALUE);
if (safe_str_eq(name, XML_OP_ATTR_START_DELAY)) {
int start_delay = unpack_start_delay(value, NULL);
crm_xml_add_int(attr, XML_NVPAIR_ATTR_VALUE, start_delay);
} else if (safe_str_eq(name, XML_OP_ATTR_ORIGIN)) {
long long start_delay = 0;
if (unpack_interval_origin(value, xml_obj, interval_ms, now,
&start_delay)) {
crm_xml_add(attr, XML_NVPAIR_ATTR_NAME,
XML_OP_ATTR_START_DELAY);
crm_xml_add_ll(attr, XML_NVPAIR_ATTR_VALUE, start_delay);
}
} else if (safe_str_eq(name, XML_ATTR_TIMEOUT)) {
int timeout = unpack_timeout(value);
crm_xml_add_int(attr, XML_NVPAIR_ATTR_VALUE, timeout);
}
}
}
}
#endif
/*!
* \brief Unpack operation XML into an action structure
*
* Unpack an operation's meta-attributes (normalizing the interval, timeout,
* and start delay values as integer milliseconds), requirements, and
* failure policy.
*
* \param[in,out] action Action to unpack into
* \param[in] xml_obj Operation XML (or NULL if all defaults)
* \param[in] container Resource that contains affected resource, if any
* \param[in] data_set Cluster state
*/
void
unpack_operation(pe_action_t * action, xmlNode * xml_obj, pe_resource_t * container,
pe_working_set_t * data_set)
{
guint interval_ms = 0;
int timeout = 0;
char *value_ms = NULL;
const char *value = NULL;
const char *field = NULL;
char *default_timeout = NULL;
#if ENABLE_VERSIONED_ATTRS
pe_rsc_action_details_t *rsc_details = NULL;
#endif
CRM_CHECK(action && action->rsc, return);
// Cluster-wide
pe__unpack_dataset_nvpairs(data_set->op_defaults, XML_TAG_META_SETS, NULL,
action->meta, NULL, FALSE, data_set);
// Probe timeouts default differently, so handle timeout default later
default_timeout = g_hash_table_lookup(action->meta, XML_ATTR_TIMEOUT);
if (default_timeout) {
default_timeout = strdup(default_timeout);
g_hash_table_remove(action->meta, XML_ATTR_TIMEOUT);
}
if (xml_obj) {
xmlAttrPtr xIter = NULL;
// take precedence over defaults
pe__unpack_dataset_nvpairs(xml_obj, XML_TAG_META_SETS, NULL,
action->meta, NULL, TRUE, data_set);
#if ENABLE_VERSIONED_ATTRS
rsc_details = pe_rsc_action_details(action);
pe_unpack_versioned_attributes(data_set->input, xml_obj,
XML_TAG_ATTR_SETS, NULL,
rsc_details->versioned_parameters,
data_set->now, NULL);
pe_unpack_versioned_attributes(data_set->input, xml_obj,
XML_TAG_META_SETS, NULL,
rsc_details->versioned_meta,
data_set->now, NULL);
#endif
/* Anything set as an XML property has highest precedence.
* This ensures we use the name and interval from the tag.
*/
for (xIter = xml_obj->properties; xIter; xIter = xIter->next) {
const char *prop_name = (const char *)xIter->name;
const char *prop_value = crm_element_value(xml_obj, prop_name);
g_hash_table_replace(action->meta, strdup(prop_name), strdup(prop_value));
}
}
g_hash_table_remove(action->meta, "id");
// Normalize interval to milliseconds
field = XML_LRM_ATTR_INTERVAL;
value = g_hash_table_lookup(action->meta, field);
if (value != NULL) {
interval_ms = crm_parse_interval_spec(value);
} else if ((xml_obj == NULL) && !strcmp(action->task, RSC_STATUS)) {
/* An orphaned recurring monitor will not have any XML. However, we
* want the interval to be set, so the action can be properly detected
* as a recurring monitor. Parse it from the key in this case.
*/
parse_op_key(action->uuid, NULL, NULL, &interval_ms);
}
if (interval_ms > 0) {
value_ms = crm_strdup_printf("%u", interval_ms);
g_hash_table_replace(action->meta, strdup(field), value_ms);
} else if (value) {
g_hash_table_remove(action->meta, field);
}
// Handle timeout default, now that we know the interval
if (g_hash_table_lookup(action->meta, XML_ATTR_TIMEOUT)) {
free(default_timeout);
} else {
// Probe timeouts default to minimum-interval monitor's
if (safe_str_eq(action->task, RSC_STATUS) && (interval_ms == 0)) {
xmlNode *min_interval_mon = find_min_interval_mon(action->rsc, FALSE);
if (min_interval_mon) {
value = crm_element_value(min_interval_mon, XML_ATTR_TIMEOUT);
if (value) {
crm_trace("\t%s defaults to minimum-interval monitor's timeout '%s'",
action->uuid, value);
free(default_timeout);
default_timeout = strdup(value);
}
}
}
if (default_timeout) {
g_hash_table_insert(action->meta, strdup(XML_ATTR_TIMEOUT),
default_timeout);
}
}
if (safe_str_neq(action->task, RSC_START)
&& safe_str_neq(action->task, RSC_PROMOTE)) {
action->needs = rsc_req_nothing;
value = "nothing (not start/promote)";
} else if (is_set(action->rsc->flags, pe_rsc_needs_fencing)) {
action->needs = rsc_req_stonith;
value = "fencing (resource)";
} else if (is_set(action->rsc->flags, pe_rsc_needs_quorum)) {
action->needs = rsc_req_quorum;
value = "quorum (resource)";
} else {
action->needs = rsc_req_nothing;
value = "nothing (resource)";
}
pe_rsc_trace(action->rsc, "\tAction %s requires: %s", action->uuid, value);
value = unpack_operation_on_fail(action);
if (value == NULL) {
} else if (safe_str_eq(value, "block")) {
action->on_fail = action_fail_block;
g_hash_table_insert(action->meta, strdup(XML_OP_ATTR_ON_FAIL), strdup("block"));
value = "block"; // The above could destroy the original string
} else if (safe_str_eq(value, "fence")) {
action->on_fail = action_fail_fence;
value = "node fencing";
if (is_set(data_set->flags, pe_flag_stonith_enabled) == FALSE) {
pcmk__config_err("Resetting '" XML_OP_ATTR_ON_FAIL "' for "
"operation '%s' to 'stop' because 'fence' is not "
"valid when fencing is disabled", action->uuid);
action->on_fail = action_fail_stop;
action->fail_role = RSC_ROLE_STOPPED;
value = "stop resource";
}
} else if (safe_str_eq(value, "standby")) {
action->on_fail = action_fail_standby;
value = "node standby";
} else if (safe_str_eq(value, "ignore")
|| safe_str_eq(value, "nothing")) {
action->on_fail = action_fail_ignore;
value = "ignore";
} else if (safe_str_eq(value, "migrate")) {
action->on_fail = action_fail_migrate;
value = "force migration";
} else if (safe_str_eq(value, "stop")) {
action->on_fail = action_fail_stop;
action->fail_role = RSC_ROLE_STOPPED;
value = "stop resource";
} else if (safe_str_eq(value, "restart")) {
action->on_fail = action_fail_recover;
value = "restart (and possibly migrate)";
} else if (safe_str_eq(value, "restart-container")) {
if (container) {
action->on_fail = action_fail_restart_container;
value = "restart container (and possibly migrate)";
} else {
value = NULL;
}
} else {
pe_err("Resource %s: Unknown failure type (%s)", action->rsc->id, value);
value = NULL;
}
/* defaults */
if (value == NULL && container) {
action->on_fail = action_fail_restart_container;
value = "restart container (and possibly migrate) (default)";
/* For remote nodes, ensure that any failure that results in dropping an
* active connection to the node results in fencing of the node.
*
* There are only two action failures that don't result in fencing.
* 1. probes - probe failures are expected.
