diff --git a/cts/cts-fencing.in b/cts/cts-fencing.in index 2885d1982f..65a9916629 100644 --- a/cts/cts-fencing.in +++ b/cts/cts-fencing.in @@ -1,1527 +1,1527 @@ #!@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 2012-2018 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 for node 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 of node3 by") test.add_stonith_log_pattern("Operation off of node3 by") ### 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 for node node3 originating from client") - test.add_stonith_log_pattern("Merging stonith action off for node node3 originating from client") - test.add_stonith_log_pattern("Merging stonith action off for node node3 originating from client") - test.add_stonith_log_pattern("Merging stonith action off for node 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") + 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 of node3 by") test.add_stonith_log_pattern("Operation off of node3 by") test.add_stonith_log_pattern("Operation off of node3 by") test.add_stonith_log_pattern("Operation off of node3 by") test.add_stonith_log_pattern("Operation off of node3 by") ### 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 for node node3 originating from client") - test.add_stonith_log_pattern("Merging stonith action off for node node3 originating from client") - test.add_stonith_log_pattern("Merging stonith action off for node node3 originating from client") - test.add_stonith_log_pattern("Merging stonith action off for node 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") + 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 of node3 by") test.add_stonith_log_pattern("Operation off of node3 by") test.add_stonith_log_pattern("Operation off of node3 by") test.add_stonith_log_pattern("Operation off of node3 by") test.add_stonith_log_pattern("Operation off of node3 by") 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 for node node3 originating from client") + 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)") 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 of 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 of node_fake") test.add_stonith_log_pattern("perform op 'node_fake off' with 'true1'") test.add_stonith_log_pattern("perform op 'node_fake off' with 'true2'") test.add_stonith_log_pattern("Remapped off of 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) for node_fake") test.add_stonith_log_pattern("Ignoring true2 'on' failure (no capable peers) for node_fake") test.add_stonith_log_pattern("Undoing remap of reboot of 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 of node_fake") test.add_stonith_log_pattern("perform op 'node_fake off' with 'true1'") test.add_stonith_log_pattern("perform op 'node_fake off' with 'true2'") test.add_stonith_log_pattern("Remapped off of node_fake complete, remapping to on") test.add_stonith_log_pattern("Undoing remap of reboot of node_fake") test.add_stonith_neg_log_pattern("perform op 'node_fake on' with") 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 op 'node_fake reboot' with 'false1'") test.add_stonith_log_pattern("Remapping multiple-device reboot of node_fake") test.add_stonith_log_pattern("perform op 'node_fake off' with 'true1'") test.add_stonith_log_pattern("perform op 'node_fake off' with 'true2'") test.add_stonith_log_pattern("Remapped off of node_fake complete, remapping to on") test.add_stonith_log_pattern("Ignoring true1 'on' failure (no capable peers) for node_fake") test.add_stonith_log_pattern("Ignoring true2 'on' failure (no capable peers) for node_fake") test.add_stonith_log_pattern("Undoing remap of reboot of 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 op 'node_fake reboot' with 'false1'") test.add_stonith_log_pattern("Remapping multiple-device reboot of node_fake") test.add_stonith_log_pattern("perform op 'node_fake off' with 'true1'") test.add_stonith_log_pattern("perform op 'node_fake off' with '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 of node_fake") test.add_stonith_log_pattern("perform op 'node_fake reboot' with '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/daemons/fenced/fenced_commands.c b/daemons/fenced/fenced_commands.c index bbf37754aa..9d3f9246bb 100644 --- a/daemons/fenced/fenced_commands.c +++ b/daemons/fenced/fenced_commands.c @@ -1,2729 +1,2729 @@ /* * Copyright 2009-2019 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 #if SUPPORT_CIBSECRETS # include #endif #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 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; 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); } return delay_max_ms; } static int get_action_delay_base(stonith_device_t * device, const char * action) { const char *value = NULL; int delay_base_ms = 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); } return delay_base_ms; } /*! * \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; 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 ? " for node " : "", cmd->victim ? cmd->victim : "", + 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 %dms", 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 (replace_secret_params(device->id, device->params) < 0) { /* 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 ? " for node " : "", cmd->victim ? cmd->victim : "", + 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; 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%s%s on %s for remote peer %s with op id (%s) (timeout=%ds)", + 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%s%s on %s for %s (timeout=%ds)", + 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; } 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%s%s on %s for %dms (timeout=%ds, base=%dms, " + 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); cmd->delay_id = g_timeout_add(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); + 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", + 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", + 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); + 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_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", + 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", + 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", + 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", + 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 " + 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", + 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_TRACE); 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) { 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: %d (%s)%s%s", cmd->action, pid, cmd->id, cmd->client_name, cmd->victim, cmd->device, 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: %d (%s)%s%s", cmd->action, pid, cmd->device, 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) { 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); + 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_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); crm_log_xml_trace(reply, "Reply"); 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 on %s", cmd->action, device->id); + 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); schedule_stonith_command(cmd, next_device); /* Prevent cmd from being freed */ cmd = NULL; goto done; } stonith_send_async_reply(cmd, output, rc, pid); 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 for node %s originating from client %s with identical stonith request from client %s", + ("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); 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 (g_list_length(devices) > 0) { /* 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); 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(crm_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); crm_ipcs_send(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) { crm_ipcs_send_ack(client, id, flags, "ack", __FUNCTION__, __LINE__); } 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 %s from %s", op, client ? client->name : remote_peer); crm_log_xml_warn(request, "UnknownOp"); } 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(crm_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(crm_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_TRACE)) { 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/lib/fencing/st_client.c b/lib/fencing/st_client.c index b7c4443b94..3d0f805b27 100644 --- a/lib/fencing/st_client.c +++ b/lib/fencing/st_client.c @@ -1,2597 +1,2597 @@ /* * 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. */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #if SUPPORT_CIBSECRETS # include #endif 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 for node '%s' as '%s=%s'", action, victim, param, + 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) { 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); rc = stonith_send_command(stonith, STONITH_OP_FENCE, data, NULL, call_options, timeout); free_xml(data); return rc; } 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_TRACE); 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); } } /*! * \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)); } 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"; static 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); { 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 (!crm_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 = replace_secret_params(rsc_id, params_table); if (rc < 0) { crm_warn("Could not replace secret parameters for validation of %s: %s", agent, pcmk_strerror(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->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; }