diff --git a/cts/cts-fencing.in b/cts/cts-fencing.in index babfb63514..5cd9f7b8fe 100644 --- a/cts/cts-fencing.in +++ b/cts/cts-fencing.in @@ -1,1624 +1,1624 @@ #!@PYTHON@ """ Regression tests for Pacemaker's fencer """ __copyright__ = "Copyright 2012-2021 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@" COROSYNC_CONF = "@PCMK__COROSYNC_CONF@" 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/support:%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): if rng_file == None: return ["xmllint", "-"] else: 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, check_stderr=False, stdin=None): """ Get text output from pipes """ if stdin is not None: pipe_outputs = pipes.communicate(input=stdin.encode()) else: pipe_outputs = pipes.communicate() output = pipe_outputs[0].decode(sys.stdout.encoding) if check_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, timeout=2, force_wait=0, logdir="/tmp"): self.name = name self.description = description self.cmds = [] self.verbose = verbose self.timeout = timeout self.force_wait = force_wait self.logpath = os.path.join(logdir, "pacemaker-fenced.log") 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, check_rng=True, check_stderr=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, "check_rng" : check_rng, "check_stderr" : check_stderr, } ) 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(self.logpath): os.remove(self.logpath) cmd = "pacemaker-fenced %s -l %s" % (self.stonith_options, self.logpath) self.stonith_process = subprocess.Popen(shlex.split(cmd)) logfile = None init_time = time.time() update_time = init_time while True: time.sleep(0.1) if self.force_wait == 0 and logfile == None \ and os.path.exists(self.logpath): logfile = io.open(self.logpath, 'rt', encoding = "ISO-8859-1") if self.force_wait == 0 and logfile != None: for line in logfile.readlines(): if "successfully started" in line: return now = time.time() if self.timeout > 0 and (now - init_time) >= self.timeout: if self.force_wait == 0: print("\tDaemon pacemaker-fenced doesn't seem to have been initialized within %fs." "\n\tConsider specifying a longer '--timeout' value." %(self.timeout)) return if self.verbose and (now - update_time) >= 5: print("Waiting for pacemaker-fenced to be initialized: %fs ..." %(now - init_time)) update_time = now 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(self.logpath, '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") 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, check_rng=True, check_stderr=True): """ Add a simple command to be executed as part of this test """ self.__new_cmd(cmd, args, CrmExit.OK, "", validate=validate, check_rng=check_rng, check_stderr=check_stderr) 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, check_stderr=args['check_stderr']) 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']: if args['check_rng']: rng_file = rng_directory() + "/api/api-result.rng" else: rng_file = None 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, check_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 except XmlValidationError as e: print("Step %d FAILED - xmllint failed: %s" % (i, 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, timeout=2, force_wait=0, logdir="/tmp"): self.tests = [] self.verbose = verbose self.timeout = timeout self.force_wait = force_wait self.logdir = logdir self.autogen_corosync_cfg = not os.path.exists(COROSYNC_CONF) def new_test(self, name, description, with_cpg=0): """ Create a named test """ test = Test(name, description, self.verbose, with_cpg, self.timeout, self.force_wait, self.logdir) 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 fencing action 'off' targeting node3 originating from client") ### the pattern below signifies that both the original and duplicate operation completed test.add_stonith_log_pattern("Operation 'off' targeting node3 by ") test.add_stonith_log_pattern("Operation 'off' targeting 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 fencing action 'off' targeting node3 originating from client") test.add_stonith_log_pattern("Merging fencing action 'off' targeting node3 originating from client") test.add_stonith_log_pattern("Merging fencing action 'off' targeting node3 originating from client") test.add_stonith_log_pattern("Merging fencing action 'off' targeting node3 originating from client") ### the pattern below signifies that both the original and duplicate operation completed test.add_stonith_log_pattern("Operation 'off' targeting node3 by ") test.add_stonith_log_pattern("Operation 'off' targeting node3 by ") test.add_stonith_log_pattern("Operation 'off' targeting node3 by ") test.add_stonith_log_pattern("Operation 'off' targeting node3 by ") test.add_stonith_log_pattern("Operation 'off' targeting 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 fencing action 'off' targeting node3 originating from client") test.add_stonith_log_pattern("Merging fencing action 'off' targeting node3 originating from client") test.add_stonith_log_pattern("Merging fencing action 'off' targeting node3 originating from client") test.add_stonith_log_pattern("Merging fencing action 'off' targeting node3 originating from client") ### the pattern below signifies that both the original and duplicate operation completed test.add_stonith_log_pattern("Operation 'off' targeting node3 by ") test.add_stonith_log_pattern("Operation 'off' targeting node3 by ") test.add_stonith_log_pattern("Operation 'off' targeting node3 by ") test.add_stonith_log_pattern("Operation 'off' targeting node3 by ") test.add_stonith_log_pattern("Operation 'off' targeting 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 fencing 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("targeting node3 using false1 returned ") test.add_stonith_log_pattern("targeting node3 using false2 returned ") test.add_stonith_log_pattern("targeting node3 using 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("targeting node3 using 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("targeting node3 using 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("targeting node3 using false returned -201") + test.add_stonith_log_pattern("targeting node3 using false returned 1") test.add_stonith_log_pattern("targeting node3 using 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("targeting node3 using false1 returned -201") - test.add_stonith_log_pattern("targeting node3 using false2 returned -201") + test.add_stonith_log_pattern("targeting node3 using false1 returned 1") + test.add_stonith_log_pattern("targeting node3 using false2 returned 1") test.add_stonith_log_pattern("targeting node3 using true3 returned 0") test.add_stonith_log_pattern("targeting node3 using 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("targeting node3 using false1 returned -201") + test.add_stonith_log_pattern("targeting node3 using false1 returned 1") test.add_stonith_neg_log_pattern("targeting node3 using false2 returned ") test.add_stonith_log_pattern("targeting node3 using true3 returned 0") test.add_stonith_log_pattern("targeting node3 using 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("targeting node3 using 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("targeting node2 using true1 returned 0") test.add_stonith_log_pattern("targeting pcmk3 using 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", 'action="off" target="node3" .* status="success"') # 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 support reboot") - test.add_stonith_log_pattern("using true1 returned 0 (OK)") + test.add_stonith_log_pattern("using true1 returned 0") # 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("using true1 returned 0 (OK)") + test.add_stonith_log_pattern("using true1 returned 0") # make sure requested fencing delay is applied only for the first device in the first level # make sure static delay from pcmk_delay_base is added for test_type in test_types: if test_type["use_cpg"] == 0: continue test = self.new_test("%s_topology_delay" % test_type["prefix"], "Verify requested fencing delay is applied only for the first device in the first level and pcmk_delay_base is added.", 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\" -o \"pcmk_delay_base=1\"") test.add_cmd("stonith_admin", "--output-as=xml -R false1 -a fence_dummy -o \"mode=fail\" -o \"pcmk_host_list=node1 node2 node3\" -o \"pcmk_delay_base=1\"") test.add_cmd("stonith_admin", "--output-as=xml -R true2 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=node1 node2 node3\"") test.add_cmd("stonith_admin", "--output-as=xml -R true3 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=node1 node2 node3\"") test.add_cmd("stonith_admin", "--output-as=xml -r node3 -i 1 -v true1") test.add_cmd("stonith_admin", "--output-as=xml -r node3 -i 1 -v false1") test.add_cmd("stonith_admin", "--output-as=xml -r node3 -i 2 -v true2") test.add_cmd("stonith_admin", "--output-as=xml -r node3 -i 2 -v true3") test.add_cmd("stonith_admin", "--output-as=xml -F node3 --delay 1") test.add_stonith_log_pattern("Delaying 'off' action targeting node3 using true1 for 2s | timeout=120s requested_delay=1s base=1s max=1s") test.add_stonith_log_pattern("Delaying 'off' action targeting node3 using false1 for 1s | timeout=120s requested_delay=0s base=1s max=1s") test.add_stonith_neg_log_pattern("Delaying 'off' action targeting node3 using true2") test.add_stonith_neg_log_pattern("Delaying 'off' action targeting node3 using true3") def build_nodeid_tests(self): """ Register tests that use a corosync node id """ our_uname = localname() ### verify nodeid is supplied when nodeid is in the metadata parameters test = self.new_test("cpg_supply_nodeid", "Verify nodeid is given when fence agent has nodeid as parameter", 1) test.add_cmd("stonith_admin", "--output-as=xml -R true1 -a fence_dummy -o \"mode=pass\" -o \"pcmk_host_list=%s\"" % (our_uname)) test.add_cmd("stonith_admin", "--output-as=xml -F %s -t 3" % (our_uname)) test.add_stonith_log_pattern("as nodeid with fence action 'off' targeting %s" % (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("as nodeid with fence action 'off' targeting regr-test") ### 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("as nodeid with fence action 'off' targeting %s" % (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("using true1 returned 0 (OK)") - test.add_stonith_log_pattern("using true2 returned 0 (OK)") + test.add_stonith_log_pattern("using true1 returned 0") + test.add_stonith_log_pattern("using true2 returned 0") ### 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("using true1 returned 0 (OK)") - test.add_stonith_log_pattern("using true2 returned 0 (OK)") + test.add_stonith_log_pattern("using true1 returned 0") + test.add_stonith_log_pattern("using true2 returned 0") ### 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("using true1 returned 0 (OK)") - test.add_stonith_log_pattern("using true2 returned 0 (OK)") - test.add_stonith_log_pattern("using true3 returned 0 (OK)") - test.add_stonith_log_pattern("using true4 returned 0 (OK)") + test.add_stonith_log_pattern("using true1 returned 0") + test.add_stonith_log_pattern("using true2 returned 0") + test.add_stonith_log_pattern("using true3 returned 0") + test.add_stonith_log_pattern("using true4 returned 0") 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 target") ### 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 target") ### 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 target") ### 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 target") def build_remap_tests(self): """ Register tests that verify remapping of reboots to off-on """ test = self.new_test("cpg_remap_simple", "Verify sequential topology reboot is remapped to all-off-then-all-on", 1) test.add_cmd("stonith_admin", """--output-as=xml -R true1 -a fence_dummy -o "mode=pass" -o "pcmk_host_list=node_fake" """ """-o "pcmk_off_timeout=1" -o "pcmk_reboot_timeout=10" """) test.add_cmd("stonith_admin", """--output-as=xml -R true2 -a fence_dummy -o "mode=pass" -o "pcmk_host_list=node_fake" """ """-o "pcmk_off_timeout=2" -o "pcmk_reboot_timeout=20" """) test.add_cmd("stonith_admin", "--output-as=xml -r node_fake -i 1 -v true1 -v true2") test.add_cmd("stonith_admin", "--output-as=xml -B node_fake -t 5") test.add_stonith_log_pattern("Remapping multiple-device reboot targeting node_fake") # timeout should be sum of off timeouts (1+2=3), not reboot timeouts (10+20=30) test.add_stonith_log_pattern("Total timeout set to 3 for peer's fencing targeting node_fake") test.add_stonith_log_pattern("perform 'off' action targeting node_fake using true1") test.add_stonith_log_pattern("perform 'off' action targeting node_fake using true2") test.add_stonith_log_pattern("Remapped 'off' targeting node_fake complete, remapping to 'on'") # fence_dummy sets "on" as an on_target action test.add_stonith_log_pattern("Ignoring true1 'on' failure (no capable peers) targeting node_fake") test.add_stonith_log_pattern("Ignoring true2 'on' failure (no capable peers) targeting node_fake") test.add_stonith_log_pattern("Undoing remap of reboot targeting node_fake") test = self.new_test("cpg_remap_automatic", "Verify remapped topology reboot skips automatic 'on'", 1) test.add_cmd("stonith_admin", """--output-as=xml -R true1 -a fence_dummy_auto_unfence """ """-o "mode=pass" -o "pcmk_host_list=node_fake" """) test.add_cmd("stonith_admin", """--output-as=xml -R true2 -a fence_dummy_auto_unfence """ """-o "mode=pass" -o "pcmk_host_list=node_fake" """) test.add_cmd("stonith_admin", "--output-as=xml -r node_fake -i 1 -v true1 -v true2") test.add_cmd("stonith_admin", "--output-as=xml -B node_fake -t 5") test.add_stonith_log_pattern("Remapping multiple-device reboot targeting node_fake") test.add_stonith_log_pattern("perform 'off' action targeting node_fake using true1") test.add_stonith_log_pattern("perform 'off' action targeting node_fake using true2") test.add_stonith_log_pattern("Remapped 'off' targeting node_fake complete, remapping to 'on'") test.add_stonith_log_pattern("Undoing remap of reboot targeting node_fake") test.add_stonith_neg_log_pattern("perform 'on' action targeting node_fake using") test.add_stonith_neg_log_pattern("'on' failure") test = self.new_test("cpg_remap_complex_1", "Verify remapped topology reboot in second level works if non-remapped first level fails", 1) test.add_cmd("stonith_admin", """--output-as=xml -R false1 -a fence_dummy -o "mode=fail" -o "pcmk_host_list=node_fake" """) test.add_cmd("stonith_admin", """--output-as=xml -R true1 -a fence_dummy -o "mode=pass" -o "pcmk_host_list=node_fake" """) test.add_cmd("stonith_admin", """--output-as=xml -R true2 -a fence_dummy -o "mode=pass" -o "pcmk_host_list=node_fake" """) test.add_cmd("stonith_admin", "--output-as=xml -r node_fake -i 1 -v false1") test.add_cmd("stonith_admin", "--output-as=xml -r node_fake -i 2 -v true1 -v true2") test.add_cmd("stonith_admin", "--output-as=xml -B node_fake -t 5") test.add_stonith_log_pattern("perform 'reboot' action targeting node_fake using false1") test.add_stonith_log_pattern("Remapping multiple-device reboot targeting node_fake") test.add_stonith_log_pattern("perform 'off' action targeting node_fake using true1") test.add_stonith_log_pattern("perform 'off' action targeting node_fake using true2") test.add_stonith_log_pattern("Remapped 'off' targeting node_fake complete, remapping to 'on'") test.add_stonith_log_pattern("Ignoring true1 'on' failure (no capable peers) targeting node_fake") test.add_stonith_log_pattern("Ignoring true2 'on' failure (no capable peers) targeting node_fake") test.add_stonith_log_pattern("Undoing remap of reboot targeting node_fake") test = self.new_test("cpg_remap_complex_2", "Verify remapped topology reboot failure in second level proceeds to third level", 1) test.add_cmd("stonith_admin", """--output-as=xml -R false1 -a fence_dummy -o "mode=fail" -o "pcmk_host_list=node_fake" """) test.add_cmd("stonith_admin", """--output-as=xml -R false2 -a fence_dummy -o "mode=fail" -o "pcmk_host_list=node_fake" """) test.add_cmd("stonith_admin", """--output-as=xml -R true1 -a fence_dummy -o "mode=pass" -o "pcmk_host_list=node_fake" """) test.add_cmd("stonith_admin", """--output-as=xml -R true2 -a fence_dummy -o "mode=pass" -o "pcmk_host_list=node_fake" """) test.add_cmd("stonith_admin", """--output-as=xml -R true3 -a fence_dummy -o "mode=pass" -o "pcmk_host_list=node_fake" """) test.add_cmd("stonith_admin", "--output-as=xml -r node_fake -i 1 -v false1") test.add_cmd("stonith_admin", "--output-as=xml -r node_fake -i 2 -v true1 -v false2 -v true3") test.add_cmd("stonith_admin", "--output-as=xml -r node_fake -i 3 -v true2") test.add_cmd("stonith_admin", "--output-as=xml -B node_fake -t 5") test.add_stonith_log_pattern("perform 'reboot' action targeting node_fake using false1") test.add_stonith_log_pattern("Remapping multiple-device reboot targeting node_fake") test.add_stonith_log_pattern("perform 'off' action targeting node_fake using true1") test.add_stonith_log_pattern("perform 'off' action targeting node_fake using false2") test.add_stonith_log_pattern("Attempted to execute agent fence_dummy (off) the maximum number of times") test.add_stonith_log_pattern("Undoing remap of reboot targeting node_fake") test.add_stonith_log_pattern("perform 'reboot' action targeting node_fake using true2") test.add_stonith_neg_log_pattern("node_fake with true3") def build_query_tests(self): """ run stonith_admin --metadata for the fence_dummy agent and check command output """ test = self.new_test("get_metadata", "Run stonith_admin --metadata for the fence_dummy agent", 1) test.add_cmd_check_stdout("stonith_admin", "--output-as=xml -a fence_dummy --metadata", ' #include #include #include #include #include #include // lrmd_event_data_t, lrmd_rsc_info_t, etc. #include #include #include #include #include #include #include #define START_DELAY_THRESHOLD 5 * 60 * 1000 #define MAX_LRM_REG_FAILS 30 struct delete_event_s { int rc; const char *rsc; lrm_state_t *lrm_state; }; static gboolean is_rsc_active(lrm_state_t * lrm_state, const char *rsc_id); static gboolean build_active_RAs(lrm_state_t * lrm_state, xmlNode * rsc_list); static gboolean stop_recurring_actions(gpointer key, gpointer value, gpointer user_data); static lrmd_event_data_t *construct_op(lrm_state_t * lrm_state, xmlNode * rsc_op, const char *rsc_id, const char *operation); static void do_lrm_rsc_op(lrm_state_t *lrm_state, lrmd_rsc_info_t *rsc, const char *operation, xmlNode *msg); static gboolean lrm_state_verify_stopped(lrm_state_t * lrm_state, enum crmd_fsa_state cur_state, int log_level); static int do_update_resource(const char *node_name, lrmd_rsc_info_t *rsc, lrmd_event_data_t *op, time_t lock_time); static void lrm_connection_destroy(void) { if (pcmk_is_set(fsa_input_register, R_LRM_CONNECTED)) { crm_crit("Connection to executor failed"); register_fsa_input(C_FSA_INTERNAL, I_ERROR, NULL); controld_clear_fsa_input_flags(R_LRM_CONNECTED); } else { crm_info("Disconnected from executor"); } } static char * make_stop_id(const char *rsc, int call_id) { return crm_strdup_printf("%s:%d", rsc, call_id); } static void copy_instance_keys(gpointer key, gpointer value, gpointer user_data) { if (strstr(key, CRM_META "_") == NULL) { g_hash_table_replace(user_data, strdup((const char *)key), strdup((const char *)value)); } } static void copy_meta_keys(gpointer key, gpointer value, gpointer user_data) { if (strstr(key, CRM_META "_") != NULL) { g_hash_table_replace(user_data, strdup((const char *)key), strdup((const char *)value)); } } /*! * \internal * \brief Remove a recurring operation from a resource's history * * \param[in,out] history Resource history to modify * \param[in] op Operation to remove * * \return TRUE if the operation was found and removed, FALSE otherwise */ static gboolean history_remove_recurring_op(rsc_history_t *history, const lrmd_event_data_t *op) { GList *iter; for (iter = history->recurring_op_list; iter != NULL; iter = iter->next) { lrmd_event_data_t *existing = iter->data; if ((op->interval_ms == existing->interval_ms) && pcmk__str_eq(op->rsc_id, existing->rsc_id, pcmk__str_none) && pcmk__str_eq(op->op_type, existing->op_type, pcmk__str_casei)) { history->recurring_op_list = g_list_delete_link(history->recurring_op_list, iter); lrmd_free_event(existing); return TRUE; } } return FALSE; } /*! * \internal * \brief Free all recurring operations in resource history * * \param[in,out] history Resource history to modify */ static void history_free_recurring_ops(rsc_history_t *history) { GList *iter; for (iter = history->recurring_op_list; iter != NULL; iter = iter->next) { lrmd_free_event(iter->data); } g_list_free(history->recurring_op_list); history->recurring_op_list = NULL; } /*! * \internal * \brief Free resource history * * \param[in,out] history Resource history to free */ void history_free(gpointer data) { rsc_history_t *history = (rsc_history_t*)data; if (history->stop_params) { g_hash_table_destroy(history->stop_params); } /* Don't need to free history->rsc.id because it's set to history->id */ free(history->rsc.type); free(history->rsc.standard); free(history->rsc.provider); lrmd_free_event(history->failed); lrmd_free_event(history->last); free(history->id); history_free_recurring_ops(history); free(history); } static void update_history_cache(lrm_state_t * lrm_state, lrmd_rsc_info_t * rsc, lrmd_event_data_t * op) { int target_rc = 0; rsc_history_t *entry = NULL; if (op->rsc_deleted) { crm_debug("Purged history for '%s' after %s", op->rsc_id, op->op_type); controld_delete_resource_history(op->rsc_id, lrm_state->node_name, NULL, crmd_cib_smart_opt()); return; } if (pcmk__str_eq(op->op_type, RSC_NOTIFY, pcmk__str_casei)) { return; } crm_debug("Updating history for '%s' with %s op", op->rsc_id, op->op_type); entry = g_hash_table_lookup(lrm_state->resource_history, op->rsc_id); if (entry == NULL && rsc) { entry = calloc(1, sizeof(rsc_history_t)); entry->id = strdup(op->rsc_id); g_hash_table_insert(lrm_state->resource_history, entry->id, entry); entry->rsc.id = entry->id; entry->rsc.type = strdup(rsc->type); entry->rsc.standard = strdup(rsc->standard); if (rsc->provider) { entry->rsc.provider = strdup(rsc->provider); } else { entry->rsc.provider = NULL; } } else if (entry == NULL) { crm_info("Resource %s no longer exists, not updating cache", op->rsc_id); return; } entry->last_callid = op->call_id; target_rc = rsc_op_expected_rc(op); if (op->op_status == PCMK_EXEC_CANCELLED) { if (op->interval_ms > 0) { crm_trace("Removing cancelled recurring op: " PCMK__OP_FMT, op->rsc_id, op->op_type, op->interval_ms); history_remove_recurring_op(entry, op); return; } else { crm_trace("Skipping " PCMK__OP_FMT " rc=%d, status=%d", op->rsc_id, op->op_type, op->interval_ms, op->rc, op->op_status); } } else if (did_rsc_op_fail(op, target_rc)) { /* Store failed monitors here, otherwise the block below will cause them * to be forgotten when a stop happens. */ if (entry->failed) { lrmd_free_event(entry->failed); } entry->failed = lrmd_copy_event(op); } else if (op->interval_ms == 0) { if (entry->last) { lrmd_free_event(entry->last); } entry->last = lrmd_copy_event(op); if (op->params && pcmk__strcase_any_of(op->op_type, CRMD_ACTION_START, CRMD_ACTION_RELOAD, CRMD_ACTION_RELOAD_AGENT, CRMD_ACTION_STATUS, NULL)) { if (entry->stop_params) { g_hash_table_destroy(entry->stop_params); } entry->stop_params = pcmk__strkey_table(free, free); g_hash_table_foreach(op->params, copy_instance_keys, entry->stop_params); } } if (op->interval_ms > 0) { /* Ensure there are no duplicates */ history_remove_recurring_op(entry, op); crm_trace("Adding recurring op: " PCMK__OP_FMT, op->rsc_id, op->op_type, op->interval_ms); entry->recurring_op_list = g_list_prepend(entry->recurring_op_list, lrmd_copy_event(op)); } else if (entry->recurring_op_list && !pcmk__str_eq(op->op_type, RSC_STATUS, pcmk__str_casei)) { crm_trace("Dropping %d recurring ops because of: " PCMK__OP_FMT, g_list_length(entry->recurring_op_list), op->rsc_id, op->op_type, op->interval_ms); history_free_recurring_ops(entry); } } /*! * \internal * \brief Send a direct OK ack for a resource task * * \param[in] lrm_state LRM connection * \param[in] input Input message being ack'ed * \param[in] rsc_id ID of affected resource * \param[in] rsc Affected resource (if available) * \param[in] task Operation task being ack'ed * \param[in] ack_host Name of host to send ack to * \param[in] ack_sys IPC system name to ack */ static void send_task_ok_ack(lrm_state_t *lrm_state, ha_msg_input_t *input, const char *rsc_id, lrmd_rsc_info_t *rsc, const char *task, const char *ack_host, const char *ack_sys) { lrmd_event_data_t *op = construct_op(lrm_state, input->xml, rsc_id, task); lrmd__set_result(op, PCMK_OCF_OK, PCMK_EXEC_DONE, NULL); controld_ack_event_directly(ack_host, ack_sys, rsc, op, rsc_id); lrmd_free_event(op); } static inline const char * op_node_name(lrmd_event_data_t *op) { return op->remote_nodename? op->remote_nodename : fsa_our_uname; } void lrm_op_callback(lrmd_event_data_t * op) { CRM_CHECK(op != NULL, return); switch (op->type) { case lrmd_event_disconnect: if (op->remote_nodename == NULL) { /* If this is the local executor IPC connection, set the right * bits in the controller when the connection goes down. */ lrm_connection_destroy(); } break; case lrmd_event_exec_complete: { lrm_state_t *lrm_state = lrm_state_find(op_node_name(op)); CRM_ASSERT(lrm_state != NULL); process_lrm_event(lrm_state, op, NULL, NULL); } break; default: break; } } static void try_local_executor_connect(long long action, fsa_data_t *msg_data, lrm_state_t *lrm_state) { int rc = pcmk_rc_ok; crm_debug("Connecting to the local executor"); // If we can connect, great rc = controld_connect_local_executor(lrm_state); if (rc == pcmk_rc_ok) { controld_set_fsa_input_flags(R_LRM_CONNECTED); crm_info("Connection to the local executor established"); return; } // Otherwise, if we can try again, set a timer to do so if (lrm_state->num_lrm_register_fails < MAX_LRM_REG_FAILS) { crm_warn("Failed to connect to the local executor %d time%s " "(%d max): %s", lrm_state->num_lrm_register_fails, pcmk__plural_s(lrm_state->num_lrm_register_fails), MAX_LRM_REG_FAILS, pcmk_rc_str(rc)); controld_start_timer(wait_timer); crmd_fsa_stall(FALSE); return; } // Otherwise give up crm_err("Failed to connect to the executor the max allowed " "%d time%s: %s", lrm_state->num_lrm_register_fails, pcmk__plural_s(lrm_state->num_lrm_register_fails), pcmk_rc_str(rc)); register_fsa_error(C_FSA_INTERNAL, I_ERROR, NULL); } /* A_LRM_CONNECT */ void do_lrm_control(long long action, enum crmd_fsa_cause cause, enum crmd_fsa_state cur_state, enum crmd_fsa_input current_input, fsa_data_t * msg_data) { /* This only pertains to local executor connections. Remote connections are * handled as resources within the scheduler. Connecting and disconnecting * from remote executor instances is handled differently. */ lrm_state_t *lrm_state = NULL; if(fsa_our_uname == NULL) { return; /* Nothing to do */ } lrm_state = lrm_state_find_or_create(fsa_our_uname); if (lrm_state == NULL) { register_fsa_error(C_FSA_INTERNAL, I_ERROR, NULL); return; } if (action & A_LRM_DISCONNECT) { if (lrm_state_verify_stopped(lrm_state, cur_state, LOG_INFO) == FALSE) { if (action == A_LRM_DISCONNECT) { crmd_fsa_stall(FALSE); return; } } controld_clear_fsa_input_flags(R_LRM_CONNECTED); crm_info("Disconnecting from the executor"); lrm_state_disconnect(lrm_state); lrm_state_reset_tables(lrm_state, FALSE); crm_notice("Disconnected from the executor"); } if (action & A_LRM_CONNECT) { try_local_executor_connect(action, msg_data, lrm_state); } if (action & ~(A_LRM_CONNECT | A_LRM_DISCONNECT)) { crm_err("Unexpected action %s in %s", fsa_action2string(action), __func__); } } static gboolean lrm_state_verify_stopped(lrm_state_t * lrm_state, enum crmd_fsa_state cur_state, int log_level) { int counter = 0; gboolean rc = TRUE; const char *when = "lrm disconnect"; GHashTableIter gIter; const char *key = NULL; rsc_history_t *entry = NULL; active_op_t *pending = NULL; crm_debug("Checking for active resources before exit"); if (cur_state == S_TERMINATE) { log_level = LOG_ERR; when = "shutdown"; } else if (pcmk_is_set(fsa_input_register, R_SHUTDOWN)) { when = "shutdown... waiting"; } if (lrm_state->pending_ops && lrm_state_is_connected(lrm_state) == TRUE) { guint removed = g_hash_table_foreach_remove( lrm_state->pending_ops, stop_recurring_actions, lrm_state); guint nremaining = g_hash_table_size(lrm_state->pending_ops); if (removed || nremaining) { crm_notice("Stopped %u recurring operation%s at %s (%u remaining)", removed, pcmk__plural_s(removed), when, nremaining); } } if (lrm_state->pending_ops) { g_hash_table_iter_init(&gIter, lrm_state->pending_ops); while (g_hash_table_iter_next(&gIter, NULL, (void **)&pending)) { /* Ignore recurring actions in the shutdown calculations */ if (pending->interval_ms == 0) { counter++; } } } if (counter > 0) { do_crm_log(log_level, "%d pending executor operation%s at %s", counter, pcmk__plural_s(counter), when); if ((cur_state == S_TERMINATE) || !pcmk_is_set(fsa_input_register, R_SENT_RSC_STOP)) { g_hash_table_iter_init(&gIter, lrm_state->pending_ops); while (g_hash_table_iter_next(&gIter, (gpointer*)&key, (gpointer*)&pending)) { do_crm_log(log_level, "Pending action: %s (%s)", key, pending->op_key); } } else { rc = FALSE; } return rc; } if (lrm_state->resource_history == NULL) { return rc; } if (pcmk_is_set(fsa_input_register, R_SHUTDOWN)) { /* At this point we're not waiting, we're just shutting down */ when = "shutdown"; } counter = 0; g_hash_table_iter_init(&gIter, lrm_state->resource_history); while (g_hash_table_iter_next(&gIter, NULL, (gpointer*)&entry)) { if (is_rsc_active(lrm_state, entry->id) == FALSE) { continue; } counter++; if (log_level == LOG_ERR) { crm_info("Found %s active at %s", entry->id, when); } else { crm_trace("Found %s active at %s", entry->id, when); } if (lrm_state->pending_ops) { GHashTableIter hIter; g_hash_table_iter_init(&hIter, lrm_state->pending_ops); while (g_hash_table_iter_next(&hIter, (gpointer*)&key, (gpointer*)&pending)) { if (pcmk__str_eq(entry->id, pending->rsc_id, pcmk__str_none)) { crm_notice("%sction %s (%s) incomplete at %s", pending->interval_ms == 0 ? "A" : "Recurring a", key, pending->op_key, when); } } } } if (counter) { crm_err("%d resource%s active at %s", counter, (counter == 1)? " was" : "s were", when); } return rc; } /*! * \internal * \brief Build XML and string of parameters meeting some criteria, for digest * * \param[in] op Executor event with parameter table to use * \param[in] metadata Parsed meta-data for executed resource agent * \param[in] param_type Flag used for selection criteria * \param[out] result Will be set to newly created XML with selected * parameters as attributes * * \return Newly allocated space-separated string of parameter names * \note Selection criteria varies by param_type: for the restart digest, we * want parameters that are *not* marked reloadable (OCF 1.1) or that * *are* marked unique (pre-1.1), for both string and XML results; for the * secure digest, we want parameters that *are* marked private for the * string, but parameters that are *not* marked private for the XML. * \note It is the caller's responsibility to free the string return value with * free() and the XML result with free_xml(). */ static char * build_parameter_list(const lrmd_event_data_t *op, const struct ra_metadata_s *metadata, enum ra_param_flags_e param_type, xmlNode **result) { char *list = NULL; size_t len = 0; *result = create_xml_node(NULL, XML_TAG_PARAMS); for (GList *iter = metadata->ra_params; iter != NULL; iter = iter->next) { struct ra_param_s *param = (struct ra_param_s *) iter->data; bool accept_for_list = false; bool accept_for_xml = false; switch (param_type) { case ra_param_reloadable: accept_for_list = !pcmk_is_set(param->rap_flags, param_type); accept_for_xml = accept_for_list; break; case ra_param_unique: accept_for_list = pcmk_is_set(param->rap_flags, param_type); accept_for_xml = accept_for_list; break; case ra_param_private: accept_for_list = pcmk_is_set(param->rap_flags, param_type); accept_for_xml = !accept_for_list; break; } if (accept_for_list) { crm_trace("Attr %s is %s", param->rap_name, ra_param_flag2text(param_type)); if (list == NULL) { // We will later search for " WORD ", so start list with a space pcmk__add_word(&list, &len, " "); } pcmk__add_word(&list, &len, param->rap_name); } else { crm_trace("Rejecting %s for %s", param->rap_name, ra_param_flag2text(param_type)); } if (accept_for_xml) { const char *v = g_hash_table_lookup(op->params, param->rap_name); if (v != NULL) { crm_trace("Adding attr %s=%s to the xml result", param->rap_name, v); crm_xml_add(*result, param->rap_name, v); } } } if (list != NULL) { // We will later search for " WORD ", so end list with a space pcmk__add_word(&list, &len, " "); } return list; } static void append_restart_list(lrmd_event_data_t *op, struct ra_metadata_s *metadata, xmlNode *update, const char *version) { char *list = NULL; char *digest = NULL; xmlNode *restart = NULL; CRM_LOG_ASSERT(op->params != NULL); if (op->interval_ms > 0) { /* monitors are not reloadable */ return; } if (pcmk_is_set(metadata->ra_flags, ra_supports_reload_agent)) { // Add parameters not marked reloadable to the "op-force-restart" list list = build_parameter_list(op, metadata, ra_param_reloadable, &restart); } else if (pcmk_is_set(metadata->ra_flags, ra_supports_legacy_reload)) { /* @COMPAT pre-OCF-1.1 resource agents * * Before OCF 1.1, Pacemaker abused "unique=0" to indicate * reloadability. Add any parameters with unique="1" to the * "op-force-restart" list. */ list = build_parameter_list(op, metadata, ra_param_unique, &restart); } else { // Resource does not support agent reloads return; } digest = calculate_operation_digest(restart, version); /* Add "op-force-restart" and "op-restart-digest" to indicate the resource supports reload, * no matter if it actually supports any parameters with unique="1"). */ crm_xml_add(update, XML_LRM_ATTR_OP_RESTART, list? list: ""); crm_xml_add(update, XML_LRM_ATTR_RESTART_DIGEST, digest); crm_trace("%s: %s, %s", op->rsc_id, digest, list); crm_log_xml_trace(restart, "restart digest source"); free_xml(restart); free(digest); free(list); } static void append_secure_list(lrmd_event_data_t *op, struct ra_metadata_s *metadata, xmlNode *update, const char *version) { char *list = NULL; char *digest = NULL; xmlNode *secure = NULL; CRM_LOG_ASSERT(op->params != NULL); /* * To keep XML_LRM_ATTR_OP_SECURE short, we want it to contain the * secure parameters but XML_LRM_ATTR_SECURE_DIGEST to be based on * the insecure ones */ list = build_parameter_list(op, metadata, ra_param_private, &secure); if (list != NULL) { digest = calculate_operation_digest(secure, version); crm_xml_add(update, XML_LRM_ATTR_OP_SECURE, list); crm_xml_add(update, XML_LRM_ATTR_SECURE_DIGEST, digest); crm_trace("%s: %s, %s", op->rsc_id, digest, list); crm_log_xml_trace(secure, "secure digest source"); } else { crm_trace("%s: no secure parameters", op->rsc_id); } free_xml(secure); free(digest); free(list); } static gboolean build_operation_update(xmlNode * parent, lrmd_rsc_info_t * rsc, lrmd_event_data_t * op, const char *node_name, const char *src) { int target_rc = 0; xmlNode *xml_op = NULL; struct ra_metadata_s *metadata = NULL; const char *caller_version = NULL; lrm_state_t *lrm_state = NULL; uint32_t metadata_source = controld_metadata_from_agent; if (op == NULL) { return FALSE; } target_rc = rsc_op_expected_rc(op); /* there is a small risk in formerly mixed clusters that it will * be sub-optimal. * * however with our upgrade policy, the update we send should * still be completely supported anyway */ caller_version = g_hash_table_lookup(op->params, XML_ATTR_CRM_VERSION); CRM_LOG_ASSERT(caller_version != NULL); if(caller_version == NULL) { caller_version = CRM_FEATURE_SET; } crm_trace("Building %s operation update with originator version: %s", op->rsc_id, caller_version); xml_op = pcmk__create_history_xml(parent, op, caller_version, target_rc, fsa_our_uname, src, LOG_DEBUG); if (xml_op == NULL) { return TRUE; } if ((rsc == NULL) || (op->params == NULL) || !crm_op_needs_metadata(rsc->standard, op->op_type)) { crm_trace("No digests needed for %s action on %s (params=%p rsc=%p)", op->op_type, op->rsc_id, op->params, rsc); return TRUE; } lrm_state = lrm_state_find(node_name); if (lrm_state == NULL) { crm_warn("Cannot calculate digests for operation " PCMK__OP_FMT " because we have no connection to executor for %s", op->rsc_id, op->op_type, op->interval_ms, node_name); return TRUE; } /* Getting meta-data from cache is OK unless this is a successful start * action -- always refresh from the agent for those, in case the * resource agent was updated. * * @TODO Only refresh the meta-data after starts if the agent actually * changed (using something like inotify, or a hash or modification time of * the agent executable). */ if ((op->op_status != PCMK_EXEC_DONE) || (op->rc != target_rc) || !pcmk__str_eq(op->op_type, CRMD_ACTION_START, pcmk__str_none)) { metadata_source |= controld_metadata_from_cache; } metadata = controld_get_rsc_metadata(lrm_state, rsc, metadata_source); if (metadata == NULL) { return TRUE; } #if ENABLE_VERSIONED_ATTRS crm_xml_add(xml_op, XML_ATTR_RA_VERSION, metadata->ra_version); #endif crm_trace("Including additional digests for %s:%s:%s", rsc->standard, rsc->provider, rsc->type); append_restart_list(op, metadata, xml_op, caller_version); append_secure_list(op, metadata, xml_op, caller_version); return TRUE; } static gboolean is_rsc_active(lrm_state_t * lrm_state, const char *rsc_id) { rsc_history_t *entry = NULL; entry = g_hash_table_lookup(lrm_state->resource_history, rsc_id); if (entry == NULL || entry->last == NULL) { return FALSE; } crm_trace("Processing %s: %s.%d=%d", rsc_id, entry->last->op_type, entry->last->interval_ms, entry->last->rc); if (entry->last->rc == PCMK_OCF_OK && pcmk__str_eq(entry->last->op_type, CRMD_ACTION_STOP, pcmk__str_casei)) { return FALSE; } else if (entry->last->rc == PCMK_OCF_OK && pcmk__str_eq(entry->last->op_type, CRMD_ACTION_MIGRATE, pcmk__str_casei)) { // A stricter check is too complex ... leave that to the scheduler return FALSE; } else if (entry->last->rc == PCMK_OCF_NOT_RUNNING) { return FALSE; } else if ((entry->last->interval_ms == 0) && (entry->last->rc == PCMK_OCF_NOT_CONFIGURED)) { /* Badly configured resources can't be reliably stopped */ return FALSE; } return TRUE; } static gboolean build_active_RAs(lrm_state_t * lrm_state, xmlNode * rsc_list) { GHashTableIter iter; rsc_history_t *entry = NULL; g_hash_table_iter_init(&iter, lrm_state->resource_history); while (g_hash_table_iter_next(&iter, NULL, (void **)&entry)) { GList *gIter = NULL; xmlNode *xml_rsc = create_xml_node(rsc_list, XML_LRM_TAG_RESOURCE); crm_xml_add(xml_rsc, XML_ATTR_ID, entry->id); crm_xml_add(xml_rsc, XML_ATTR_TYPE, entry->rsc.type); crm_xml_add(xml_rsc, XML_AGENT_ATTR_CLASS, entry->rsc.standard); crm_xml_add(xml_rsc, XML_AGENT_ATTR_PROVIDER, entry->rsc.provider); if (entry->last && entry->last->params) { const char *container = g_hash_table_lookup(entry->last->params, CRM_META"_"XML_RSC_ATTR_CONTAINER); if (container) { crm_trace("Resource %s is a part of container resource %s", entry->id, container); crm_xml_add(xml_rsc, XML_RSC_ATTR_CONTAINER, container); } } build_operation_update(xml_rsc, &(entry->rsc), entry->failed, lrm_state->node_name, __func__); build_operation_update(xml_rsc, &(entry->rsc), entry->last, lrm_state->node_name, __func__); for (gIter = entry->recurring_op_list; gIter != NULL; gIter = gIter->next) { build_operation_update(xml_rsc, &(entry->rsc), gIter->data, lrm_state->node_name, __func__); } } return FALSE; } static xmlNode * do_lrm_query_internal(lrm_state_t *lrm_state, int update_flags) { xmlNode *xml_state = NULL; xmlNode *xml_data = NULL; xmlNode *rsc_list = NULL; crm_node_t *peer = NULL; peer = crm_get_peer_full(0, lrm_state->node_name, CRM_GET_PEER_ANY); CRM_CHECK(peer != NULL, return NULL); xml_state = create_node_state_update(peer, update_flags, NULL, __func__); if (xml_state == NULL) { return NULL; } xml_data = create_xml_node(xml_state, XML_CIB_TAG_LRM); crm_xml_add(xml_data, XML_ATTR_ID, peer->uuid); rsc_list = create_xml_node(xml_data, XML_LRM_TAG_RESOURCES); /* Build a list