* 2. start - a start failure indicates that an active connection does not already
* exist. The user can set op on-fail=fence if they really want to fence start
* failures. */
} else if (((value == NULL) || !is_set(action->rsc->flags, pe_rsc_managed)) &&
(pe__resource_is_remote_conn(action->rsc, data_set) &&
!(safe_str_eq(action->task, CRMD_ACTION_STATUS) && (interval_ms == 0)) &&
(safe_str_neq(action->task, CRMD_ACTION_START)))) {
if (!is_set(action->rsc->flags, pe_rsc_managed)) {
action->on_fail = action_fail_stop;
action->fail_role = RSC_ROLE_STOPPED;
value = "stop unmanaged remote node (enforcing default)";
} else {
if (is_set(data_set->flags, pe_flag_stonith_enabled)) {
value = "fence remote node (default)";
} else {
value = "recover remote node connection (default)";
}
if (action->rsc->remote_reconnect_ms) {
action->fail_role = RSC_ROLE_STOPPED;
}
action->on_fail = action_fail_reset_remote;
}
} else if (value == NULL && safe_str_eq(action->task, CRMD_ACTION_STOP)) {
if (is_set(data_set->flags, pe_flag_stonith_enabled)) {
action->on_fail = action_fail_fence;
value = "resource fence (default)";
} else {
action->on_fail = action_fail_block;
value = "resource block (default)";
}
} else if (value == NULL) {
action->on_fail = action_fail_recover;
value = "restart (and possibly migrate) (default)";
}
pe_rsc_trace(action->rsc, "\t%s failure handling: %s", action->task, value);
value = NULL;
if (xml_obj != NULL) {
value = g_hash_table_lookup(action->meta, "role_after_failure");
if (value) {
pe_warn_once(pe_wo_role_after,
"Support for role_after_failure is deprecated and will be removed in a future release");
}
}
if (value != NULL && action->fail_role == RSC_ROLE_UNKNOWN) {
action->fail_role = text2role(value);
}
/* defaults */
if (action->fail_role == RSC_ROLE_UNKNOWN) {
if (safe_str_eq(action->task, CRMD_ACTION_PROMOTE)) {
action->fail_role = RSC_ROLE_SLAVE;
} else {
action->fail_role = RSC_ROLE_STARTED;
}
}
pe_rsc_trace(action->rsc, "\t%s failure results in: %s", action->task,
role2text(action->fail_role));
value = g_hash_table_lookup(action->meta, XML_OP_ATTR_START_DELAY);
if (value) {
unpack_start_delay(value, action->meta);
} else {
long long start_delay = 0;
value = g_hash_table_lookup(action->meta, XML_OP_ATTR_ORIGIN);
if (unpack_interval_origin(value, xml_obj, interval_ms, data_set->now,
&start_delay)) {
g_hash_table_replace(action->meta, strdup(XML_OP_ATTR_START_DELAY),
crm_strdup_printf("%lld", start_delay));
}
}
value = g_hash_table_lookup(action->meta, XML_ATTR_TIMEOUT);
timeout = unpack_timeout(value);
g_hash_table_replace(action->meta, strdup(XML_ATTR_TIMEOUT), crm_itoa(timeout));
#if ENABLE_VERSIONED_ATTRS
unpack_versioned_meta(rsc_details->versioned_meta, xml_obj, interval_ms,
data_set->now);
#endif
}
static xmlNode *
find_rsc_op_entry_helper(pe_resource_t * rsc, const char *key, gboolean include_disabled)
{
guint interval_ms = 0;
gboolean do_retry = TRUE;
char *local_key = NULL;
const char *name = NULL;
const char *value = NULL;
const char *interval_spec = NULL;
char *match_key = NULL;
xmlNode *op = NULL;
xmlNode *operation = NULL;
retry:
for (operation = __xml_first_child_element(rsc->ops_xml); operation != NULL;
operation = __xml_next_element(operation)) {
if (crm_str_eq((const char *)operation->name, "op", TRUE)) {
name = crm_element_value(operation, "name");
interval_spec = crm_element_value(operation, XML_LRM_ATTR_INTERVAL);
value = crm_element_value(operation, "enabled");
if (!include_disabled && value && crm_is_true(value) == FALSE) {
continue;
}
interval_ms = crm_parse_interval_spec(interval_spec);
match_key = pcmk__op_key(rsc->id, name, interval_ms);
if (safe_str_eq(key, match_key)) {
op = operation;
}
free(match_key);
if (rsc->clone_name) {
match_key = pcmk__op_key(rsc->clone_name, name, interval_ms);
if (safe_str_eq(key, match_key)) {
op = operation;
}
free(match_key);
}
if (op != NULL) {
free(local_key);
return op;
}
}
}
free(local_key);
if (do_retry == FALSE) {
return NULL;
}
do_retry = FALSE;
if (strstr(key, CRMD_ACTION_MIGRATE) || strstr(key, CRMD_ACTION_MIGRATED)) {
local_key = pcmk__op_key(rsc->id, "migrate", 0);
key = local_key;
goto retry;
} else if (strstr(key, "_notify_")) {
local_key = pcmk__op_key(rsc->id, "notify", 0);
key = local_key;
goto retry;
}
return NULL;
}
xmlNode *
find_rsc_op_entry(pe_resource_t * rsc, const char *key)
{
return find_rsc_op_entry_helper(rsc, key, FALSE);
}
void
print_node(const char *pre_text, pe_node_t * node, gboolean details)
{
if (node == NULL) {
crm_trace("%s%s: ", pre_text == NULL ? "" : pre_text, pre_text == NULL ? "" : ": ");
return;
}
CRM_ASSERT(node->details);
crm_trace("%s%s%sNode %s: (weight=%d, fixed=%s)",
pre_text == NULL ? "" : pre_text,
pre_text == NULL ? "" : ": ",
node->details->online ? "" : "Unavailable/Unclean ",
node->details->uname, node->weight, node->fixed ? "True" : "False");
if (details) {
int log_level = LOG_TRACE;
char *pe_mutable = strdup("\t\t");
GListPtr gIter = node->details->running_rsc;
crm_trace("\t\t===Node Attributes");
g_hash_table_foreach(node->details->attrs, print_str_str, pe_mutable);
free(pe_mutable);
crm_trace("\t\t=== Resources");
for (; gIter != NULL; gIter = gIter->next) {
pe_resource_t *rsc = (pe_resource_t *) gIter->data;
rsc->fns->print(rsc, "\t\t", pe_print_log|pe_print_pending,
&log_level);
}
}
}
/*
* Used by the HashTable for-loop
*/
void
print_str_str(gpointer key, gpointer value, gpointer user_data)
{
crm_trace("%s%s %s ==> %s",
user_data == NULL ? "" : (char *)user_data,
user_data == NULL ? "" : ": ", (char *)key, (char *)value);
}
void
pe_free_action(pe_action_t * action)
{
if (action == NULL) {
return;
}
g_list_free_full(action->actions_before, free); /* pe_action_wrapper_t* */
g_list_free_full(action->actions_after, free); /* pe_action_wrapper_t* */
if (action->extra) {
g_hash_table_destroy(action->extra);
}
if (action->meta) {
g_hash_table_destroy(action->meta);
}
#if ENABLE_VERSIONED_ATTRS
if (action->rsc) {
pe_free_rsc_action_details(action);
}
#endif
free(action->cancel_task);
free(action->reason);
free(action->task);
free(action->uuid);
free(action->node);
free(action);
}
GListPtr
find_recurring_actions(GListPtr input, pe_node_t * not_on_node)
{
const char *value = NULL;
GListPtr result = NULL;
GListPtr gIter = input;
CRM_CHECK(input != NULL, return NULL);
for (; gIter != NULL; gIter = gIter->next) {
pe_action_t *action = (pe_action_t *) gIter->data;
value = g_hash_table_lookup(action->meta, XML_LRM_ATTR_INTERVAL_MS);
if (value == NULL) {
/* skip */
} else if (safe_str_eq(value, "0")) {
/* skip */
} else if (safe_str_eq(CRMD_ACTION_CANCEL, action->task)) {
/* skip */
} else if (not_on_node == NULL) {
crm_trace("(null) Found: %s", action->uuid);
result = g_list_prepend(result, action);
} else if (action->node == NULL) {
/* skip */
} else if (action->node->details != not_on_node->details) {
crm_trace("Found: %s", action->uuid);
result = g_list_prepend(result, action);
}
}
return result;
}
enum action_tasks
get_complex_task(pe_resource_t * rsc, const char *name, gboolean allow_non_atomic)
{
enum action_tasks task = text2task(name);
if (rsc == NULL) {
return task;
} else if (allow_non_atomic == FALSE || rsc->variant == pe_native) {
switch (task) {
case stopped_rsc:
case started_rsc:
case action_demoted:
case action_promoted:
crm_trace("Folding %s back into its atomic counterpart for %s", name, rsc->id);
return task - 1;
break;
default:
break;
}
}
return task;
}
pe_action_t *
find_first_action(GListPtr input, const char *uuid, const char *task, pe_node_t * on_node)
{
GListPtr gIter = NULL;
CRM_CHECK(uuid || task, return NULL);
for (gIter = input; gIter != NULL; gIter = gIter->next) {
pe_action_t *action = (pe_action_t *) gIter->data;
if (uuid != NULL && safe_str_neq(uuid, action->uuid)) {
continue;
} else if (task != NULL && safe_str_neq(task, action->task)) {
continue;
} else if (on_node == NULL) {
return action;
} else if (action->node == NULL) {
continue;
} else if (on_node->details == action->node->details) {
return action;
}
}
return NULL;
}
GListPtr
find_actions(GListPtr input, const char *key, const pe_node_t *on_node)
{
GListPtr gIter = input;
GListPtr result = NULL;
CRM_CHECK(key != NULL, return NULL);
for (; gIter != NULL; gIter = gIter->next) {
pe_action_t *action = (pe_action_t *) gIter->data;
if (safe_str_neq(key, action->uuid)) {
crm_trace("%s does not match action %s", key, action->uuid);
continue;
} else if (on_node == NULL) {
crm_trace("Action %s matches (ignoring node)", key);
result = g_list_prepend(result, action);
} else if (action->node == NULL) {
crm_trace("Action %s matches (unallocated, assigning to %s)",
key, on_node->details->uname);
action->node = node_copy(on_node);
result = g_list_prepend(result, action);
} else if (on_node->details == action->node->details) {
crm_trace("Action %s on %s matches", key, on_node->details->uname);
result = g_list_prepend(result, action);
} else {
crm_trace("Action %s on node %s does not match requested node %s",
key, action->node->details->uname,
on_node->details->uname);
}
}
return result;
}
GList *
find_actions_exact(GList *input, const char *key, const pe_node_t *on_node)
{
GList *result = NULL;
CRM_CHECK(key != NULL, return NULL);
if (on_node == NULL) {
crm_trace("Not searching for action %s because node not specified",
key);
return NULL;
}
for (GList *gIter = input; gIter != NULL; gIter = gIter->next) {
pe_action_t *action = (pe_action_t *) gIter->data;
if (action->node == NULL) {
crm_trace("Skipping comparison of %s vs action %s without node",
key, action->uuid);
} else if (safe_str_neq(key, action->uuid)) {
crm_trace("Desired action %s doesn't match %s", key, action->uuid);
} else if (safe_str_neq(on_node->details->id,
action->node->details->id)) {
crm_trace("Action %s desired node ID %s doesn't match %s",
key, on_node->details->id, action->node->details->id);
} else {
crm_trace("Action %s matches", key);
result = g_list_prepend(result, action);
}
}
return result;
}
/*!