of active (not always running) resources */ build_active_RAs(lrm_state, rsc_list); crm_log_xml_trace(xml_state, "Current executor state"); return xml_state; } xmlNode * controld_query_executor_state(const char *node_name) { lrm_state_t *lrm_state = lrm_state_find(node_name); if (!lrm_state) { crm_err("Could not find executor state for node %s", node_name); return NULL; } return do_lrm_query_internal(lrm_state, node_update_cluster|node_update_peer); } /*! * \internal * \brief Map standard Pacemaker return code to operation status and OCF code * * \param[out] event Executor event whose status and return code should be set * \param[in] rc Standard Pacemaker return code */ void controld_rc2event(lrmd_event_data_t *event, int rc) { /* This is called for cleanup requests from controller peers/clients, not * for resource actions, so no exit reason is needed. */ switch (rc) { case pcmk_rc_ok: lrmd__set_result(event, PCMK_OCF_OK, PCMK_EXEC_DONE, NULL); break; case EACCES: lrmd__set_result(event, PCMK_OCF_INSUFFICIENT_PRIV, PCMK_EXEC_ERROR, NULL); break; default: lrmd__set_result(event, PCMK_OCF_UNKNOWN_ERROR, PCMK_EXEC_ERROR, NULL); break; } } /*! * \internal * \brief Trigger a new transition after CIB status was deleted * * If a CIB status delete was not expected (as part of the transition graph), * trigger a new transition by updating the (arbitrary) "last-lrm-refresh" * cluster property. * * \param[in] from_sys IPC name that requested the delete * \param[in] rsc_id Resource whose status was deleted (for logging only) */ void controld_trigger_delete_refresh(const char *from_sys, const char *rsc_id) { if (!pcmk__str_eq(from_sys, CRM_SYSTEM_TENGINE, pcmk__str_casei)) { char *now_s = crm_strdup_printf("%lld", (long long) time(NULL)); crm_debug("Triggering a refresh after %s cleaned %s", from_sys, rsc_id); update_attr_delegate(fsa_cib_conn, cib_none, XML_CIB_TAG_CRMCONFIG, NULL, NULL, NULL, NULL, "last-lrm-refresh", now_s, FALSE, NULL, NULL); free(now_s); } } static void notify_deleted(lrm_state_t * lrm_state, ha_msg_input_t * input, const char *rsc_id, int rc) { lrmd_event_data_t *op = NULL; const char *from_sys = crm_element_value(input->msg, F_CRM_SYS_FROM); const char *from_host = crm_element_value(input->msg, F_CRM_HOST_FROM); crm_info("Notifying %s on %s that %s was%s deleted", from_sys, (from_host? from_host : "localhost"), rsc_id, ((rc == pcmk_ok)? "" : " not")); op = construct_op(lrm_state, input->xml, rsc_id, CRMD_ACTION_DELETE); controld_rc2event(op, pcmk_legacy2rc(rc)); controld_ack_event_directly(from_host, from_sys, NULL, op, rsc_id); lrmd_free_event(op); controld_trigger_delete_refresh(from_sys, rsc_id); } static gboolean lrm_remove_deleted_rsc(gpointer key, gpointer value, gpointer user_data) { struct delete_event_s *event = user_data; struct pending_deletion_op_s *op = value; if (pcmk__str_eq(event->rsc, op->rsc, pcmk__str_none)) { notify_deleted(event->lrm_state, op->input, event->rsc, event->rc); return TRUE; } return FALSE; } static gboolean lrm_remove_deleted_op(gpointer key, gpointer value, gpointer user_data) { const char *rsc = user_data; active_op_t *pending = value; if (pcmk__str_eq(rsc, pending->rsc_id, pcmk__str_none)) { crm_info("Removing op %s:%d for deleted resource %s", pending->op_key, pending->call_id, rsc); return TRUE; } return FALSE; } static void delete_rsc_entry(lrm_state_t * lrm_state, ha_msg_input_t * input, const char *rsc_id, GHashTableIter * rsc_gIter, int rc, const char *user_name) { struct delete_event_s event; CRM_CHECK(rsc_id != NULL, return); if (rc == pcmk_ok) { char *rsc_id_copy = strdup(rsc_id); if (rsc_gIter) { g_hash_table_iter_remove(rsc_gIter); } else { g_hash_table_remove(lrm_state->resource_history, rsc_id_copy); } controld_delete_resource_history(rsc_id_copy, lrm_state->node_name, user_name, crmd_cib_smart_opt()); g_hash_table_foreach_remove(lrm_state->pending_ops, lrm_remove_deleted_op, rsc_id_copy); free(rsc_id_copy); } if (input) { notify_deleted(lrm_state, input, rsc_id, rc); } event.rc = rc; event.rsc = rsc_id; event.lrm_state = lrm_state; g_hash_table_foreach_remove(lrm_state->deletion_ops, lrm_remove_deleted_rsc, &event); } /*! * \internal * \brief Erase an LRM history entry from the CIB, given the operation data * * \param[in] lrm_state LRM state of the desired node * \param[in] op Operation whose history should be deleted */ static void erase_lrm_history_by_op(lrm_state_t *lrm_state, lrmd_event_data_t *op) { xmlNode *xml_top = NULL; CRM_CHECK(op != NULL, return); xml_top = create_xml_node(NULL, XML_LRM_TAG_RSC_OP); crm_xml_add_int(xml_top, XML_LRM_ATTR_CALLID, op->call_id); crm_xml_add(xml_top, XML_ATTR_TRANSITION_KEY, op->user_data); if (op->interval_ms > 0) { char *op_id = pcmk__op_key(op->rsc_id, op->op_type, op->interval_ms); /* Avoid deleting last_failure too (if it was a result of this recurring op failing) */ crm_xml_add(xml_top, XML_ATTR_ID, op_id); free(op_id); } crm_debug("Erasing resource operation history for " PCMK__OP_FMT " (call=%d)", op->rsc_id, op->op_type, op->interval_ms, op->call_id); fsa_cib_conn->cmds->remove(fsa_cib_conn, XML_CIB_TAG_STATUS, xml_top, cib_quorum_override); crm_log_xml_trace(xml_top, "op:cancel"); free_xml(xml_top); } /* Define xpath to find LRM resource history entry by node and resource */ #define XPATH_HISTORY \ "/" XML_TAG_CIB "/" XML_CIB_TAG_STATUS \ "/" XML_CIB_TAG_STATE "[@" XML_ATTR_UNAME "='%s']" \ "/" XML_CIB_TAG_LRM "/" XML_LRM_TAG_RESOURCES \ "/" XML_LRM_TAG_RESOURCE "[@" XML_ATTR_ID "='%s']" \ "/" XML_LRM_TAG_RSC_OP /* ... and also by operation key */ #define XPATH_HISTORY_ID XPATH_HISTORY \ "[@" XML_ATTR_ID "='%s']" /* ... and also by operation key and operation call ID */ #define XPATH_HISTORY_CALL XPATH_HISTORY \ "[@" XML_ATTR_ID "='%s' and @" XML_LRM_ATTR_CALLID "='%d']" /* ... and also by operation key and original operation key */ #define XPATH_HISTORY_ORIG XPATH_HISTORY \ "[@" XML_ATTR_ID "='%s' and @" XML_LRM_ATTR_TASK_KEY "='%s']" /*! * \internal * \brief Erase an LRM history entry from the CIB, given operation identifiers * * \param[in] lrm_state LRM state of the node to clear history for * \param[in] rsc_id Name of resource to clear history for * \param[in] key Operation key of operation to clear history for * \param[in] orig_op If specified, delete only if it has this original op * \param[in] call_id If specified, delete entry only if it has this call ID */ static void erase_lrm_history_by_id(lrm_state_t *lrm_state, const char *rsc_id, const char *key, const char *orig_op, int call_id) { char *op_xpath = NULL; CRM_CHECK((rsc_id != NULL) && (key != NULL), return); if (call_id > 0) { op_xpath = crm_strdup_printf(XPATH_HISTORY_CALL, lrm_state->node_name, rsc_id, key, call_id); } else if (orig_op) { op_xpath = crm_strdup_printf(XPATH_HISTORY_ORIG, lrm_state->node_name, rsc_id, key, orig_op); } else { op_xpath = crm_strdup_printf(XPATH_HISTORY_ID, lrm_state->node_name, rsc_id, key); } crm_debug("Erasing resource operation history for %s on %s (call=%d)", key, rsc_id, call_id); fsa_cib_conn->cmds->remove(fsa_cib_conn, op_xpath, NULL, cib_quorum_override | cib_xpath); free(op_xpath); } static inline gboolean last_failed_matches_op(rsc_history_t *entry, const char *op, guint interval_ms) { if (entry == NULL) { return FALSE; } if (op == NULL) { return TRUE; } return (pcmk__str_eq(op, entry->failed->op_type, pcmk__str_casei) && (interval_ms == entry->failed->interval_ms)); } /*! * \internal * \brief Clear a resource's last failure * * Erase a resource's last failure on a particular node from both the * LRM resource history in the CIB, and the resource history remembered * for the LRM state. * * \param[in] rsc_id Resource name * \param[in] node_name Node name * \param[in] operation If specified, only clear if matching this operation * \param[in] interval_ms If operation is specified, it has this interval */ void lrm_clear_last_failure(const char *rsc_id, const char *node_name, const char *operation, guint interval_ms) { char *op_key = NULL; char *orig_op_key = NULL; lrm_state_t *lrm_state = NULL; lrm_state = lrm_state_find(node_name); if (lrm_state == NULL) { return; } /* Erase from CIB */ op_key = pcmk__op_key(rsc_id, "last_failure", 0); if (operation) { orig_op_key = pcmk__op_key(rsc_id, operation, interval_ms); } erase_lrm_history_by_id(lrm_state, rsc_id, op_key, orig_op_key, 0); free(op_key); free(orig_op_key); /* Remove from memory */ if (lrm_state->resource_history) { rsc_history_t *entry = g_hash_table_lookup(lrm_state->resource_history, rsc_id); if (last_failed_matches_op(entry, operation, interval_ms)) { lrmd_free_event(entry->failed); entry->failed = NULL; } } } /* Returns: gboolean - cancellation is in progress */ static gboolean cancel_op(lrm_state_t * lrm_state, const char *rsc_id, const char *key, int op, gboolean remove) { int rc = pcmk_ok; char *local_key = NULL; active_op_t *pending = NULL; CRM_CHECK(op != 0, return FALSE); CRM_CHECK(rsc_id != NULL, return FALSE); if (key == NULL) { local_key = make_stop_id(rsc_id, op); key = local_key; } pending = g_hash_table_lookup(lrm_state->pending_ops, key); if (pending) { if (remove && !pcmk_is_set(pending->flags, active_op_remove)) { controld_set_active_op_flags(pending, active_op_remove); crm_debug("Scheduling %s for removal", key); } if (pcmk_is_set(pending->flags, active_op_cancelled)) { crm_debug("Operation %s already cancelled", key); free(local_key); return FALSE; } controld_set_active_op_flags(pending, active_op_cancelled); } else { crm_info("No pending op found for %s", key); free(local_key); return FALSE; } crm_debug("Cancelling op %d for %s (%s)", op, rsc_id, key); rc = lrm_state_cancel(lrm_state, pending->rsc_id, pending->op_type, pending->interval_ms); if (rc == pcmk_ok) { crm_debug("Op %d for %s (%s): cancelled", op, rsc_id, key); free(local_key); return TRUE; } crm_debug("Op %d for %s (%s): Nothing to cancel", op, rsc_id, key); /* The caller needs to make sure the entry is * removed from the pending_ops list * * Usually by returning TRUE inside the worker function * supplied to g_hash_table_foreach_remove() * * Not removing the entry from pending_ops will block * the node from shutting down */ free(local_key); return FALSE; } struct cancel_data { gboolean done; gboolean remove; const char *key; lrmd_rsc_info_t *rsc; lrm_state_t *lrm_state; }; static gboolean cancel_action_by_key(gpointer key, gpointer value, gpointer user_data) { gboolean remove = FALSE; struct cancel_data *data = user_data; active_op_t *op = value; if (pcmk__str_eq(op->op_key, data->key, pcmk__str_none)) { data->done = TRUE; remove = !cancel_op(data->lrm_state, data->rsc->id, key, op->call_id, data->remove); } return remove; } static gboolean cancel_op_key(lrm_state_t * lrm_state, lrmd_rsc_info_t * rsc, const char *key, gboolean remove) { guint removed = 0; struct cancel_data data; CRM_CHECK(rsc != NULL, return FALSE); CRM_CHECK(key != NULL, return FALSE); data.key = key; data.rsc = rsc; data.done = FALSE; data.remove = remove; data.lrm_state = lrm_state; removed = g_hash_table_foreach_remove(lrm_state->pending_ops, cancel_action_by_key, &data); crm_trace("Removed %u op cache entries, new size: %u", removed, g_hash_table_size(lrm_state->pending_ops)); return data.done; } /*! * \internal * \brief Retrieve resource information from LRM * * \param[in] lrm_state LRM connection to use * \param[in] rsc_xml XML containing resource configuration * \param[in] do_create If true, register resource with LRM if not already * \param[out] rsc_info Where to store resource information obtained from LRM * * \retval pcmk_ok Success (and rsc_info holds newly allocated result) * \retval -EINVAL Required information is missing from arguments * \retval -ENOTCONN No active connection to LRM * \retval -ENODEV Resource not found * \retval -errno Error communicating with executor when registering resource * * \note Caller is responsible for freeing result on success. */ static int get_lrm_resource(lrm_state_t *lrm_state, xmlNode *rsc_xml, gboolean do_create, lrmd_rsc_info_t **rsc_info) { const char *id = ID(rsc_xml); CRM_CHECK(lrm_state && rsc_xml && rsc_info, return -EINVAL); CRM_CHECK(id, return -EINVAL); if (lrm_state_is_connected(lrm_state) == FALSE) { return -ENOTCONN; } crm_trace("Retrieving resource information for %s from the executor", id); *rsc_info = lrm_state_get_rsc_info(lrm_state, id, 0); // If resource isn't known by ID, try clone name, if provided if (!