* \brief Find all actions of given type for a resource
*
* \param[in] rsc Resource to search
* \param[in] node Find only actions scheduled on this node
* \param[in] task Action name to search for
* \param[in] require_node If TRUE, NULL node or action node will not match
*
* \return List of actions found (or NULL if none)
* \note If node is not NULL and require_node is FALSE, matching actions
* without a node will be assigned to node.
*/
GList *
pe__resource_actions(const pe_resource_t *rsc, const pe_node_t *node,
const char *task, bool require_node)
{
GList *result = NULL;
char *key = pcmk__op_key(rsc->id, task, 0);
if (require_node) {
result = find_actions_exact(rsc->actions, key, node);
} else {
result = find_actions(rsc->actions, key, node);
}
free(key);
return result;
}
static void
resource_node_score(pe_resource_t * rsc, pe_node_t * node, int score, const char *tag)
{
pe_node_t *match = NULL;
if ((rsc->exclusive_discover || (node->rsc_discover_mode == pe_discover_never))
&& safe_str_eq(tag, "symmetric_default")) {
/* This string comparision may be fragile, but exclusive resources and
* exclusive nodes should not have the symmetric_default constraint
* applied to them.
*/
return;
} else if (rsc->children) {
GListPtr gIter = rsc->children;
for (; gIter != NULL; gIter = gIter->next) {
pe_resource_t *child_rsc = (pe_resource_t *) gIter->data;
resource_node_score(child_rsc, node, score, tag);
}
}
pe_rsc_trace(rsc, "Setting %s for %s on %s: %d", tag, rsc->id, node->details->uname, score);
match = pe_hash_table_lookup(rsc->allowed_nodes, node->details->id);
if (match == NULL) {
match = node_copy(node);
g_hash_table_insert(rsc->allowed_nodes, (gpointer) match->details->id, match);
}
match->weight = merge_weights(match->weight, score);
}
void
resource_location(pe_resource_t * rsc, pe_node_t * node, int score, const char *tag,
pe_working_set_t * data_set)
{
if (node != NULL) {
resource_node_score(rsc, node, score, tag);
} else if (data_set != NULL) {
GListPtr gIter = data_set->nodes;
for (; gIter != NULL; gIter = gIter->next) {
pe_node_t *node_iter = (pe_node_t *) gIter->data;
resource_node_score(rsc, node_iter, score, tag);
}
} else {
GHashTableIter iter;
pe_node_t *node_iter = NULL;
g_hash_table_iter_init(&iter, rsc->allowed_nodes);
while (g_hash_table_iter_next(&iter, NULL, (void **)&node_iter)) {
resource_node_score(rsc, node_iter, score, tag);
}
}
if (node == NULL && score == -INFINITY) {
if (rsc->allocated_to) {
crm_info("Deallocating %s from %s", rsc->id, rsc->allocated_to->details->uname);
free(rsc->allocated_to);
rsc->allocated_to = NULL;
}
}
}
#define sort_return(an_int, why) do { \
free(a_uuid); \
free(b_uuid); \
crm_trace("%s (%d) %c %s (%d) : %s", \
a_xml_id, a_call_id, an_int>0?'>':an_int<0?'<':'=', \
b_xml_id, b_call_id, why); \
return an_int; \
} while(0)
gint
sort_op_by_callid(gconstpointer a, gconstpointer b)
{
int a_call_id = -1;
int b_call_id = -1;
char *a_uuid = NULL;
char *b_uuid = NULL;
const xmlNode *xml_a = a;
const xmlNode *xml_b = b;
const char *a_xml_id = crm_element_value(xml_a, XML_ATTR_ID);
const char *b_xml_id = crm_element_value(xml_b, XML_ATTR_ID);
if (safe_str_eq(a_xml_id, b_xml_id)) {
/* We have duplicate lrm_rsc_op entries in the status
* section which is unlikely to be a good thing
* - we can handle it easily enough, but we need to get
* to the bottom of why it's happening.
*/
pe_err("Duplicate lrm_rsc_op entries named %s", a_xml_id);
sort_return(0, "duplicate");
}
crm_element_value_int(xml_a, XML_LRM_ATTR_CALLID, &a_call_id);
crm_element_value_int(xml_b, XML_LRM_ATTR_CALLID, &b_call_id);
if (a_call_id == -1 && b_call_id == -1) {
/* both are pending ops so it doesn't matter since
* stops are never pending
*/
sort_return(0, "pending");
} else if (a_call_id >= 0 && a_call_id < b_call_id) {
sort_return(-1, "call id");
} else if (b_call_id >= 0 && a_call_id > b_call_id) {
sort_return(1, "call id");
} else if (b_call_id >= 0 && a_call_id == b_call_id) {
/*
* The op and last_failed_op are the same
* Order on last-rc-change
*/
time_t last_a = -1;
time_t last_b = -1;
crm_element_value_epoch(xml_a, XML_RSC_OP_LAST_CHANGE, &last_a);
crm_element_value_epoch(xml_b, XML_RSC_OP_LAST_CHANGE, &last_b);
crm_trace("rc-change: %lld vs %lld",
(long long) last_a, (long long) last_b);
if (last_a >= 0 && last_a < last_b) {
sort_return(-1, "rc-change");
} else if (last_b >= 0 && last_a > last_b) {
sort_return(1, "rc-change");
}
sort_return(0, "rc-change");
} else {
/* One of the inputs is a pending operation
* Attempt to use XML_ATTR_TRANSITION_MAGIC to determine its age relative to the other
*/
int a_id = -1;
int b_id = -1;
const char *a_magic = crm_element_value(xml_a, XML_ATTR_TRANSITION_MAGIC);
const char *b_magic = crm_element_value(xml_b, XML_ATTR_TRANSITION_MAGIC);
CRM_CHECK(a_magic != NULL && b_magic != NULL, sort_return(0, "No magic"));
if (!decode_transition_magic(a_magic, &a_uuid, &a_id, NULL, NULL, NULL,
NULL)) {
sort_return(0, "bad magic a");
}
if (!decode_transition_magic(b_magic, &b_uuid, &b_id, NULL, NULL, NULL,
NULL)) {
sort_return(0, "bad magic b");
}
/* try to determine the relative age of the operation...
* some pending operations (e.g. a start) may have been superseded
* by a subsequent stop
*
* [a|b]_id == -1 means it's a shutdown operation and _always_ comes last
*/
if (safe_str_neq(a_uuid, b_uuid) || a_id == b_id) {
/*
* some of the logic in here may be redundant...
*
* if the UUID from the TE doesn't match then one better
* be a pending operation.
* pending operations don't survive between elections and joins
* because we query the LRM directly
*/
if (b_call_id == -1) {
sort_return(-1, "transition + call");
} else if (a_call_id == -1) {
sort_return(1, "transition + call");
}
} else if ((a_id >= 0 && a_id < b_id) || b_id == -1) {
sort_return(-1, "transition");
} else if ((b_id >= 0 && a_id > b_id) || a_id == -1) {
sort_return(1, "transition");
}
}
/* we should never end up here */
CRM_CHECK(FALSE, sort_return(0, "default"));
}
time_t
get_effective_time(pe_working_set_t * data_set)
{
if(data_set) {
if (data_set->now == NULL) {
crm_trace("Recording a new 'now'");
data_set->now = crm_time_new(NULL);
}
return crm_time_get_seconds_since_epoch(data_set->now);
}
crm_trace("Defaulting to 'now'");
return time(NULL);
}
gboolean
get_target_role(pe_resource_t * rsc, enum rsc_role_e * role)
{
enum rsc_role_e local_role = RSC_ROLE_UNKNOWN;
const char *value = g_hash_table_lookup(rsc->meta, XML_RSC_ATTR_TARGET_ROLE);
CRM_CHECK(role != NULL, return FALSE);
if (value == NULL || safe_str_eq("started", value)
|| safe_str_eq("default", value)) {
return FALSE;
}
local_role = text2role(value);
if (local_role == RSC_ROLE_UNKNOWN) {
pcmk__config_err("Ignoring '" XML_RSC_ATTR_TARGET_ROLE "' for %s "
"because '%s' is not valid", rsc->id, value);
return FALSE;
} else if (local_role > RSC_ROLE_STARTED) {
if (is_set(uber_parent(rsc)->flags, pe_rsc_promotable)) {
if (local_role > RSC_ROLE_SLAVE) {
/* This is what we'd do anyway, just leave the default to avoid messing up the placement algorithm */
return FALSE;
}
} else {
pcmk__config_err("Ignoring '" XML_RSC_ATTR_TARGET_ROLE "' for %s "
"because '%s' only makes sense for promotable "
"clones", rsc->id, value);
return FALSE;
}
}
*role = local_role;
return TRUE;
}
gboolean
order_actions(pe_action_t * lh_action, pe_action_t * rh_action, enum pe_ordering order)
{
GListPtr gIter = NULL;
pe_action_wrapper_t *wrapper = NULL;
GListPtr list = NULL;
if (order == pe_order_none) {
return FALSE;
}
if (lh_action == NULL || rh_action == NULL) {
return FALSE;
}
crm_trace("Ordering Action %s before %s", lh_action->uuid, rh_action->uuid);
/* Ensure we never create a dependency on ourselves... it's happened */
CRM_ASSERT(lh_action != rh_action);
/* Filter dups, otherwise update_action_states() has too much work to do */
gIter = lh_action->actions_after;
for (; gIter != NULL; gIter = gIter->next) {
pe_action_wrapper_t *after = (pe_action_wrapper_t *) gIter->data;
if (after->action == rh_action && (after->type & order)) {
return FALSE;
}
}
wrapper = calloc(1, sizeof(pe_action_wrapper_t));
wrapper->action = rh_action;
wrapper->type = order;
list = lh_action->actions_after;
list = g_list_prepend(list, wrapper);
lh_action->actions_after = list;
wrapper = NULL;
/* order |= pe_order_implies_then; */
/* order ^= pe_order_implies_then; */
wrapper = calloc(1, sizeof(pe_action_wrapper_t));
wrapper->action = lh_action;
wrapper->type = order;
list = rh_action->actions_before;
list = g_list_prepend(list, wrapper);
rh_action->actions_before = list;
return TRUE;
}
pe_action_t *
get_pseudo_op(const char *name, pe_working_set_t * data_set)
{
pe_action_t *op = NULL;
if(data_set->singletons) {
op = g_hash_table_lookup(data_set->singletons, name);
}
if (op == NULL) {
op = custom_action(NULL, strdup(name), name, NULL, TRUE, TRUE, data_set);
set_bit(op->flags, pe_action_pseudo);
set_bit(op->flags, pe_action_runnable);
}
return op;
}
void
destroy_ticket(gpointer data)
{
pe_ticket_t *ticket = data;
if (ticket->state) {
g_hash_table_destroy(ticket->state);
}
free(ticket->id);
free(ticket);
}
pe_ticket_t *
ticket_new(const char *ticket_id, pe_working_set_t * data_set)
{
pe_ticket_t *ticket = NULL;
if (ticket_id == NULL || strlen(ticket_id) == 0) {
return NULL;
}
if (data_set->tickets == NULL) {
data_set->tickets =
g_hash_table_new_full(crm_str_hash, g_str_equal, free,
destroy_ticket);
}
ticket = g_hash_table_lookup(data_set->tickets, ticket_id);
if (ticket == NULL) {
ticket = calloc(1, sizeof(pe_ticket_t));
if (ticket == NULL) {
crm_err("Cannot allocate ticket '%s'", ticket_id);
return NULL;
}
crm_trace("Creaing ticket entry for %s", ticket_id);
ticket->id = strdup(ticket_id);
ticket->granted = FALSE;
ticket->last_granted = -1;
ticket->standby = FALSE;
ticket->state = crm_str_table_new();
g_hash_table_insert(data_set->tickets, strdup(ticket->id), ticket);
}
return ticket;
}
static void
filter_parameters(xmlNode * param_set, const char *param_string, bool need_present)
{
if (param_set && param_string) {
xmlAttrPtr xIter = param_set->properties;
while (xIter) {
const char *prop_name = (const char *)xIter->name;
char *name = crm_strdup_printf(" %s ", prop_name);
char *match = strstr(param_string, name);
free(name);
// Do now, because current entry might get removed below
xIter = xIter->next;
if (need_present && match == NULL) {
crm_trace("%s not found in %s", prop_name, param_string);
xml_remove_prop(param_set, prop_name);
} else if (need_present == FALSE && match) {
crm_trace("%s found in %s", prop_name, param_string);
xml_remove_prop(param_set, prop_name);
}
}
}
}
#if ENABLE_VERSIONED_ATTRS
static void
append_versioned_params(xmlNode *versioned_params, const char *ra_version, xmlNode *params)
{
GHashTable *hash = pe_unpack_versioned_parameters(versioned_params, ra_version);
char *key = NULL;
char *value = NULL;
GHashTableIter iter;
g_hash_table_iter_init(&iter, hash);
while (g_hash_table_iter_next(&iter, (gpointer *) &key, (gpointer *) &value)) {
crm_xml_add(params, key, value);
}
g_hash_table_destroy(hash);
}
#endif
/*!
* \internal
* \brief Calculate action digests and store in node's digest cache
*
* \param[in] rsc Resource that action was for
* \param[in] task Name of action performed
* \param[in] key Action's task key
* \param[in] node Node action was performed on
* \param[in] xml_op XML of operation in CIB status (if available)
* \param[in] calc_secure Whether to calculate secure digest
* \param[in] data_set Cluster working set
*
* \return Pointer to node's digest cache entry
*/
static op_digest_cache_t *
rsc_action_digest(pe_resource_t *rsc, const char *task, const char *key,
pe_node_t *node, xmlNode *xml_op, bool calc_secure,
pe_working_set_t *data_set)
{
op_digest_cache_t *data = NULL;
data = g_hash_table_lookup(node->details->digest_cache, key);
if (data == NULL) {
GHashTable *local_rsc_params = crm_str_table_new();
pe_action_t *action = custom_action(rsc, strdup(key), task, node, TRUE, FALSE, data_set);
#if ENABLE_VERSIONED_ATTRS
xmlNode *local_versioned_params = create_xml_node(NULL, XML_TAG_RSC_VER_ATTRS);
const char *ra_version = NULL;
#endif
const char *op_version;
const char *restart_list = NULL;
const char *secure_list = " passwd password ";
data = calloc(1, sizeof(op_digest_cache_t));
CRM_ASSERT(data != NULL);
get_rsc_attributes(local_rsc_params, rsc, node, data_set);
#if ENABLE_VERSIONED_ATTRS
pe_get_versioned_attributes(local_versioned_params, rsc, node, data_set);
#endif
data->params_all = create_xml_node(NULL, XML_TAG_PARAMS);
// REMOTE_CONTAINER_HACK: Allow remote nodes that start containers with pacemaker remote inside
if (pe__add_bundle_remote_name(rsc, data->params_all,
XML_RSC_ATTR_REMOTE_RA_ADDR)) {
crm_trace("Set address for bundle connection %s (on %s)",
rsc->id, node->details->uname);
}
g_hash_table_foreach(local_rsc_params, hash2field, data->params_all);
g_hash_table_foreach(action->extra, hash2field, data->params_all);
g_hash_table_foreach(rsc->parameters, hash2field, data->params_all);
g_hash_table_foreach(action->meta, hash2metafield, data->params_all);
if(xml_op) {
secure_list = crm_element_value(xml_op, XML_LRM_ATTR_OP_SECURE);
restart_list = crm_element_value(xml_op, XML_LRM_ATTR_OP_RESTART);
op_version = crm_element_value(xml_op, XML_ATTR_CRM_VERSION);
#if ENABLE_VERSIONED_ATTRS
ra_version = crm_element_value(xml_op, XML_ATTR_RA_VERSION);
#endif
} else {
op_version = CRM_FEATURE_SET;
}
#if ENABLE_VERSIONED_ATTRS
append_versioned_params(local_versioned_params, ra_version, data->params_all);
append_versioned_params(rsc->versioned_parameters, ra_version, data->params_all);
{
pe_rsc_action_details_t *details = pe_rsc_action_details(action);
append_versioned_params(details->versioned_parameters, ra_version, data->params_all);
}
#endif
pcmk__filter_op_for_digest(data->params_all);
g_hash_table_destroy(local_rsc_params);
pe_free_action(action);
data->digest_all_calc = calculate_operation_digest(data->params_all, op_version);
if (calc_secure) {
data->params_secure = copy_xml(data->params_all);
if(secure_list) {
filter_parameters(data->params_secure, secure_list, FALSE);
}
data->digest_secure_calc = calculate_operation_digest(data->params_secure, op_version);
}
if(xml_op && crm_element_value(xml_op, XML_LRM_ATTR_RESTART_DIGEST) != NULL) {
data->params_restart = copy_xml(data->params_all);
if (restart_list) {
filter_parameters(data->params_restart, restart_list, TRUE);
}
data->digest_restart_calc = calculate_operation_digest(data->params_restart, op_version);
}
g_hash_table_insert(node->details->digest_cache, strdup(key), data);
}
return data;
}
op_digest_cache_t *
rsc_action_digest_cmp(pe_resource_t * rsc, xmlNode * xml_op, pe_node_t * node,
pe_working_set_t * data_set)
{
op_digest_cache_t *data = NULL;
char *key = NULL;
guint interval_ms = 0;
const char *op_version;
const char *task = crm_element_value(xml_op, XML_LRM_ATTR_TASK);
const char *digest_all;
const char *digest_restart;
CRM_ASSERT(node != NULL);
op_version = crm_element_value(xml_op, XML_ATTR_CRM_VERSION);
digest_all = crm_element_value(xml_op, XML_LRM_ATTR_OP_DIGEST);
digest_restart = crm_element_value(xml_op, XML_LRM_ATTR_RESTART_DIGEST);
crm_element_value_ms(xml_op, XML_LRM_ATTR_INTERVAL_MS, &interval_ms);
key = pcmk__op_key(rsc->id, task, interval_ms);
data = rsc_action_digest(rsc, task, key, node, xml_op,
is_set(data_set->flags, pe_flag_sanitized),
data_set);
data->rc = RSC_DIGEST_MATCH;
if (digest_restart && data->digest_restart_calc && strcmp(data->digest_restart_calc, digest_restart) != 0) {
pe_rsc_info(rsc, "Parameters to %s on %s changed: was %s vs. now %s (restart:%s) %s",
key, node->details->uname,
crm_str(digest_restart), data->digest_restart_calc,
op_version, crm_element_value(xml_op, XML_ATTR_TRANSITION_MAGIC));
data->rc = RSC_DIGEST_RESTART;
} else if (digest_all == NULL) {
/* it is unknown what the previous op digest was */
data->rc = RSC_DIGEST_UNKNOWN;
} else if (strcmp(digest_all, data->digest_all_calc) != 0) {
pe_rsc_info(rsc, "Parameters to %s on %s changed: was %s vs. now %s (%s:%s) %s",
key, node->details->uname,
crm_str(digest_all), data->digest_all_calc,
(interval_ms > 0)? "reschedule" : "reload",
op_version, crm_element_value(xml_op, XML_ATTR_TRANSITION_MAGIC));
data->rc = RSC_DIGEST_ALL;
}
free(key);
return data;
}
/*!