*rsc_info) { const char *long_id = crm_element_value(rsc_xml, XML_ATTR_ID_LONG); if (long_id) { *rsc_info = lrm_state_get_rsc_info(lrm_state, long_id, 0); } } if ((*rsc_info == NULL) && do_create) { const char *class = crm_element_value(rsc_xml, XML_AGENT_ATTR_CLASS); const char *provider = crm_element_value(rsc_xml, XML_AGENT_ATTR_PROVIDER); const char *type = crm_element_value(rsc_xml, XML_ATTR_TYPE); int rc; crm_trace("Registering resource %s with the executor", id); rc = lrm_state_register_rsc(lrm_state, id, class, provider, type, lrmd_opt_drop_recurring); if (rc != pcmk_ok) { fsa_data_t *msg_data = NULL; crm_err("Could not register resource %s with the executor on %s: %s " CRM_XS " rc=%d", id, lrm_state->node_name, pcmk_strerror(rc), rc); /* Register this as an internal error if this involves the local * executor. Otherwise, we're likely dealing with an unresponsive * remote node, which is not an FSA failure. */ if (lrm_state_is_local(lrm_state) == TRUE) { register_fsa_error(C_FSA_INTERNAL, I_FAIL, NULL); } return rc; } *rsc_info = lrm_state_get_rsc_info(lrm_state, id, 0); } return *rsc_info? pcmk_ok : -ENODEV; } static void delete_resource(lrm_state_t * lrm_state, const char *id, lrmd_rsc_info_t * rsc, GHashTableIter * gIter, const char *sys, const char *user, ha_msg_input_t * request, gboolean unregister) { int rc = pcmk_ok; crm_info("Removing resource %s from executor for %s%s%s", id, sys, (user? " as " : ""), (user? user : "")); if (rsc && unregister) { rc = lrm_state_unregister_rsc(lrm_state, id, 0); } if (rc == pcmk_ok) { crm_trace("Resource %s deleted from executor", id); } else if (rc == -EINPROGRESS) { crm_info("Deletion of resource '%s' from executor is pending", id); if (request) { struct pending_deletion_op_s *op = NULL; char *ref = crm_element_value_copy(request->msg, XML_ATTR_REFERENCE); op = calloc(1, sizeof(struct pending_deletion_op_s)); op->rsc = strdup(rsc->id); op->input = copy_ha_msg_input(request); g_hash_table_insert(lrm_state->deletion_ops, ref, op); } return; } else { crm_warn("Could not delete '%s' from executor for %s%s%s: %s " CRM_XS " rc=%d", id, sys, (user? " as " : ""), (user? user : ""), pcmk_strerror(rc), rc); } delete_rsc_entry(lrm_state, request, id, gIter, rc, user); } static int get_fake_call_id(lrm_state_t *lrm_state, const char *rsc_id) { int call_id = 999999999; rsc_history_t *entry = NULL; if(lrm_state) { entry = g_hash_table_lookup(lrm_state->resource_history, rsc_id); } /* Make sure the call id is greater than the last successful operation, * otherwise the failure will not result in a possible recovery of the resource * as it could appear the failure occurred before the successful start */ if (entry) { call_id = entry->last_callid + 1; } if (call_id < 0) { call_id = 1; } return call_id; } static void fake_op_status(lrm_state_t *lrm_state, lrmd_event_data_t *op, int op_status, enum ocf_exitcode op_exitcode, const char *exit_reason) { op->call_id = get_fake_call_id(lrm_state, op->rsc_id); op->t_run = time(NULL); op->t_rcchange = op->t_run; lrmd__set_result(op, op_exitcode, op_status, exit_reason); } static void force_reprobe(lrm_state_t *lrm_state, const char *from_sys, const char *from_host, const char *user_name, gboolean is_remote_node) { GHashTableIter gIter; rsc_history_t *entry = NULL; crm_info("Clearing resource history on node %s", lrm_state->node_name); g_hash_table_iter_init(&gIter, lrm_state->resource_history); while (g_hash_table_iter_next(&gIter, NULL, (void **)&entry)) { /* only unregister the resource during a reprobe if it is not a remote connection * resource. otherwise unregistering the connection will terminate remote-node * membership */ gboolean unregister = TRUE; if (is_remote_lrmd_ra(NULL, NULL, entry->id)) { lrm_state_t *remote_lrm_state = lrm_state_find(entry->id); if (remote_lrm_state) { /* when forcing a reprobe, make sure to clear remote node before * clearing the remote node's connection resource */ force_reprobe(remote_lrm_state, from_sys, from_host, user_name, TRUE); } unregister = FALSE; } delete_resource(lrm_state, entry->id, &entry->rsc, &gIter, from_sys, user_name, NULL, unregister); } /* Now delete the copy in the CIB */ controld_delete_node_state(lrm_state->node_name, controld_section_lrm, cib_scope_local); /* Finally, _delete_ the value in pacemaker-attrd -- setting it to FALSE * would result in the scheduler sending us back here again */ update_attrd(lrm_state->node_name, CRM_OP_PROBED, NULL, user_name, is_remote_node); } /*! * \internal * \brief Fail a requested action without actually executing it * * For an action that can't be executed, process it similarly to an actual * execution result, with specified error status (except for notify actions, * which will always be treated as successful). * * \param[in] lrm_state Executor connection that action is for * \param[in] action Action XML from request * \param[in] rc Desired return code to use * \param[in] op_status Desired operation status to use * \param[in] exit_reason Human-friendly detail, if error */ static void synthesize_lrmd_failure(lrm_state_t *lrm_state, xmlNode *action, int op_status, enum ocf_exitcode rc, const char *exit_reason) { lrmd_event_data_t *op = NULL; const char *operation = crm_element_value(action, XML_LRM_ATTR_TASK); const char *target_node = crm_element_value(action, XML_LRM_ATTR_TARGET); xmlNode *xml_rsc = find_xml_node(action, XML_CIB_TAG_RESOURCE, TRUE); if ((xml_rsc == NULL) || (ID(xml_rsc) == NULL)) { /* @TODO Should we do something else, like direct ack? */ crm_info("Can't fake %s failure (%d) on %s without resource configuration", crm_element_value(action, XML_LRM_ATTR_TASK_KEY), rc, target_node); return; } else if(operation == NULL) { /* This probably came from crm_resource -C, nothing to do */ crm_info("Can't fake %s failure (%d) on %s without operation", ID(xml_rsc), rc, target_node); return; } op = construct_op(lrm_state, action, ID(xml_rsc), operation); if (pcmk__str_eq(operation, RSC_NOTIFY, pcmk__str_casei)) { // Notifications can't fail fake_op_status(lrm_state, op, PCMK_EXEC_DONE, PCMK_OCF_OK, NULL); } else { fake_op_status(lrm_state, op, op_status, rc, exit_reason); } crm_info("Faking " PCMK__OP_FMT " result (%d) on %s", op->rsc_id, op->op_type, op->interval_ms, op->rc, target_node); // Process the result as if it came from the LRM process_lrm_event(lrm_state, op, NULL, action); lrmd_free_event(op); } /*! * \internal * \brief Get target of an LRM operation * * \param[in] xml LRM operation data XML * * \return LRM operation target node name (local node or Pacemaker Remote node) */ static const char * lrm_op_target(xmlNode *xml) { const char *target = NULL; if (xml) { target = crm_element_value(xml, XML_LRM_ATTR_TARGET); } if (target == NULL) { target = fsa_our_uname; } return target; } static void fail_lrm_resource(xmlNode *xml, lrm_state_t *lrm_state, const char *user_name, const char *from_host, const char *from_sys) { lrmd_event_data_t *op = NULL; lrmd_rsc_info_t *rsc = NULL; xmlNode *xml_rsc = find_xml_node(xml, XML_CIB_TAG_RESOURCE, TRUE); CRM_CHECK(xml_rsc != NULL, return); /* The executor simply executes operations and reports the results, without * any concept of success or failure, so to fail a resource, we must fake * what a failure looks like. * * To do this, we create a fake executor operation event for the resource, * and pass that event to the executor client callback so it will be * processed as if it came from the executor. */ op = construct_op(lrm_state, xml, ID(xml_rsc), "asyncmon"); free((char*) op->user_data); op->user_data = NULL; op->interval_ms = 0; if (user_name && !pcmk__is_privileged(user_name)) { crm_err("%s does not have permission to fail %s", user_name, ID(xml_rsc)); fake_op_status(lrm_state, op, PCMK_EXEC_ERROR, PCMK_OCF_INSUFFICIENT_PRIV, "Unprivileged user cannot fail resources"); controld_ack_event_directly(from_host, from_sys, NULL, op, ID(xml_rsc)); lrmd_free_event(op); return; } if (get_lrm_resource(lrm_state, xml_rsc, TRUE, &rsc) == pcmk_ok) { crm_info("Failing resource %s...", rsc->id); fake_op_status(lrm_state, op, PCMK_EXEC_DONE, PCMK_OCF_UNKNOWN_ERROR, "Simulated failure"); process_lrm_event(lrm_state, op, NULL, xml); op->rc = PCMK_OCF_OK; // The request to fail the resource succeeded lrmd_free_rsc_info(rsc); } else { crm_info("Cannot find/create resource in order to fail it..."); crm_log_xml_warn(xml, "bad input"); fake_op_status(lrm_state, op, PCMK_EXEC_ERROR, PCMK_OCF_UNKNOWN_ERROR, "Cannot fail unknown resource"); } controld_ack_event_directly(from_host, from_sys, NULL, op, ID(xml_rsc)); lrmd_free_event(op); } static void handle_refresh_op(lrm_state_t *lrm_state, const char *user_name, const char *from_host, const char *from_sys) { int rc = pcmk_ok; xmlNode *fragment = do_lrm_query_internal(lrm_state, node_update_all); fsa_cib_update(XML_CIB_TAG_STATUS, fragment, cib_quorum_override, rc, user_name); crm_info("Forced a local resource history refresh: call=%d", rc); if (!pcmk__str_eq(CRM_SYSTEM_CRMD, from_sys, pcmk__str_casei)) { xmlNode *reply = create_request(CRM_OP_INVOKE_LRM, fragment, from_host, from_sys, CRM_SYSTEM_LRMD, fsa_our_uuid); crm_debug("ACK'ing refresh from %s (%s)", from_sys, from_host); if (relay_message(reply, TRUE) == FALSE) { crm_log_xml_err(reply, "Unable to route reply"); } free_xml(reply); } free_xml(fragment); } static void handle_query_op(xmlNode *msg, lrm_state_t *lrm_state) { xmlNode *data = do_lrm_query_internal(lrm_state, node_update_all); xmlNode *reply = create_reply(msg, data); if (relay_message(reply, TRUE) == FALSE) { crm_err("Unable to route reply"); crm_log_xml_err(reply, "reply"); } free_xml(reply); free_xml(data); } static void handle_reprobe_op(lrm_state_t *lrm_state, const char *from_sys, const char *from_host, const char *user_name, gboolean is_remote_node) { crm_notice("Forcing the status of all resources to be redetected"); force_reprobe(lrm_state, from_sys, from_host, user_name, is_remote_node); if (!pcmk__strcase_any_of(from_sys, CRM_SYSTEM_PENGINE, CRM_SYSTEM_TENGINE, NULL)) { xmlNode *reply = create_request(CRM_OP_INVOKE_LRM, NULL, from_host, from_sys, CRM_SYSTEM_LRMD, fsa_our_uuid); crm_debug("ACK'ing re-probe from %s (%s)", from_sys, from_host); if (relay_message(reply, TRUE) == FALSE) { crm_log_xml_err(reply, "Unable to route reply"); } free_xml(reply); } } static bool do_lrm_cancel(ha_msg_input_t *input, lrm_state_t *lrm_state, lrmd_rsc_info_t *rsc, const char *from_host, const char *from_sys) { char *op_key = NULL; char *meta_key = NULL; int call = 0; const char *call_id = NULL; const char *op_task = NULL; guint interval_ms = 0; gboolean in_progress = FALSE; xmlNode *params = find_xml_node(input->xml, XML_TAG_ATTRS, TRUE); CRM_CHECK(params != NULL, return FALSE); meta_key = crm_meta_name(XML_LRM_ATTR_TASK); op_task = crm_element_value(params, meta_key); free(meta_key); CRM_CHECK(op_task != NULL, return FALSE); meta_key = crm_meta_name(XML_LRM_ATTR_INTERVAL_MS); if (crm_element_value_ms(params, meta_key, &interval_ms) != pcmk_ok) { free(meta_key); return FALSE; } free(meta_key); op_key = pcmk__op_key(rsc->id, op_task, interval_ms); meta_key = crm_meta_name(XML_LRM_ATTR_CALLID); call_id = crm_element_value(params, meta_key); free(meta_key); crm_debug("Scheduler requested op %s (call=%s) be cancelled", op_key, (call_id? call_id : "NA")); pcmk__scan_min_int(call_id, &call, 0); if (call == 0) { // Normal case when the scheduler cancels a recurring op in_progress = cancel_op_key(lrm_state, rsc, op_key, TRUE); } else { // Normal case when the scheduler cancels an orphan op in_progress = cancel_op(lrm_state, rsc->id, NULL, call, TRUE); } // Acknowledge cancellation operation if for a remote connection resource if (!in_progress || is_remote_lrmd_ra(NULL, NULL, rsc->id)) { char *op_id = make_stop_id(rsc->id, call); if (is_remote_lrmd_ra(NULL, NULL, rsc->id) == FALSE) { crm_info("Nothing known about operation %d for %s", call, op_key); } erase_lrm_history_by_id(lrm_state, rsc->id, op_key, NULL, call); send_task_ok_ack(lrm_state, input, rsc->id, rsc, op_task, from_host, from_sys); /* needed at least for cancellation of a remote operation */ g_hash_table_remove(lrm_state->pending_ops, op_id); free(op_id); } else { /* No ack is needed since abcdaa8, but peers with older versions * in a rolling upgrade need one. We didn't bump the feature set * at that commit, so we can only compare against the previous * CRM version (3.0.8). If any peers have feature set 3.0.9 but * not abcdaa8, they will time out waiting for the ack (no * released versions of Pacemaker are affected). */ const char *peer_version = crm_element_value(params, XML_ATTR_CRM_VERSION); if (compare_version(peer_version, "3.0.8") <= 0) { crm_info("Sending compatibility ack for %s cancellation to %s (CRM version %s)", op_key, from_host, peer_version); send_task_ok_ack(lrm_state, input, rsc->id, rsc, op_task, from_host, from_sys); } } free(op_key); return TRUE; } static void do_lrm_delete(ha_msg_input_t *input, lrm_state_t *lrm_state, lrmd_rsc_info_t *rsc, const char *from_sys, const char *from_host, bool crm_rsc_delete, const char *user_name) { gboolean unregister = TRUE; int cib_rc = controld_delete_resource_history(rsc->id, lrm_state->node_name, user_name, cib_dryrun|cib_sync_call); if (cib_rc != pcmk_rc_ok) { lrmd_event_data_t *op = NULL; op = construct_op(lrm_state, input->xml, rsc->id, CRMD_ACTION_DELETE); /* These are resource clean-ups, not actions, so no exit reason is * needed. */ lrmd__set_result(op, pcmk_rc2ocf(cib_rc), PCMK_EXEC_ERROR, NULL); controld_ack_event_directly(from_host, from_sys, NULL, op, rsc->id); lrmd_free_event(op); return; } if (crm_rsc_delete && is_remote_lrmd_ra(NULL, NULL, rsc->id)) { unregister = FALSE; } delete_resource(lrm_state, rsc->id, rsc, NULL, from_sys, user_name, input, unregister); } /* A_LRM_INVOKE */ void do_lrm_invoke(long long action, enum crmd_fsa_cause cause, enum crmd_fsa_state cur_state, enum crmd_fsa_input current_input, fsa_data_t * msg_data) { lrm_state_t *lrm_state = NULL; const char *crm_op = NULL; const char *from_sys = NULL; const char *from_host = NULL; const char *operation = NULL; ha_msg_input_t *input = fsa_typed_data(fsa_dt_ha_msg); const char *user_name = NULL; const char *target_node = NULL; gboolean is_remote_node = FALSE; bool crm_rsc_delete = FALSE; target_node = lrm_op_target(input->xml); is_remote_node = !pcmk__str_eq(target_node, fsa_our_uname, pcmk__str_casei); lrm_state = lrm_state_find(target_node); if ((lrm_state == NULL) && is_remote_node) { crm_err("Failing