* \internal
* \brief Create an unfencing summary for use in special node attribute
*
* Create a string combining a fence device's resource ID, agent type, and
* parameter digest (whether for all parameters or just non-private parameters).
* This can be stored in a special node attribute, allowing us to detect changes
* in either the agent type or parameters, to know whether unfencing must be
* redone or can be safely skipped when the device's history is cleaned.
*
* \param[in] rsc_id Fence device resource ID
* \param[in] agent_type Fence device agent
* \param[in] param_digest Fence device parameter digest
*
* \return Newly allocated string with unfencing digest
* \note The caller is responsible for freeing the result.
*/
static inline char *
create_unfencing_summary(const char *rsc_id, const char *agent_type,
const char *param_digest)
{
return crm_strdup_printf("%s:%s:%s", rsc_id, agent_type, param_digest);
}
/*!
* \internal
* \brief Check whether a node can skip unfencing
*
* Check whether a fence device's current definition matches a node's
* stored summary of when it was last unfenced by the device.
*
* \param[in] rsc_id Fence device's resource ID
* \param[in] agent Fence device's agent type
* \param[in] digest_calc Fence device's current parameter digest
* \param[in] node_summary Value of node's special unfencing node attribute
* (a comma-separated list of unfencing summaries for
* all devices that have unfenced this node)
*
* \return TRUE if digest matches, FALSE otherwise
*/
static bool
unfencing_digest_matches(const char *rsc_id, const char *agent,
const char *digest_calc, const char *node_summary)
{
bool matches = FALSE;
if (rsc_id && agent && digest_calc && node_summary) {
char *search_secure = create_unfencing_summary(rsc_id, agent,
digest_calc);
/* The digest was calculated including the device ID and agent,
* so there is no risk of collision using strstr().
*/
matches = (strstr(node_summary, search_secure) != NULL);
crm_trace("Calculated unfencing digest '%s' %sfound in '%s'",
search_secure, matches? "" : "not ", node_summary);
free(search_secure);
}
return matches;
}
/* Magic string to use as action name for digest cache entries used for
* unfencing checks. This is not a real action name (i.e. "on"), so
* check_action_definition() won't confuse these entries with real actions.
*/
#define STONITH_DIGEST_TASK "stonith-on"
/*!
* \internal
* \brief Calculate fence device digests and digest comparison result
*
* \param[in] rsc Fence device resource
* \param[in] agent Fence device's agent type
* \param[in] node Node with digest cache to use
* \param[in] data_set Cluster working set
*
* \return Node's digest cache entry
*/
static op_digest_cache_t *
fencing_action_digest_cmp(pe_resource_t *rsc, const char *agent,
pe_node_t *node, pe_working_set_t *data_set)
{
const char *node_summary = NULL;
// Calculate device's current parameter digests
char *key = pcmk__op_key(rsc->id, STONITH_DIGEST_TASK, 0);
op_digest_cache_t *data = rsc_action_digest(rsc, STONITH_DIGEST_TASK, key,
node, NULL, TRUE, data_set);
free(key);
// Check whether node has special unfencing summary node attribute
node_summary = pe_node_attribute_raw(node, CRM_ATTR_DIGESTS_ALL);
if (node_summary == NULL) {
data->rc = RSC_DIGEST_UNKNOWN;
return data;
}
// Check whether full parameter digest matches
if (unfencing_digest_matches(rsc->id, agent, data->digest_all_calc,
node_summary)) {
data->rc = RSC_DIGEST_MATCH;
return data;
}
// Check whether secure parameter digest matches
node_summary = pe_node_attribute_raw(node, CRM_ATTR_DIGESTS_SECURE);
if (unfencing_digest_matches(rsc->id, agent, data->digest_secure_calc,
node_summary)) {
data->rc = RSC_DIGEST_MATCH;
if (is_set(data_set->flags, pe_flag_stdout)) {
printf("Only 'private' parameters to %s for unfencing %s changed\n",
rsc->id, node->details->uname);
}
return data;
}
// Parameters don't match
data->rc = RSC_DIGEST_ALL;
if (is_set(data_set->flags, (pe_flag_sanitized|pe_flag_stdout))
&& data->digest_secure_calc) {
char *digest = create_unfencing_summary(rsc->id, agent,
data->digest_secure_calc);
printf("Parameters to %s for unfencing %s changed, try '%s'\n",
rsc->id, node->details->uname, digest);
free(digest);
}
return data;
}
const char *rsc_printable_id(pe_resource_t *rsc)
{
if (is_not_set(rsc->flags, pe_rsc_unique)) {
return ID(rsc->xml);
}
return rsc->id;
}
void
clear_bit_recursive(pe_resource_t * rsc, unsigned long long flag)
{
GListPtr gIter = rsc->children;
clear_bit(rsc->flags, flag);
for (; gIter != NULL; gIter = gIter->next) {
pe_resource_t *child_rsc = (pe_resource_t *) gIter->data;
clear_bit_recursive(child_rsc, flag);
}
}
void
set_bit_recursive(pe_resource_t * rsc, unsigned long long flag)
{
GListPtr gIter = rsc->children;
set_bit(rsc->flags, flag);
for (; gIter != NULL; gIter = gIter->next) {
pe_resource_t *child_rsc = (pe_resource_t *) gIter->data;
set_bit_recursive(child_rsc, flag);
}
}
static GListPtr
find_unfencing_devices(GListPtr candidates, GListPtr matches)
{
for (GListPtr gIter = candidates; gIter != NULL; gIter = gIter->next) {
pe_resource_t *candidate = gIter->data;
const char *provides = g_hash_table_lookup(candidate->meta, XML_RSC_ATTR_PROVIDES);
const char *requires = g_hash_table_lookup(candidate->meta, XML_RSC_ATTR_REQUIRES);
if(candidate->children) {
matches = find_unfencing_devices(candidate->children, matches);
} else if (is_not_set(candidate->flags, pe_rsc_fence_device)) {
continue;
} else if (crm_str_eq(provides, "unfencing", FALSE) || crm_str_eq(requires, "unfencing", FALSE)) {
matches = g_list_prepend(matches, candidate);
}
}
return matches;
}
+static int
+node_priority_fencing_delay(pe_node_t * node, pe_working_set_t * data_set)
+{
+ int member_count = 0;
+ int online_count = 0;
+ int top_priority = 0;
+ int lowest_priority = 0;
+ GListPtr gIter = NULL;
+
+ // `priority-fencing-delay` is disabled
+ if (data_set->priority_fencing_delay < 0) {
+ return -1;
+ }
+
+ /* No need to delay fencing if the fencing target is not a normal cluster
+ * member, for example if it's a remote node or a guest node. */
+ if (node->details->type != node_member) {
+ return 0;
+ }
+
+ // No need to delay fencing if the fencing target is in our partition
+ if (node->details->online) {
+ return 0;
+ }
+
+ for (gIter = data_set->nodes; gIter != NULL; gIter = gIter->next) {
+ pe_node_t *n = gIter->data;
+
+ if (n->details->type != node_member) {
+ continue;
+ }
+
+ member_count ++;
+
+ if (n->details->online) {
+ online_count++;
+ }
+
+ if (member_count == 1
+ || n->details->priority > top_priority) {
+ top_priority = n->details->priority;
+ }
+
+ if (member_count == 1
+ || n->details->priority < lowest_priority) {
+ lowest_priority = n->details->priority;
+ }
+ }
+
+ // No need to delay if we have more than half of the cluster members
+ if (online_count > member_count / 2) {
+ return 0;
+ }
+
+ /* All the nodes have equal priority.
+ * Any configured corresponding `pcmk_delay_base/max` will be applied. */
+ if (lowest_priority == top_priority) {
+ return -1;
+ }
+
+ if (node->details->priority < top_priority) {
+ return 0;
+ }
+
+ return data_set->priority_fencing_delay;
+}
pe_action_t *
-pe_fence_op(pe_node_t * node, const char *op, bool optional, const char *reason, pe_working_set_t * data_set)
+pe_fence_op(pe_node_t * node, const char *op, bool optional, const char *reason,
+ bool priority_delay, pe_working_set_t * data_set)
{
char *op_key = NULL;
pe_action_t *stonith_op = NULL;
if(op == NULL) {
op = data_set->stonith_action;
}
op_key = crm_strdup_printf("%s-%s-%s", CRM_OP_FENCE, node->details->uname, op);
if(data_set->singletons) {
stonith_op = g_hash_table_lookup(data_set->singletons, op_key);
}
if(stonith_op == NULL) {
stonith_op = custom_action(NULL, op_key, CRM_OP_FENCE, node, TRUE, TRUE, data_set);
add_hash_param(stonith_op->meta, XML_LRM_ATTR_TARGET, node->details->uname);
add_hash_param(stonith_op->meta, XML_LRM_ATTR_TARGET_UUID, node->details->id);
add_hash_param(stonith_op->meta, "stonith_action", op);
if (pe__is_guest_or_remote_node(node)
&& is_set(data_set->flags, pe_flag_enable_unfencing)) {
/* Extra work to detect device changes on remotes
*
* We may do this for all nodes in the future, but for now
* the check_action_definition() based stuff works fine.