action because local node has never had connection to remote node %s", target_node); synthesize_lrmd_failure(NULL, input->xml, PCMK_EXEC_NOT_CONNECTED, PCMK_OCF_UNKNOWN_ERROR, "Local node has no connection to remote"); return; } CRM_ASSERT(lrm_state != NULL); user_name = pcmk__update_acl_user(input->msg, F_CRM_USER, NULL); crm_op = crm_element_value(input->msg, F_CRM_TASK); from_sys = crm_element_value(input->msg, F_CRM_SYS_FROM); if (!pcmk__str_eq(from_sys, CRM_SYSTEM_TENGINE, pcmk__str_casei)) { from_host = crm_element_value(input->msg, F_CRM_HOST_FROM); } crm_trace("Executor %s command from %s as user %s", crm_op, from_sys, user_name); if (pcmk__str_eq(crm_op, CRM_OP_LRM_DELETE, pcmk__str_casei)) { if (!pcmk__str_eq(from_sys, CRM_SYSTEM_TENGINE, pcmk__str_casei)) { crm_rsc_delete = TRUE; // from crm_resource } operation = CRMD_ACTION_DELETE; } else if (pcmk__str_eq(crm_op, CRM_OP_LRM_FAIL, pcmk__str_casei)) { fail_lrm_resource(input->xml, lrm_state, user_name, from_host, from_sys); return; } else if (input->xml != NULL) { operation = crm_element_value(input->xml, XML_LRM_ATTR_TASK); } if (pcmk__str_eq(crm_op, CRM_OP_LRM_REFRESH, pcmk__str_casei)) { handle_refresh_op(lrm_state, user_name, from_host, from_sys); } else if (pcmk__str_eq(crm_op, CRM_OP_LRM_QUERY, pcmk__str_casei)) { handle_query_op(input->msg, lrm_state); } else if (pcmk__str_eq(operation, CRM_OP_PROBED, pcmk__str_casei)) { update_attrd(lrm_state->node_name, CRM_OP_PROBED, XML_BOOLEAN_TRUE, user_name, is_remote_node); } else if (pcmk__str_eq(crm_op, CRM_OP_REPROBE, pcmk__str_casei) || pcmk__str_eq(operation, CRM_OP_REPROBE, pcmk__str_casei)) { handle_reprobe_op(lrm_state, from_sys, from_host, user_name, is_remote_node); } else if (operation != NULL) { lrmd_rsc_info_t *rsc = NULL; xmlNode *xml_rsc = find_xml_node(input->xml, XML_CIB_TAG_RESOURCE, TRUE); gboolean create_rsc = !pcmk__str_eq(operation, CRMD_ACTION_DELETE, pcmk__str_casei); int rc; // We can't return anything meaningful without a resource ID CRM_CHECK(xml_rsc && ID(xml_rsc), return); rc = get_lrm_resource(lrm_state, xml_rsc, create_rsc, &rsc); if (rc == -ENOTCONN) { synthesize_lrmd_failure(lrm_state, input->xml, PCMK_EXEC_NOT_CONNECTED, PCMK_OCF_UNKNOWN_ERROR, "Not connected to remote executor"); return; } else if ((rc < 0) && !create_rsc) { /* Delete of malformed or nonexistent resource * (deleting something that does not exist is a success) */ crm_notice("Not registering resource '%s' for a %s event " CRM_XS " get-rc=%d (%s) transition-key=%s", ID(xml_rsc), operation, rc, pcmk_strerror(rc), ID(input->xml)); delete_rsc_entry(lrm_state, input, ID(xml_rsc), NULL, pcmk_ok, user_name); return; } else if (rc == -EINVAL) { // Resource operation on malformed resource crm_err("Invalid resource definition for %s", ID(xml_rsc)); crm_log_xml_warn(input->msg, "invalid resource"); synthesize_lrmd_failure(lrm_state, input->xml, PCMK_EXEC_ERROR, PCMK_OCF_NOT_CONFIGURED, // fatal error "Invalid resource definition"); return; } else if (rc < 0) { // Error communicating with the executor crm_err("Could not register resource '%s' with executor: %s " CRM_XS " rc=%d", ID(xml_rsc), pcmk_strerror(rc), rc); crm_log_xml_warn(input->msg, "failed registration"); synthesize_lrmd_failure(lrm_state, input->xml, PCMK_EXEC_ERROR, PCMK_OCF_INVALID_PARAM, // hard error "Could not register resource with executor"); return; } if (pcmk__str_eq(operation, CRMD_ACTION_CANCEL, pcmk__str_casei)) { if (!do_lrm_cancel(input, lrm_state, rsc, from_host, from_sys)) { crm_log_xml_warn(input->xml, "Bad command"); } } else if (pcmk__str_eq(operation, CRMD_ACTION_DELETE, pcmk__str_casei)) { do_lrm_delete(input, lrm_state, rsc, from_sys, from_host, crm_rsc_delete, user_name); } else if (pcmk__str_any_of(operation, CRMD_ACTION_RELOAD, CRMD_ACTION_RELOAD_AGENT, NULL)) { /* Pre-2.1.0 DCs will schedule reload actions only, and 2.1.0+ DCs * will schedule reload-agent actions only. In either case, we need * to map that to whatever the resource agent actually supports. * Default to the OCF 1.1 name. */ struct ra_metadata_s *md = NULL; const char *reload_name = CRMD_ACTION_RELOAD_AGENT; md = controld_get_rsc_metadata(lrm_state, rsc, controld_metadata_from_cache); if ((md != NULL) && pcmk_is_set(md->ra_flags, ra_supports_legacy_reload)) { reload_name = CRMD_ACTION_RELOAD; } do_lrm_rsc_op(lrm_state, rsc, reload_name, input->xml); } else { do_lrm_rsc_op(lrm_state, rsc, operation, input->xml); } lrmd_free_rsc_info(rsc); } else { crm_err("Cannot perform operation %s of unknown type", crm_str(crm_op)); register_fsa_error(C_FSA_INTERNAL, I_ERROR, NULL); } } #if ENABLE_VERSIONED_ATTRS static void resolve_versioned_parameters(lrm_state_t *lrm_state, const char *rsc_id, const xmlNode *rsc_op, GHashTable *params) { /* Resource info *should* already be cached, so we don't get * executor call */ lrmd_rsc_info_t *rsc = lrm_state_get_rsc_info(lrm_state, rsc_id, 0); struct ra_metadata_s *metadata; metadata = controld_get_rsc_metadata(lrm_state, rsc, controld_metadata_from_cache); if (metadata) { xmlNode *versioned_attrs = NULL; GHashTable *hash = NULL; char *key = NULL; char *value = NULL; GHashTableIter iter; versioned_attrs = first_named_child(rsc_op, XML_TAG_OP_VER_ATTRS); hash = pe_unpack_versioned_parameters(versioned_attrs, metadata->ra_version); g_hash_table_iter_init(&iter, hash); while (g_hash_table_iter_next(&iter, (gpointer *) &key, (gpointer *) &value)) { g_hash_table_iter_steal(&iter); g_hash_table_replace(params, key, value); } g_hash_table_destroy(hash); versioned_attrs = first_named_child(rsc_op, XML_TAG_OP_VER_META); hash = pe_unpack_versioned_parameters(versioned_attrs, metadata->ra_version); g_hash_table_iter_init(&iter, hash); while (g_hash_table_iter_next(&iter, (gpointer *) &key, (gpointer *) &value)) { g_hash_table_replace(params, crm_meta_name(key), strdup(value)); if (pcmk__str_eq(key, XML_ATTR_TIMEOUT, pcmk__str_casei)) { pcmk__scan_min_int(value, &op->timeout, 0); } else if (pcmk__str_eq(key, XML_OP_ATTR_START_DELAY, pcmk__str_casei)) { pcmk__scan_min_int(value, &op->start_delay, 0); } } g_hash_table_destroy(hash); versioned_attrs = first_named_child(rsc_op, XML_TAG_RSC_VER_ATTRS); hash = pe_unpack_versioned_parameters(versioned_attrs, metadata->ra_version); g_hash_table_iter_init(&iter, hash); while (g_hash_table_iter_next(&iter, (gpointer *) &key, (gpointer *) &value)) { g_hash_table_iter_steal(&iter); g_hash_table_replace(params, key, value); } g_hash_table_destroy(hash); } lrmd_free_rsc_info(rsc); } #endif static lrmd_event_data_t * construct_op(lrm_state_t *lrm_state, xmlNode *rsc_op, const char *rsc_id, const char *operation) { lrmd_event_data_t *op = NULL; const char *op_delay = NULL; const char *op_timeout = NULL; GHashTable *params = NULL; xmlNode *primitive = NULL; const char *class = NULL; const char *transition = NULL; CRM_ASSERT(rsc_id && operation); op = lrmd_new_event(rsc_id, operation, 0); op->type = lrmd_event_exec_complete; op->timeout = 0; op->start_delay = 0; lrmd__set_result(op, PCMK_OCF_UNKNOWN, PCMK_EXEC_PENDING, NULL); if (rsc_op == NULL) { CRM_LOG_ASSERT(pcmk__str_eq(CRMD_ACTION_STOP, operation, pcmk__str_casei)); op->user_data = NULL; /* the stop_all_resources() case * by definition there is no DC (or they'd be shutting * us down). * So we should put our version here. */ op->params = pcmk__strkey_table(free, free); g_hash_table_insert(op->params, strdup(XML_ATTR_CRM_VERSION), strdup(CRM_FEATURE_SET)); crm_trace("Constructed %s op for %s", operation, rsc_id); return op; } params = xml2list(rsc_op); g_hash_table_remove(params, CRM_META "_op_target_rc"); op_delay = crm_meta_value(params, XML_OP_ATTR_START_DELAY); pcmk__scan_min_int(op_delay, &op->start_delay, 0); op_timeout = crm_meta_value(params, XML_ATTR_TIMEOUT); pcmk__scan_min_int(op_timeout, &op->timeout, 0); if (pcmk__guint_from_hash(params, CRM_META "_" XML_LRM_ATTR_INTERVAL_MS, 0, &(op->interval_ms)) != pcmk_rc_ok) { op->interval_ms = 0; } /* Use pcmk_monitor_timeout instead of meta timeout for stonith recurring monitor, if set */ primitive = find_xml_node(rsc_op, XML_CIB_TAG_RESOURCE, FALSE); class = crm_element_value(primitive, XML_AGENT_ATTR_CLASS); if (pcmk_is_set(pcmk_get_ra_caps(class), pcmk_ra_cap_fence_params) && pcmk__str_eq(operation, CRMD_ACTION_STATUS, pcmk__str_casei) && (op->interval_ms > 0)) { op_timeout = g_hash_table_lookup(params, "pcmk_monitor_timeout"); if (op_timeout != NULL) { op->timeout = crm_get_msec(op_timeout); } } #if ENABLE_VERSIONED_ATTRS if (lrm_state && !is_remote_lrmd_ra(NULL, NULL, rsc_id) && !pcmk__strcase_any_of(op_type, CRMD_ACTION_METADATA, CRMD_ACTION_DELETE, NULL)) { resolve_versioned_parameters(lrm_state, rsc_id, rsc_op, params); } #endif if (!pcmk__str_eq(operation, RSC_STOP, pcmk__str_casei)) { op->params = params; } else { rsc_history_t *entry = NULL; if (lrm_state) { entry = g_hash_table_lookup(lrm_state->resource_history, rsc_id); } /* If we do not have stop parameters cached, use * whatever we are given */ if (!entry || !entry->stop_params) { op->params = params; } else { /* Copy the cached parameter list so that we stop the resource * with the old attributes, not the new ones */ op->params = pcmk__strkey_table(free, free); g_hash_table_foreach(params, copy_meta_keys, op->params); g_hash_table_foreach(entry->stop_params, copy_instance_keys, op->params); g_hash_table_destroy(params); params = NULL; } } /* sanity */ if (op->timeout <= 0) { op->timeout = op->interval_ms; } if (op->start_delay < 0) { op->start_delay = 0; } transition = crm_element_value(rsc_op, XML_ATTR_TRANSITION_KEY); CRM_CHECK(transition != NULL, return op); op->user_data = strdup(transition); if (op->interval_ms != 0) { if (pcmk__strcase_any_of(operation, CRMD_ACTION_START, CRMD_ACTION_STOP, NULL)) { crm_err("Start and Stop actions cannot have an interval: %u", op->interval_ms); op->interval_ms = 0; } } crm_trace("Constructed %s op for %s: interval=%u", operation, rsc_id, op->interval_ms); return op; } /*! * \internal * \brief Send a (synthesized) event result * * Reply with a synthesized event result directly, as opposed to going through * the executor. * * \param[in] to_host Host to send result to * \param[in] to_sys IPC name to send result to (NULL for transition engine) * \param[in] rsc Type information about resource the result is for * \param[in] op Event with result to send * \param[in] rsc_id ID of resource the result is for */ void controld_ack_event_directly(const char *to_host, const char *to_sys, lrmd_rsc_info_t *rsc, lrmd_event_data_t *op, const char *rsc_id) { xmlNode *reply = NULL; xmlNode *update, *iter; crm_node_t *peer = NULL; CRM_CHECK(op != NULL, return); if (op->rsc_id == NULL) { CRM_ASSERT(rsc_id != NULL); op->rsc_id = strdup(rsc_id); } if (to_sys == NULL) { to_sys = CRM_SYSTEM_TENGINE; } peer = crm_get_peer(0, fsa_our_uname); update = create_node_state_update(peer, node_update_none, NULL, __func__); iter = create_xml_node(update, XML_CIB_TAG_LRM); crm_xml_add(iter, XML_ATTR_ID, fsa_our_uuid); iter = create_xml_node(iter, XML_LRM_TAG_RESOURCES); iter = create_xml_node(iter, XML_LRM_TAG_RESOURCE); crm_xml_add(iter, XML_ATTR_ID, op->rsc_id); build_operation_update(iter, rsc, op, fsa_our_uname, __func__); reply = create_request(CRM_OP_INVOKE_LRM, update, to_host, to_sys, CRM_SYSTEM_LRMD, NULL); crm_log_xml_trace(update, "[direct ACK]"); crm_debug("ACK'ing resource op " PCMK__OP_FMT " from %s: %s", op->rsc_id, op->op_type, op->interval_ms, op->user_data, crm_element_value(reply, XML_ATTR_REFERENCE)); if (relay_message(reply, TRUE) == FALSE) { crm_log_xml_err(reply, "Unable to route reply"); } free_xml(update); free_xml(reply); } gboolean verify_stopped(enum crmd_fsa_state cur_state, int log_level) { gboolean res = TRUE; GList *lrm_state_list = lrm_state_get_list(); GList *state_entry; for (state_entry = lrm_state_list; state_entry != NULL; state_entry = state_entry->next) { lrm_state_t *lrm_state = state_entry->data; if (!lrm_state_verify_stopped(lrm_state, cur_state, log_level)) { /* keep iterating through all even when false is returned */ res = FALSE; } } controld_set_fsa_input_flags(R_SENT_RSC_STOP); g_list_free(lrm_state_list); lrm_state_list = NULL; return res; } struct stop_recurring_action_s { lrmd_rsc_info_t *rsc; lrm_state_t *lrm_state; }; static gboolean stop_recurring_action_by_rsc(gpointer key, gpointer value, gpointer user_data) { gboolean remove = FALSE; struct stop_recurring_action_s *event = user_data; active_op_t *op = value; if ((op->interval_ms != 0) && pcmk__str_eq(op->rsc_id, event->rsc->id, pcmk__str_none)) { crm_debug("Cancelling op %d for %s (%s)", op->call_id, op->rsc_id, (char*)key); remove = !cancel_op(event->lrm_state, event->rsc->id, key, op->call_id, FALSE); } return remove; } static gboolean stop_recurring_actions(gpointer key, gpointer value, gpointer user_data) { gboolean remove = FALSE; lrm_state_t *lrm_state = user_data; active_op_t *op = value; if (op->interval_ms != 0) { crm_info("Cancelling op %d for %s (%s)", op->call_id, op->rsc_id, (const char *) key); remove = !cancel_op(lrm_state, op->rsc_id, key, op->call_id, FALSE); } return remove; } static void record_pending_op(const char *node_name, lrmd_rsc_info_t *rsc, lrmd_event_data_t *op) { const char *record_pending = NULL; CRM_CHECK(node_name != NULL, return); CRM_CHECK(rsc != NULL, return); CRM_CHECK(op != NULL, return); // Never record certain operation types as pending if ((op->op_type == NULL) || (op->params == NULL) || !controld_action_is_recordable(op->op_type)) { return; } // defaults to true record_pending = crm_meta_value(op->params, XML_OP_ATTR_PENDING); if (record_pending && !crm_is_true(record_pending)) { return; } op->call_id = -1; lrmd__set_result(op, PCMK_OCF_UNKNOWN, PCMK_EXEC_PENDING, NULL); op->t_run = time(NULL); op->t_rcchange = op->t_run; /* write a "pending" entry to the CIB, inhibit notification */ crm_debug("Recording pending op " PCMK__OP_FMT " on %s in the CIB", op->rsc_id, op->op_type, op->interval_ms, node_name); do_update_resource(node_name, rsc, op, 0); } static void do_lrm_rsc_op(lrm_state_t *lrm_state, lrmd_rsc_info_t *rsc, const char *operation, xmlNode *msg) { int rc; int call_id = 0; char *op_id = NULL; lrmd_event_data_t *op = NULL; lrmd_key_value_t *params = NULL; fsa_data_t *msg_data = NULL; const char *transition = NULL; gboolean stop_recurring = FALSE; const char *nack_reason = NULL; CRM_CHECK(rsc != NULL, return); CRM_CHECK(operation != NULL, return); if (msg != NULL) { transition = crm_element_value(msg, XML_ATTR_TRANSITION_KEY); if (transition == NULL) { crm_log_xml_err(msg, "Missing transition number"); } } op = construct_op(lrm_state, msg, rsc->id, operation); CRM_CHECK(op != NULL, return); if (is_remote_lrmd_ra(NULL, NULL, rsc->id) && (op->interval_ms == 0) && strcmp(operation, CRMD_ACTION_MIGRATE) == 0) { /* pcmk remote connections are a special use case. * We never ever want to stop monitoring a connection resource until * the entire migration has completed. If the connection is unexpectedly * severed, even during a migration, this is an event we must detect.