*/
long max = 1024;
long digests_all_offset = 0;
long digests_secure_offset = 0;
char *digests_all = calloc(max, sizeof(char));
char *digests_secure = calloc(max, sizeof(char));
GListPtr matches = find_unfencing_devices(data_set->resources, NULL);
for (GListPtr gIter = matches; gIter != NULL; gIter = gIter->next) {
pe_resource_t *match = gIter->data;
const char *agent = g_hash_table_lookup(match->meta,
XML_ATTR_TYPE);
op_digest_cache_t *data = NULL;
data = fencing_action_digest_cmp(match, agent, node, data_set);
if(data->rc == RSC_DIGEST_ALL) {
optional = FALSE;
crm_notice("Unfencing %s (remote): because the definition of %s changed", node->details->uname, match->id);
if (is_set(data_set->flags, pe_flag_stdout)) {
fprintf(stdout, " notice: Unfencing %s (remote): because the definition of %s changed\n", node->details->uname, match->id);
}
}
digests_all_offset += snprintf(
digests_all+digests_all_offset, max-digests_all_offset,
"%s:%s:%s,", match->id, agent, data->digest_all_calc);
digests_secure_offset += snprintf(
digests_secure+digests_secure_offset, max-digests_secure_offset,
"%s:%s:%s,", match->id, agent, data->digest_secure_calc);
}
g_hash_table_insert(stonith_op->meta,
strdup(XML_OP_ATTR_DIGESTS_ALL),
digests_all);
g_hash_table_insert(stonith_op->meta,
strdup(XML_OP_ATTR_DIGESTS_SECURE),
digests_secure);
}
} else {
free(op_key);
}
+ if (data_set->priority_fencing_delay >= 0
+
+ /* It's a suitable case where `priority-fencing-delay` applies.
+ * At least add `priority-fencing-delay` field as an indicator. */
+ && (priority_delay
+
+ /* Re-calculate priority delay for the suitable case when
+ * pe_fence_op() is called again by stage6() after node priority has
+ * been actually calculated with native_add_running() */
+ || g_hash_table_lookup(stonith_op->meta,
+ XML_CONFIG_ATTR_PRIORITY_FENCING_DELAY) != NULL)) {
+
+ /* Add `priority-fencing-delay` to the fencing op even if it's 0 for
+ * the targeting node. So that it takes precedence over any possible
+ * `pcmk_delay_base/max`.
+ */
+ char *delay_s = crm_itoa(node_priority_fencing_delay(node, data_set));
+
+ g_hash_table_insert(stonith_op->meta,
+ strdup(XML_CONFIG_ATTR_PRIORITY_FENCING_DELAY),
+ delay_s);
+ }
+
if(optional == FALSE && pe_can_fence(data_set, node)) {
pe_action_required(stonith_op, NULL, reason);
} else if(reason && stonith_op->reason == NULL) {
stonith_op->reason = strdup(reason);
}
return stonith_op;
}
void
trigger_unfencing(
pe_resource_t * rsc, pe_node_t *node, const char *reason, pe_action_t *dependency, pe_working_set_t * data_set)
{
if(is_not_set(data_set->flags, pe_flag_enable_unfencing)) {
/* No resources require it */
return;
} else if (rsc != NULL && is_not_set(rsc->flags, pe_rsc_fence_device)) {
/* Wasn't a stonith device */
return;
} else if(node
&& node->details->online
&& node->details->unclean == FALSE
&& node->details->shutdown == FALSE) {
- pe_action_t *unfence = pe_fence_op(node, "on", FALSE, reason, data_set);
+ pe_action_t *unfence = pe_fence_op(node, "on", FALSE, reason, FALSE, data_set);
if(dependency) {
order_actions(unfence, dependency, pe_order_optional);
}
} else if(rsc) {
GHashTableIter iter;
g_hash_table_iter_init(&iter, rsc->allowed_nodes);
while (g_hash_table_iter_next(&iter, NULL, (void **)&node)) {
if(node->details->online && node->details->unclean == FALSE && node->details->shutdown == FALSE) {
trigger_unfencing(rsc, node, reason, dependency, data_set);
}
}
}
}
gboolean
add_tag_ref(GHashTable * tags, const char * tag_name, const char * obj_ref)
{
pe_tag_t *tag = NULL;
GListPtr gIter = NULL;
gboolean is_existing = FALSE;
CRM_CHECK(tags && tag_name && obj_ref, return FALSE);
tag = g_hash_table_lookup(tags, tag_name);
if (tag == NULL) {
tag = calloc(1, sizeof(pe_tag_t));
if (tag == NULL) {
return FALSE;
}
tag->id = strdup(tag_name);
tag->refs = NULL;
g_hash_table_insert(tags, strdup(tag_name), tag);
}
for (gIter = tag->refs; gIter != NULL; gIter = gIter->next) {
const char *existing_ref = (const char *) gIter->data;
if (crm_str_eq(existing_ref, obj_ref, TRUE)){
is_existing = TRUE;
break;
}
}
if (is_existing == FALSE) {
tag->refs = g_list_append(tag->refs, strdup(obj_ref));
crm_trace("Added: tag=%s ref=%s", tag->id, obj_ref);
}
return TRUE;
}
void pe_action_set_flag_reason(const char *function, long line,
pe_action_t *action, pe_action_t *reason, const char *text,
enum pe_action_flags flags, bool overwrite)
{
bool unset = FALSE;
bool update = FALSE;
const char *change = NULL;
if(is_set(flags, pe_action_runnable)) {
unset = TRUE;
change = "unrunnable";
} else if(is_set(flags, pe_action_optional)) {
unset = TRUE;
change = "required";
} else if(is_set(flags, pe_action_migrate_runnable)) {
unset = TRUE;
overwrite = TRUE;
change = "unrunnable";
} else if(is_set(flags, pe_action_dangle)) {
change = "dangling";
} else if(is_set(flags, pe_action_requires_any)) {
change = "required";
} else {
crm_err("Unknown flag change to %x by %s: 0x%s",
flags, action->uuid, (reason? reason->uuid : "0"));
}
if(unset) {
if(is_set(action->flags, flags)) {
action->flags = crm_clear_bit(function, line, action->uuid, action->flags, flags);
update = TRUE;
}
} else {
if(is_not_set(action->flags, flags)) {
action->flags = crm_set_bit(function, line, action->uuid, action->flags, flags);
update = TRUE;
}
}
if((change && update) || text) {
char *reason_text = NULL;
if(reason == NULL) {
pe_action_set_reason(action, text, overwrite);
} else if(reason->rsc == NULL) {
reason_text = crm_strdup_printf("%s %s%c %s", change, reason->task, text?':':0, text?text:"");
} else {
reason_text = crm_strdup_printf("%s %s %s%c %s", change, reason->rsc->id, reason->task, text?':':0, text?text:"NA");
}
if(reason_text && action->rsc != reason->rsc) {
pe_action_set_reason(action, reason_text, overwrite);
}
free(reason_text);
}
}
void pe_action_set_reason(pe_action_t *action, const char *reason, bool overwrite)
{
if (action->reason != NULL && overwrite) {
pe_rsc_trace(action->rsc, "Changing %s reason from '%s' to '%s'",
action->uuid, action->reason, crm_str(reason));
free(action->reason);
} else if (action->reason == NULL) {
pe_rsc_trace(action->rsc, "Set %s reason to '%s'",
action->uuid, crm_str(reason));
} else {
// crm_assert(action->reason != NULL && !overwrite);
return;
}
if (reason != NULL) {
action->reason = strdup(reason);
} else {
action->reason = NULL;
}
}
/*!
* \internal
* \brief Check whether shutdown has been requested for a node
*
* \param[in] node Node to check
*
* \return TRUE if node has shutdown attribute set and nonzero, FALSE otherwise
* \note This differs from simply using node->details->shutdown in that it can
* be used before that has been determined (and in fact to determine it),
* and it can also be used to distinguish requested shutdown from implicit
* shutdown of remote nodes by virtue of their connection stopping.
*/
bool
pe__shutdown_requested(pe_node_t *node)
{
const char *shutdown = pe_node_attribute_raw(node, XML_CIB_ATTR_SHUTDOWN);
return shutdown && strcmp(shutdown, "0");
}
/*!
* \internal
* \brief Update a data set's "recheck by" time
*
* \param[in] recheck Epoch time when recheck should happen
* \param[in,out] data_set Current working set
*/
void
pe__update_recheck_time(time_t recheck, pe_working_set_t *data_set)
{
if ((recheck > get_effective_time(data_set))
&& ((data_set->recheck_by == 0)
|| (data_set->recheck_by > recheck))) {
data_set->recheck_by = recheck;
}
}
/*!
* \internal
* \brief Wrapper for pe_unpack_nvpairs() using a cluster working set
*/
void
pe__unpack_dataset_nvpairs(xmlNode *xml_obj, const char *set_name,
GHashTable *node_hash, GHashTable *hash,
const char *always_first, gboolean overwrite,
pe_working_set_t *data_set)
{
crm_time_t *next_change = crm_time_new_undefined();
pe_unpack_nvpairs(data_set->input, xml_obj, set_name, node_hash, hash,
always_first, overwrite, data_set->now, next_change);
if (crm_time_is_defined(next_change)) {
time_t recheck = (time_t) crm_time_get_seconds_since_epoch(next_change);
pe__update_recheck_time(recheck, data_set);
}
crm_time_free(next_change);
}
bool
pe__resource_is_disabled(pe_resource_t *rsc)
{
const char *target_role = NULL;
CRM_CHECK(rsc != NULL, return false);
target_role = g_hash_table_lookup(rsc->meta, XML_RSC_ATTR_TARGET_ROLE);
if (target_role) {
enum rsc_role_e target_role_e = text2role(target_role);
if ((target_role_e == RSC_ROLE_STOPPED)
|| ((target_role_e == RSC_ROLE_SLAVE)
&& is_set(uber_parent(rsc)->flags, pe_rsc_promotable))) {
return true;
}
}
return false;
}
/*!
* \internal
* \brief Create an action to clear a resource's history from CIB
*
* \param[in] rsc Resource to clear
* \param[in] node Node to clear history on
*
* \return New action to clear resource history
*/
pe_action_t *
pe__clear_resource_history(pe_resource_t *rsc, pe_node_t *node,
pe_working_set_t *data_set)
{
char *key = NULL;
CRM_ASSERT(rsc && node);
key = pcmk__op_key(rsc->id, CRM_OP_LRM_DELETE, 0);
return custom_action(rsc, key, CRM_OP_LRM_DELETE, node, FALSE, TRUE,
data_set);
}
diff --git a/tools/stonith_admin.c b/tools/stonith_admin.c
index 87b538ac94..f98e335a17 100644
--- a/tools/stonith_admin.c
+++ b/tools/stonith_admin.c
@@ -1,619 +1,625 @@
/*
* Copyright 2009-2020 the Pacemaker project contributors
*
* The version control history for this file may have further details.
*
* This source code is licensed under the GNU General Public License version 2
* or later (GPLv2+) WITHOUT ANY WARRANTY.
*/
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#include
#define SUMMARY "stonith_admin - Access the Pacemaker fencing API"
char action = 0;
struct {
gboolean as_nodeid;
gboolean broadcast;
gboolean cleanup;
gboolean installed;
gboolean metadata;
gboolean registered;
gboolean validate_cfg;
stonith_key_value_t *devices;
stonith_key_value_t *params;
int fence_level;
int timeout ;
int tolerance;
+ int delay;
char *agent;
char *confirm_host;
char *fence_host;
char *history;
char *last_fenced;
char *query;
char *reboot_host;
char *register_dev;
char *register_level;
char *targets;
char *terminate;
char *unfence_host;
char *unregister_dev;
char *unregister_level;
} options = {
- .timeout = 120
+ .timeout = 120,
+ .delay = -1
};
gboolean add_env_params(const gchar *option_name, const gchar *optarg, gpointer data, GError **error);
gboolean add_stonith_device(const gchar *option_name, const gchar *optarg, gpointer data, GError **error);
gboolean add_stonith_params(const gchar *option_name, const gchar *optarg, gpointer data, GError **error);
gboolean add_tolerance(const gchar *option_name, const gchar *optarg, gpointer data, GError **error);
gboolean set_tag(const gchar *option_name, const gchar *optarg, gpointer data, GError **error);
#define INDENT " "
/* *INDENT-OFF* */
static GOptionEntry defn_entries[] = {
{ "register", 'R', 0, G_OPTION_ARG_STRING, &options.register_dev,
"Register the named stonith device. Requires: --agent.\n"
INDENT "Optional: --option, --env-option.",
"DEVICE" },
{ "deregister", 'D', 0, G_OPTION_ARG_STRING, &options.unregister_dev,
"De-register the named stonith device.",
"DEVICE" },
{ "register-level", 'r', 0, G_OPTION_ARG_STRING, &options.register_level,
"Register a stonith level for the named target,\n"
INDENT "specified as one of NAME, @PATTERN, or ATTR=VALUE.\n"
INDENT "Requires: --index and one or more --device entries.",
"TARGET" },
{ "deregister-level", 'd', 0, G_OPTION_ARG_STRING, &options.unregister_level,
"Unregister a stonith level for the named target,\n"
INDENT "specified as for --register-level. Requires: --index",
"TARGET" },
{ NULL }
};
static GOptionEntry query_entries[] = {
{ "list", 'l', 0, G_OPTION_ARG_STRING, &options.terminate,
"List devices that can terminate the specified host.\n"
INDENT "Optional: --timeout",
"HOST" },
{ "list-registered", 'L', 0, G_OPTION_ARG_NONE, &options.registered,
"List all registered devices. Optional: --timeout.",
NULL },
{ "list-installed", 'I', 0, G_OPTION_ARG_NONE, &options.installed,
"List all installed devices. Optional: --timeout.",
NULL },
{ "list-targets", 's', 0, G_OPTION_ARG_STRING, &options.targets,
"List the targets that can be fenced by the\n"
INDENT "named device. Optional: --timeout.",
"DEVICE" },
{ "metadata", 'M', 0, G_OPTION_ARG_NONE, &options.metadata,
"Show agent metadata. Requires: --agent.\n"
INDENT "Optional: --timeout.",
NULL },
{ "query", 'Q', 0, G_OPTION_ARG_STRING, &options.query,
"Check the named device's status. Optional: --timeout.",
"DEVICE" },
{ "history", 'H', 0, G_OPTION_ARG_STRING, &options.history,
"Show last successful fencing operation for named node\n"
INDENT "(or '*' for all nodes). Optional: --timeout, --cleanup,\n"
INDENT "--quiet (show only the operation's epoch timestamp),\n"
INDENT "--verbose (show all recorded and pending operations),\n"
INDENT "--broadcast (update history from all nodes available).",
"NODE" },
{ "last", 'h', 0, G_OPTION_ARG_STRING, &options.last_fenced,
"Indicate when the named node was last fenced.\n"
INDENT "Optional: --as-node-id.",
"NODE" },
{ "validate", 'K', 0, G_OPTION_ARG_NONE, &options.validate_cfg,
"Validate a fence device configuration.\n"
INDENT "Requires: --agent. Optional: --option, --env-option,\n"
INDENT "--quiet (print no output, only return status).",
NULL },
{ NULL }
};
static GOptionEntry fence_entries[] = {
{ "fence", 'F', 0, G_OPTION_ARG_STRING, &options.fence_host,
"Fence named host. Optional: --timeout, --tolerance.",
"HOST" },
{ "unfence", 'U', 0, G_OPTION_ARG_STRING, &options.unfence_host,
"Unfence named host. Optional: --timeout, --tolerance.",
"HOST" },
{ "reboot", 'B', 0, G_OPTION_ARG_STRING, &options.reboot_host,
"Reboot named host. Optional: --timeout, --tolerance.",
"HOST" },
{ "confirm", 'C', 0, G_OPTION_ARG_STRING, &options.confirm_host,
"Tell cluster that named host is now safely down.",
"HOST", },
{ NULL }
};
static GOptionEntry addl_entries[] = {
{ "cleanup", 'c', 0, G_OPTION_ARG_NONE, &options.cleanup,
"Cleanup wherever appropriate. Requires --history.",
NULL },
{ "broadcast", 'b', 0, G_OPTION_ARG_NONE, &options.broadcast,
"Broadcast wherever appropriate.",
NULL },
{ "agent", 'a', 0, G_OPTION_ARG_STRING, &options.agent,
"The agent to use (for example, fence_xvm;\n"
INDENT "with --register, --metadata, --validate).",
"AGENT" },
{ "option", 'o', 0, G_OPTION_ARG_CALLBACK, add_stonith_params,
"Specify a device configuration parameter as NAME=VALUE\n"
INDENT "(may be specified multiple times; with --register,\n"
INDENT "--validate).",
"PARAM" },
{ "env-option", 'e', 0, G_OPTION_ARG_CALLBACK, add_env_params,
"Specify a device configuration parameter with the\n"
INDENT "specified name, using the value of the\n"
INDENT "environment variable of the same name prefixed with\n"
INDENT "OCF_RESKEY_ (may be specified multiple times;\n"
INDENT "with --register, --validate).",
"PARAM" },
{ "tag", 'T', 0, G_OPTION_ARG_CALLBACK, set_tag,
"Identify fencing operations in logs with the specified\n"
INDENT "tag; useful when multiple entities might invoke\n"
INDENT "stonith_admin (used with most commands).",
"TAG" },
{ "device", 'v', 0, G_OPTION_ARG_CALLBACK, add_stonith_device,
"Device ID (with --register-level, device to associate with\n"
INDENT "a given host and level; may be specified multiple times)"
#if SUPPORT_CIBSECRETS
"\n" INDENT "(with --validate, name to use to load CIB secrets)"
#endif
".",
"DEVICE" },
{ "index", 'i', 0, G_OPTION_ARG_INT, &options.fence_level,
"The stonith level (1-9) (with --register-level,\n"
INDENT "--deregister-level).",
"LEVEL" },
{ "timeout", 't', 0, G_OPTION_ARG_INT, &options.timeout,
"Operation timeout in seconds (default 120;\n"
INDENT "used with most commands).",
"SECONDS" },
+ { "delay", 'y', 0, G_OPTION_ARG_INT, &options.delay,
+ "Enforced fencing delay in seconds (default -1 (disabled);\n"
+ INDENT "with --fence, --reboot, --unfence).",
+ "SECONDS" },
{ "as-node-id", 'n', 0, G_OPTION_ARG_NONE, &options.as_nodeid,
"(Advanced) The supplied node is the corosync node ID\n"
INDENT "(with --last).",
NULL },
{ "tolerance", 0, 0, G_OPTION_ARG_CALLBACK, add_tolerance,
"(Advanced) Do nothing if an equivalent --fence request\n"
INDENT "succeeded less than this many seconds earlier\n"
INDENT "(with --fence, --unfence, --reboot).",
"SECONDS" },
{ NULL }
};
/* *INDENT-ON* */
static pcmk__supported_format_t formats[] = {
PCMK__SUPPORTED_FORMAT_HTML,
PCMK__SUPPORTED_FORMAT_TEXT,
PCMK__SUPPORTED_FORMAT_XML,
{ NULL, NULL, NULL }
};
static int st_opts = st_opt_sync_call | st_opt_allow_suicide;
static char *name = NULL;
gboolean
add_env_params(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) {
char *key = crm_strdup_printf("OCF_RESKEY_%s", optarg);
const char *env = getenv(key);
gboolean retval = TRUE;
if (env == NULL) {
crm_err("Invalid option: -e %s", optarg);
g_set_error(error, G_OPTION_ERROR, CRM_EX_INVALID_PARAM, "Invalid option: -e %s", optarg);
retval = FALSE;
} else {
crm_info("Got: '%s'='%s'", optarg, env);
options.params = stonith_key_value_add(options.params, optarg, env);
}
free(key);
return retval;
}
gboolean
add_stonith_device(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) {
options.devices = stonith_key_value_add(options.devices, NULL, optarg);
return TRUE;
}
gboolean
add_tolerance(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) {
options.tolerance = crm_get_msec(optarg) / 1000;
return TRUE;
}
gboolean
add_stonith_params(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) {
char *name = NULL;
char *value = NULL;
int rc = 0;
gboolean retval = TRUE;
crm_info("Scanning: -o %s", optarg);
rc = pcmk_scan_nvpair(optarg, &name, &value);
if (rc != 2) {
crm_err("Invalid option: -o %s: %s", optarg, pcmk_strerror(rc));
g_set_error(error, G_OPTION_ERROR, rc, "Invalid option: -o %s: %s", optarg, pcmk_strerror(rc));
retval = FALSE;
} else {
crm_info("Got: '%s'='%s'", name, value);
options.params = stonith_key_value_add(options.params, name, value);
}
free(name);
free(value);
return retval;
}
gboolean
set_tag(const gchar *option_name, const gchar *optarg, gpointer data, GError **error) {
free(name);
name = crm_strdup_printf("%s.%s", crm_system_name, optarg);
return TRUE;
}
static GOptionContext *
build_arg_context(pcmk__common_args_t *args, GOptionGroup **group) {
GOptionContext *context = NULL;
GOptionEntry extra_prog_entries[] = {
{ "quiet", 'q', 0, G_OPTION_ARG_NONE, &(args->quiet),
"Be less descriptive in output.",
NULL },
{ NULL }
};
context = pcmk__build_arg_context(args, "text (default), html, xml", group);
/* Add the -q option, which cannot be part of the globally supported options
* because some tools use that flag for something else.
*/
pcmk__add_main_args(context, extra_prog_entries);
pcmk__add_arg_group(context, "definition", "Device Definition Commands:",
"Show device definition help", defn_entries);
pcmk__add_arg_group(context, "queries", "Queries:",
"Show query help", query_entries);
pcmk__add_arg_group(context, "fence", "Fencing Commands:",
"Show fence help", fence_entries);
pcmk__add_arg_group(context, "additional", "Additional Options:",
"Show additional options", addl_entries);
return context;
}
int
main(int argc, char **argv)
{
int rc = 0;
bool no_connect = false;
bool required_agent = false;
char *target = NULL;
const char *device = NULL;
crm_exit_t exit_code = CRM_EX_OK;
stonith_t *st = NULL;
pcmk__output_t *out = NULL;
pcmk__common_args_t *args = pcmk__new_common_args(SUMMARY);
GError *error = NULL;
GOptionContext *context = NULL;
GOptionGroup *output_group = NULL;
gchar **processed_args = NULL;
context = build_arg_context(args, &output_group);
pcmk__register_formats(output_group, formats);
crm_log_cli_init("stonith_admin");
name = strdup(crm_system_name);
processed_args = pcmk__cmdline_preproc(argv, "adehilorstvBCDFHQRTU");
if (!g_option_context_parse_strv(context, &processed_args, &error)) {
fprintf(stderr, "%s: %s\n", g_get_prgname(), error->message);
exit_code = CRM_EX_USAGE;
goto done;
}
for (int i = 0; i < args->verbosity; i++) {
crm_bump_log_level(argc, argv);
}
rc = pcmk__output_new(&out, args->output_ty, args->output_dest, argv);
if (rc != pcmk_rc_ok) {
fprintf(stderr, "Error creating output format %s: %s\n",
args->output_ty, pcmk_rc_str(rc));
exit_code = CRM_EX_ERROR;
goto done;
}
stonith__register_messages(out);
if (args->version) {
out->version(out, false);
goto done;
}
if (options.validate_cfg) {
required_agent = true;
no_connect = true;
action = 'K';
}
if (options.installed) {
no_connect = true;
action = 'I';
}
if (options.registered) {
action = 'L';
}
if (options.register_dev != NULL) {
required_agent = true;
action = 'R';
device = options.register_dev;
}
if (options.query != NULL) {
action = 'Q';
device = options.query;
}
if (options.unregister_dev != NULL) {
action = 'D';
device = options.unregister_dev;
}
if (options.targets != NULL) {
action = 's';
device = options.targets;
}
if (options.terminate != NULL) {
action = 'L';
target = options.terminate;
}
if (options.metadata) {
no_connect = true;
required_agent = true;
action = 'M';
}
if (options.reboot_host != NULL) {
no_connect = true;
action = 'B';
target = options.reboot_host;
crm_log_args(argc, argv);
}
if (options.fence_host != NULL) {
no_connect = true;
action = 'F';
target = options.fence_host;
crm_log_args(argc, argv);
}
if (options.unfence_host != NULL) {
no_connect = true;
action = 'U';
target = options.unfence_host;
crm_log_args(argc, argv);
}
if (options.confirm_host != NULL) {
action = 'C';
target = options.confirm_host;
crm_log_args(argc, argv);
}
if (options.last_fenced != NULL) {
action = 'h';
target = options.last_fenced;
}
if (options.history != NULL) {
action = 'H';
target = options.history;
}
if (options.register_level != NULL) {
action = 'r';
target = options.register_level;
}
if (options.unregister_level != NULL) {
action = 'd';
target = options.unregister_level;
}
if (optind > argc || action == 0) {
char *help = g_option_context_get_help(context, TRUE, NULL);
out->err(out, "%s", help);
free(help);
exit_code = CRM_EX_USAGE;
goto done;
}
if (required_agent && options.agent == NULL) {
char *help = g_option_context_get_help(context, TRUE, NULL);
out->err(out, "Please specify an agent to query using -a,--agent [value]");
out->err(out, "%s", help);
free(help);
exit_code = CRM_EX_USAGE;
goto done;
}
st = stonith_api_new();
if (st == NULL) {
rc = -ENOMEM;
} else if (!no_connect) {
rc = st->cmds->connect(st, name, NULL);
}
if (rc < 0) {
out->err(out, "Could not connect to fencer: %s", pcmk_strerror(rc));
exit_code = CRM_EX_DISCONNECT;
goto done;
}
switch (action) {
case 'I':
rc = pcmk__fence_installed(out, st, options.timeout*1000);
if (rc != pcmk_rc_ok) {
out->err(out, "Failed to list installed devices: %s", pcmk_strerror(rc));
}
break;
case 'L':
rc = pcmk__fence_registered(out, st, target, options.timeout*1000);
if (rc != pcmk_rc_ok) {
out->err(out, "Failed to list registered devices: %s", pcmk_strerror(rc));
}
break;
case 'Q':
rc = st->cmds->monitor(st, st_opts, device, options.timeout);
if (rc != pcmk_rc_ok) {
rc = st->cmds->list(st, st_opts, device, NULL, options.timeout);
}
break;
case 's':
rc = pcmk__fence_list_targets(out, st, target, options.timeout*1000);
if (rc != pcmk_rc_ok) {
out->err(out, "Couldn't list targets: %s", pcmk_strerror(rc));
}
break;
case 'R':
rc = st->cmds->register_device(st, st_opts, device, NULL, options.agent,
options.params);
break;
case 'D':
rc = st->cmds->remove_device(st, st_opts, device);
break;
case 'd':
rc = pcmk__fence_unregister_level(st, target, options.fence_level);
break;
case 'r':
rc = pcmk__fence_register_level(st, target, options.fence_level, options.devices);
break;
case 'M':
rc = pcmk__fence_metadata(out, st, options.agent, options.timeout*1000);
if (rc != pcmk_rc_ok) {
out->err(out, "Can't get fence agent meta-data: %s", pcmk_strerror(rc));
}
break;
case 'C':
rc = st->cmds->confirm(st, st_opts, target);
break;
case 'B':
rc = pcmk__fence_action(st, target, "reboot", name, options.timeout*1000,
- options.tolerance*1000);
+ options.tolerance*1000, options.delay);
break;
case 'F':
rc = pcmk__fence_action(st, target, "off", name, options.timeout*1000,
- options.tolerance*1000);
+ options.tolerance*1000, options.delay);
break;
case 'U':
rc = pcmk__fence_action(st, target, "on", name, options.timeout*1000,
- options.tolerance*1000);
+ options.tolerance*1000, options.delay);
break;
case 'h':
rc = pcmk__fence_last(out, target, options.as_nodeid);
break;
case 'H':
rc = pcmk__fence_history(out, st, target, options.timeout*1000, args->quiet,
args->verbosity, options.broadcast, options.cleanup);
break;
case 'K':
device = options.devices ? options.devices->key : NULL;
rc = pcmk__fence_validate(out, st, options.agent, device, options.params,
options.timeout*1000);
break;
}
crm_info("Command returned: %s (%d)", pcmk_strerror(rc), rc);
exit_code = crm_errno2exit(rc);
done:
g_strfreev(processed_args);
g_clear_error(&error);
pcmk__free_arg_context(context);
if (out != NULL) {
out->finish(out, exit_code, true, NULL);
pcmk__output_free(out);
}
free(name);
stonith_key_value_freeall(options.params, 1, 1);
if (st != NULL) {
st->cmds->disconnect(st);
stonith_api_delete(st);
}
return exit_code;
}