*/ stop_recurring = FALSE; } else if ((op->interval_ms == 0) && strcmp(operation, CRMD_ACTION_STATUS) != 0 && strcmp(operation, CRMD_ACTION_NOTIFY) != 0) { /* stop any previous monitor operations before changing the resource state */ stop_recurring = TRUE; } if (stop_recurring == TRUE) { guint removed = 0; struct stop_recurring_action_s data; data.rsc = rsc; data.lrm_state = lrm_state; removed = g_hash_table_foreach_remove( lrm_state->pending_ops, stop_recurring_action_by_rsc, &data); if (removed) { crm_debug("Stopped %u recurring operation%s in preparation for " PCMK__OP_FMT, removed, pcmk__plural_s(removed), rsc->id, operation, op->interval_ms); } } /* now do the op */ crm_notice("Requesting local execution of %s operation for %s on %s " CRM_XS " transition_key=%s op_key=" PCMK__OP_FMT, crm_action_str(op->op_type, op->interval_ms), rsc->id, lrm_state->node_name, transition, rsc->id, operation, op->interval_ms); if (pcmk_is_set(fsa_input_register, R_SHUTDOWN) && pcmk__str_eq(operation, RSC_START, pcmk__str_casei)) { register_fsa_input(C_SHUTDOWN, I_SHUTDOWN, NULL); nack_reason = "Not attempting start due to shutdown in progress"; } else if (fsa_state != S_NOT_DC && fsa_state != S_POLICY_ENGINE /* Recalculating */ && fsa_state != S_TRANSITION_ENGINE && !pcmk__str_eq(operation, CRMD_ACTION_STOP, pcmk__str_casei)) { nack_reason = "Controller cannot attempt actions at this time"; } if (nack_reason != NULL) { crm_notice("Discarding attempt to perform action %s on %s in state %s (shutdown=%s)", operation, rsc->id, fsa_state2string(fsa_state), pcmk__btoa(pcmk_is_set(fsa_input_register, R_SHUTDOWN))); lrmd__set_result(op, PCMK_OCF_UNKNOWN_ERROR, PCMK_EXEC_INVALID, nack_reason); controld_ack_event_directly(NULL, NULL, rsc, op, rsc->id); lrmd_free_event(op); free(op_id); return; } record_pending_op(lrm_state->node_name, rsc, op); op_id = pcmk__op_key(rsc->id, op->op_type, op->interval_ms); if (op->interval_ms > 0) { /* cancel it so we can then restart it without conflict */ cancel_op_key(lrm_state, rsc, op_id, FALSE); } if (op->params) { char *key = NULL; char *value = NULL; GHashTableIter iter; g_hash_table_iter_init(&iter, op->params); while (g_hash_table_iter_next(&iter, (gpointer *) & key, (gpointer *) & value)) { params = lrmd_key_value_add(params, key, value); } } rc = controld_execute_resource_agent(lrm_state, rsc->id, op->op_type, op->user_data, op->interval_ms, op->timeout, op->start_delay, params, &call_id); if (rc == pcmk_rc_ok) { /* record all operations so we can wait * for them to complete during shutdown */ char *call_id_s = make_stop_id(rsc->id, call_id); active_op_t *pending = NULL; pending = calloc(1, sizeof(active_op_t)); crm_trace("Recording pending op: %d - %s %s", call_id, op_id, call_id_s); pending->call_id = call_id; pending->interval_ms = op->interval_ms; pending->op_type = strdup(operation); pending->op_key = strdup(op_id); pending->rsc_id = strdup(rsc->id); pending->start_time = time(NULL); pending->user_data = op->user_data? strdup(op->user_data) : NULL; if (crm_element_value_epoch(msg, XML_CONFIG_ATTR_SHUTDOWN_LOCK, &(pending->lock_time)) != pcmk_ok) { pending->lock_time = 0; } g_hash_table_replace(lrm_state->pending_ops, call_id_s, pending); if ((op->interval_ms > 0) && (op->start_delay > START_DELAY_THRESHOLD)) { int target_rc = PCMK_OCF_OK; crm_info("Faking confirmation of %s: execution postponed for over 5 minutes", op_id); decode_transition_key(op->user_data, NULL, NULL, NULL, &target_rc); lrmd__set_result(op, target_rc, PCMK_EXEC_DONE, NULL); controld_ack_event_directly(NULL, NULL, rsc, op, rsc->id); } pending->params = op->params; op->params = NULL; } else if (lrm_state_is_local(lrm_state)) { crm_err("Could not initiate %s action for resource %s locally: %s " CRM_XS " rc=%d", operation, rsc->id, pcmk_rc_str(rc), rc); fake_op_status(lrm_state, op, PCMK_EXEC_NOT_CONNECTED, PCMK_OCF_UNKNOWN_ERROR, pcmk_rc_str(rc)); process_lrm_event(lrm_state, op, NULL, NULL); register_fsa_error(C_FSA_INTERNAL, I_FAIL, NULL); } else { crm_err("Could not initiate %s action for resource %s remotely on %s: " "%s " CRM_XS " rc=%d", operation, rsc->id, lrm_state->node_name, pcmk_rc_str(rc), rc); fake_op_status(lrm_state, op, PCMK_EXEC_NOT_CONNECTED, PCMK_OCF_UNKNOWN_ERROR, pcmk_rc_str(rc)); process_lrm_event(lrm_state, op, NULL, NULL); } free(op_id); lrmd_free_event(op); } int last_resource_update = 0; static void cib_rsc_callback(xmlNode * msg, int call_id, int rc, xmlNode * output, void *user_data) { switch (rc) { case pcmk_ok: case -pcmk_err_diff_failed: case -pcmk_err_diff_resync: crm_trace("Resource update %d complete: rc=%d", call_id, rc); break; default: crm_warn("Resource update %d failed: (rc=%d) %s", call_id, rc, pcmk_strerror(rc)); } if (call_id == last_resource_update) { last_resource_update = 0; trigger_fsa(); } } /* Only successful stops, and probes that found the resource inactive, get locks * recorded in the history. This ensures the resource stays locked to the node * until it is active there again after the node comes back up. */ static bool should_preserve_lock(lrmd_event_data_t *op) { if (!controld_shutdown_lock_enabled) { return false; } if (!strcmp(op->op_type, RSC_STOP) && (op->rc == PCMK_OCF_OK)) { return true; } if (!strcmp(op->op_type, RSC_STATUS) && (op->rc == PCMK_OCF_NOT_RUNNING)) { return true; } return false; } static int do_update_resource(const char *node_name, lrmd_rsc_info_t *rsc, lrmd_event_data_t *op, time_t lock_time) { /* */ int rc = pcmk_ok; xmlNode *update, *iter = NULL; int call_opt = crmd_cib_smart_opt(); const char *uuid = NULL; CRM_CHECK(op != NULL, return 0); iter = create_xml_node(iter, XML_CIB_TAG_STATUS); update = iter; iter = create_xml_node(iter, XML_CIB_TAG_STATE); if (pcmk__str_eq(node_name, fsa_our_uname, pcmk__str_casei)) { uuid = fsa_our_uuid; } else { /* remote nodes uuid and uname are equal */ uuid = node_name; crm_xml_add(iter, XML_NODE_IS_REMOTE, "true"); } CRM_LOG_ASSERT(uuid != NULL); if(uuid == NULL) { rc = -EINVAL; goto done; } crm_xml_add(iter, XML_ATTR_UUID, uuid); crm_xml_add(iter, XML_ATTR_UNAME, node_name); crm_xml_add(iter, XML_ATTR_ORIGIN, __func__); iter = create_xml_node(iter, XML_CIB_TAG_LRM); crm_xml_add(iter, XML_ATTR_ID, uuid); iter = create_xml_node(iter, XML_LRM_TAG_RESOURCES); iter = create_xml_node(iter, XML_LRM_TAG_RESOURCE); crm_xml_add(iter, XML_ATTR_ID, op->rsc_id); build_operation_update(iter, rsc, op, node_name, __func__); if (rsc) { const char *container = NULL; crm_xml_add(iter, XML_ATTR_TYPE, rsc->type); crm_xml_add(iter, XML_AGENT_ATTR_CLASS, rsc->standard); crm_xml_add(iter, XML_AGENT_ATTR_PROVIDER, rsc->provider); if (lock_time != 0) { /* Actions on a locked resource should either preserve the lock by * recording it with the action result, or clear it. */ if (!should_preserve_lock(op)) { lock_time = 0; } crm_xml_add_ll(iter, XML_CONFIG_ATTR_SHUTDOWN_LOCK, (long long) lock_time); } if (op->params) { container = g_hash_table_lookup(op->params, CRM_META"_"XML_RSC_ATTR_CONTAINER); } if (container) { crm_trace("Resource %s is a part of container resource %s", op->rsc_id, container); crm_xml_add(iter, XML_RSC_ATTR_CONTAINER, container); } } else { crm_warn("Resource %s no longer exists in the executor", op->rsc_id); controld_ack_event_directly(NULL, NULL, rsc, op, op->rsc_id); goto cleanup; } crm_log_xml_trace(update, __func__); /* make it an asynchronous call and be done with it * * Best case: * the resource state will be discovered during * the next signup or election. * * Bad case: * we are shutting down and there is no DC at the time, * but then why were we shutting down then anyway? * (probably because of an internal error) * * Worst case: * we get shot for having resources "running" that really weren't * * the alternative however means blocking here for too long, which * isn't acceptable */ fsa_cib_update(XML_CIB_TAG_STATUS, update, call_opt, rc, NULL); if (rc > 0) { last_resource_update = rc; } done: /* the return code is a call number, not an error code */ crm_trace("Sent resource state update message: %d for %s=%u on %s", rc, op->op_type, op->interval_ms, op->rsc_id); fsa_register_cib_callback(rc, FALSE, NULL, cib_rsc_callback); cleanup: free_xml(update); return rc; } void do_lrm_event(long long action, enum crmd_fsa_cause cause, enum crmd_fsa_state cur_state, enum crmd_fsa_input cur_input, fsa_data_t * msg_data) { CRM_CHECK(FALSE, return); } static char * unescape_newlines(const char *string) { char *pch = NULL; char *ret = NULL; static const char *escaped_newline = "\\n"; if (!string) { return NULL; } ret = strdup(string); pch = strstr(ret, escaped_newline); while (pch != NULL) { /* Replace newline escape pattern with actual newline (and a space so we * don't have to shuffle the rest of the buffer) */ pch[0] = '\n'; pch[1] = ' '; pch = strstr(pch, escaped_newline); } return ret; } static bool did_lrm_rsc_op_fail(lrm_state_t *lrm_state, const char * rsc_id, const char * op_type, guint interval_ms) { rsc_history_t *entry = NULL; CRM_CHECK(lrm_state != NULL, return FALSE); CRM_CHECK(rsc_id != NULL, return FALSE); CRM_CHECK(op_type != NULL, return FALSE); entry = g_hash_table_lookup(lrm_state->resource_history, rsc_id); if (entry == NULL || entry->failed == NULL) { return FALSE; } if (pcmk__str_eq(entry->failed->rsc_id, rsc_id, pcmk__str_none) && pcmk__str_eq(entry->failed->op_type, op_type, pcmk__str_casei) && entry->failed->interval_ms == interval_ms) { return TRUE; } return FALSE; } /*! * \internal * \brief Log the result of an executor action (actual or synthesized) * * \param[in] op Executor action to log result for * \param[in] op_key Operation key for action * \param[in] node_name Name of node action was performed on, if known * \param[in] update_id Call ID for CIB update (or 0 if none) * \param[in] confirmed Whether to log that graph action was confirmed */ static void log_executor_event(lrmd_event_data_t *op, const char *op_key, const char *node_name, int update_id, gboolean confirmed) { int log_level = LOG_ERR; GString *str = g_string_sized_new(100); // reasonable starting size g_string_printf(str, "Result of %s operation for %s", crm_action_str(op->op_type, op->interval_ms), op->rsc_id); if (node_name != NULL) { g_string_append_printf(str, " on %s", node_name); } switch (op->op_status) { case PCMK_EXEC_DONE: log_level = LOG_NOTICE; g_string_append_printf(str, ": %s", services_ocf_exitcode_str(op->rc)); break; case PCMK_EXEC_TIMEOUT: g_string_append_printf(str, ": %s after %s", pcmk_exec_status_str(op->op_status), pcmk__readable_interval(op->timeout)); break; case PCMK_EXEC_CANCELLED: log_level = LOG_INFO; // Fall through default: g_string_append_printf(str, ": %s", pcmk_exec_status_str(op->op_status)); } if ((op->exit_reason != NULL) && ((op->op_status != PCMK_EXEC_DONE) || (op->rc != PCMK_OCF_OK))) { g_string_append_printf(str, " (%s)", op->exit_reason); } g_string_append(str, " " CRM_XS); if (update_id != 0) { g_string_append_printf(str, " CIB update %d,", update_id); } g_string_append_printf(str, " graph action %sconfirmed; call=%d key=%s", (confirmed? "" : "un"), op->call_id, op_key); if (op->op_status == PCMK_EXEC_DONE) { g_string_append_printf(str, " rc=%d", op->rc); } do_crm_log(log_level, "%s", str->str); g_string_free(str, TRUE); - if (op->output != NULL) { - char *prefix = crm_strdup_printf("%s-" PCMK__OP_FMT ":%d", node_name, + /* The services library has already logged the output at info or debug + * level, so just raise to notice if it looks like a failure. + */ + if ((op->output != NULL) && (op->rc != PCMK_OCF_OK)) { + char *prefix = crm_strdup_printf(PCMK__OP_FMT "@%s output", op->rsc_id, op->op_type, - op->interval_ms, op->call_id); + op->interval_ms, node_name); - if (op->rc) { - crm_log_output(LOG_NOTICE, prefix, op->output); - } else { - crm_log_output(LOG_DEBUG, prefix, op->output); - } + crm_log_output(LOG_NOTICE, prefix, op->output); free(prefix); } } void process_lrm_event(lrm_state_t *lrm_state, lrmd_event_data_t *op, active_op_t *pending, xmlNode *action_xml) { char *op_id = NULL; char *op_key = NULL; int update_id = 0; gboolean remove = FALSE; gboolean removed = FALSE; bool need_direct_ack = FALSE; lrmd_rsc_info_t *rsc = NULL; const char *node_name = NULL; CRM_CHECK(op != NULL, return); CRM_CHECK(op->rsc_id != NULL, return); // Remap new status codes for older DCs if (compare_version(fsa_our_dc_version, "3.2.0") < 0) { switch (op->op_status) { case PCMK_EXEC_NOT_CONNECTED: lrmd__set_result(op, PCMK_OCF_CONNECTION_DIED, PCMK_EXEC_ERROR, op->exit_reason); break; case PCMK_EXEC_INVALID: lrmd__set_result(op, CRM_DIRECT_NACK_RC, PCMK_EXEC_ERROR, op->exit_reason); break; default: break; } } op_id = make_stop_id(op->rsc_id, op->call_id); op_key = pcmk__op_key(op->rsc_id, op->op_type, op->interval_ms); // Get resource info if available (from executor state or action XML) if (lrm_state) { rsc = lrm_state_get_rsc_info(lrm_state, op->rsc_id, 0); } if ((rsc == NULL) && action_xml) { xmlNode *xml = find_xml_node(action_xml, XML_CIB_TAG_RESOURCE, TRUE); const char *standard = crm_element_value(xml, XML_AGENT_ATTR_CLASS); const char *provider = crm_element_value(xml, XML_AGENT_ATTR_PROVIDER); const char *type = crm_element_value(xml, XML_ATTR_TYPE); if (standard && type) { crm_info("%s agent information not cached, using %s%s%s:%s from action XML", op->rsc_id, standard, (provider? ":" : ""), (provider? provider : ""), type); rsc = lrmd_new_rsc_info(op->rsc_id, standard, provider, type); } else { crm_err("Can't process %s result because %s agent information not cached or in XML", op_key, op->rsc_id); } } // Get node name if available (from executor state or action XML) if (lrm_state) { node_name = lrm_state->node_name; } else if (action_xml) { node_name = crm_element_value(action_xml, XML_LRM_ATTR_TARGET); } if(pending == NULL) { remove = TRUE; if (lrm_state) { pending = g_hash_table_lookup(lrm_state->pending_ops, op_id); } } if (op->op_status == PCMK_EXEC_ERROR) { switch(op->rc) { case PCMK_OCF_NOT_RUNNING: case PCMK_OCF_RUNNING_PROMOTED: case PCMK_OCF_DEGRADED: case PCMK_OCF_DEGRADED_PROMOTED: // Leave it to the TE/scheduler to decide if this is an error op->op_status = PCMK_EXEC_DONE; break; default: /* Nothing to do */ break; } } if (op->op_status != PCMK_EXEC_CANCELLED) { /* We might not record the result, so directly acknowledge it to the * originator instead, so it doesn't time out waiting for the result * (especially important if part of a transition). */ need_direct_ack = TRUE; if (controld_action_is_recordable(op->op_type)) { if (node_name && rsc) { // We should record the result, and happily, we can update_id = do_update_resource(node_name, rsc, op, pending? pending->lock_time : 0); need_direct_ack = FALSE; } else if (op->rsc_deleted) { /* We shouldn't record the result (likely the resource was * refreshed, cleaned, or removed while this operation was * in flight). */ crm_notice("Not recording %s result in CIB because " "resource information was removed since it was initiated", op_key); } else { /* This shouldn't be possible; the executor didn't consider the * resource deleted, but we couldn't find resource or node * information. */ crm_err("Unable to record %s result in CIB: %s", op_key, (node_name? "No resource information" : "No node name")); } } } else if (op->interval_ms == 0) { /* A non-recurring operation was cancelled. Most likely, the * never-initiated action was removed from the executor's pending * operations list upon resource removal. */ need_direct_ack = TRUE; } else if (pending == NULL) { /* This recurring operation was cancelled, but was not pending. No * transition actions are waiting on it, nothing needs to be done. */ } else if (op->user_data == NULL) { /* This recurring operation was cancelled and pending, but we don't * have a transition key. This should never happen. */ crm_err("Recurring operation %s was cancelled without transition information", op_key); } else if (pcmk_is_set(pending->flags, active_op_remove)) { /* This recurring operation was cancelled (by us) and pending, and we * have been waiting for it to finish. */ if (lrm_state) { erase_lrm_history_by_op(lrm_state, op); } /* If the recurring operation had failed, the lrm_rsc_op is recorded as * "last_failure" which won't get erased from the cib given the logic on * purpose in erase_lrm_history_by_op(). So that the cancel action won't * have a chance to get confirmed by DC with process_op_deletion(). * Cluster transition would get stuck waiting for the remaining action * timer to time out. * * Directly acknowledge the cancel operation in this case. */ if (did_lrm_rsc_op_fail(lrm_state, pending->rsc_id, pending->op_type, pending->interval_ms)) { need_direct_ack = TRUE; } } else if (op->rsc_deleted) { /* This recurring operation was cancelled (but not by us, and the * executor does not have resource information, likely due to resource * cleanup, refresh, or removal) and pending. */ crm_debug("Recurring op %s was cancelled due to resource deletion", op_key); need_direct_ack = TRUE; } else { /* This recurring operation was cancelled (but not by us, likely by the * executor before stopping the resource) and pending. We don't need to * do anything special. */ } if (need_direct_ack) { controld_ack_event_directly(NULL, NULL, NULL, op, op->rsc_id); } if(remove == FALSE) { /* The caller will do this afterwards, but keep the logging consistent */ removed = TRUE; } else if (lrm_state && ((op->interval_ms == 0) || (op->op_status == PCMK_EXEC_CANCELLED))) { gboolean found = g_hash_table_remove(lrm_state->pending_ops, op_id); if (op->interval_ms != 0) { removed = TRUE; } else if (found) { removed = TRUE; crm_trace("Op %s (call=%d, stop-id=%s, remaining=%u): Confirmed", op_key, op->call_id, op_id, g_hash_table_size(lrm_state->pending_ops)); } } log_executor_event(op, op_key, node_name, update_id, removed); if (lrm_state) { if (!pcmk__str_eq(op->op_type, RSC_METADATA, pcmk__str_casei)) { crmd_alert_resource_op(lrm_state->node_name, op); } else if (rsc && (op->rc == PCMK_OCF_OK)) { char *metadata = unescape_newlines(op->output); metadata_cache_update(lrm_state->metadata_cache, rsc, metadata); free(metadata); } } if (op->rsc_deleted) { crm_info("Deletion of resource '%s' complete after %s", op->rsc_id, op_key); if (lrm_state) { delete_rsc_entry(lrm_state, NULL, op->rsc_id, NULL, pcmk_ok, NULL); } } /* If a shutdown was escalated while operations were pending, * then the FSA will be stalled right now... allow it to continue */ mainloop_set_trigger(fsa_source); if (lrm_state && rsc) { update_history_cache(lrm_state, rsc, op); } lrmd_free_rsc_info(rsc); free(op_key); free(op_id); } diff --git a/daemons/fenced/fenced_commands.c b/daemons/fenced/fenced_commands.c index b5ae28d90c..9d06c68dc2 100644 --- a/daemons/fenced/fenced_commands.c +++ b/daemons/fenced/fenced_commands.c @@ -1,3162 +1,3232 @@ /* * Copyright 2009-2021 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include GHashTable *device_list = NULL; GHashTable *topology = NULL; GList *cmd_list = NULL; struct device_search_s { /* target of fence action */ char *host; /* requested fence action */ char *action; /* timeout to use if a device is queried dynamically for possible targets */ int per_device_timeout; /* number of registered fencing devices at time of request */ int replies_needed; /* number of device replies received so far */ int replies_received; /* whether the target is eligible to perform requested action (or off) */ bool allow_suicide; /* private data to pass to search callback function */ void *user_data; /* function to call when all replies have been received */ void (*callback) (GList * devices, void *user_data); /* devices capable of performing requested action (or off if remapping) */ GList *capable; }; static gboolean stonith_device_dispatch(gpointer user_data); -static void st_child_done(int pid, int rc, const char *output, void *user_data); +static void st_child_done(int pid, const pcmk__action_result_t *result, + void *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); static int get_agent_metadata(const char *agent, xmlNode **metadata); static void read_action_metadata(stonith_device_t *device); typedef struct async_command_s { int id; int pid; int fd_stdout; int options; int default_timeout; /* seconds */ int timeout; /* seconds */ int start_delay; /* seconds */ int delay_id; char *op; char *origin; char *client; char *client_name; char *remote_op_id; char *victim; uint32_t victim_nodeid; char *action; char *device; GList *device_list; GList *device_next; void *internal_user_data; - void (*done_cb) (int pid, int rc, const char *output, void *user_data); + void (*done_cb) (int pid, const pcmk__action_result_t *result, + void *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 xmlNode *construct_async_reply(async_command_t *cmd, + const pcmk__action_result_t *result); static gboolean is_action_required(const char *action, stonith_device_t *device) { return device && device->automatic_unfencing && pcmk__str_eq(action, "on", pcmk__str_casei); } static int get_action_delay_max(stonith_device_t * device, const char * action) { const char *value = NULL; int delay_max = 0; if (!pcmk__strcase_any_of(action, "off", "reboot", NULL)) { return 0; } value = g_hash_table_lookup(device->params, PCMK_STONITH_DELAY_MAX); if (value) { delay_max = crm_parse_interval_spec(value) / 1000; } return delay_max; } static int get_action_delay_base(stonith_device_t *device, const char *action, const char *victim) { char *hash_value = NULL; int delay_base = 0; if (!pcmk__strcase_any_of(action, "off", "reboot", NULL)) { return 0; } hash_value = g_hash_table_lookup(device->params, PCMK_STONITH_DELAY_BASE); if (hash_value) { char *value = strdup(hash_value); char *valptr = value; CRM_ASSERT(value != NULL); if (victim) { for (char *val = strtok(value, "; \t"); val != NULL; val = strtok(NULL, "; \t")) { char *mapval = strchr(val, ':'); if (mapval == NULL || mapval[1] == 0) { crm_err("pcmk_delay_base: empty value in mapping", val); continue; } if (mapval != val && strncasecmp(victim, val, (size_t)(mapval - val)) == 0) { value = mapval + 1; crm_debug("pcmk_delay_base mapped to %s for %s", value, victim); break; } } } if (strchr(value, ':') == 0) { delay_base = crm_parse_interval_spec(value) / 1000; } free(valptr); } return delay_base; } /*! * \internal * \brief Override STONITH timeout with pcmk_*_timeout if available * * \param[in] device STONITH device to use * \param[in] action STONITH action name * \param[in] default_timeout Timeout to use if device does not have * a pcmk_*_timeout parameter for action * * \return Value of pcmk_(action)_timeout if available, otherwise default_timeout * \note For consistency, it would be nice if reboot/off/on timeouts could be * set the same way as start/stop/monitor timeouts, i.e. with an * entry in the fencing resource configuration. However that * is insufficient because fencing devices may be registered directly via * the fencer's register_device() API instead of going through the CIB * (e.g. stonith_admin uses it for its -R option, and the executor uses it * to ensure a device is registered when a command is issued). As device * properties, pcmk_*_timeout parameters can be grabbed by the fencer when * the device is registered, whether by CIB change or API call. */ static int get_action_timeout(stonith_device_t * device, const char *action, int default_timeout) { if (action && device && device->params) { char buffer[64] = { 0, }; const char *value = NULL; /* If "reboot" was requested but the device does not support it, * we will remap to "off", so check timeout for "off" instead */ if (pcmk__str_eq(action, "reboot", pcmk__str_casei) && !pcmk_is_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->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; // Value -1 means disable any static/random fencing delays crm_element_value_int(msg, F_STONITH_DELAY, &(cmd->start_delay)); cmd->origin = crm_element_value_copy(msg, F_ORIG); cmd->remote_op_id = crm_element_value_copy(msg, F_STONITH_REMOTE_OP_ID); cmd->client = crm_element_value_copy(msg, F_STONITH_CLIENTID); cmd->client_name = crm_element_value_copy(msg, F_STONITH_CLIENTNAME); cmd->op = crm_element_value_copy(msg, F_STONITH_OPERATION); cmd->action = strdup(action); cmd->victim = crm_element_value_copy(op, F_STONITH_TARGET); cmd->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, PCMK_STONITH_ACTION_LIMIT); if ((value == NULL) || (pcmk__scan_min_int(value, &action_limit, INT_MIN) != pcmk_rc_ok) || (action_limit == 0)) { action_limit = 1; } return action_limit; } static int get_active_cmds(stonith_device_t * device) { int counter = 0; GList *gIter = NULL; GList *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(int pid, void *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' [%d]%s%s using %s now running with %ds timeout", cmd->action, pid, ((cmd->victim == NULL)? "" : " targeting "), ((cmd->victim == NULL)? "" : cmd->victim), device->id, cmd->timeout); cmd->active_on = device; cmd->activating_on = NULL; } static int get_agent_metadata_cb(gpointer data) { stonith_device_t *device = data; guint period_ms; switch (get_agent_metadata(device->agent, &device->agent_metadata)) { case pcmk_rc_ok: if (device->agent_metadata) { read_action_metadata(device); stonith__device_parameter_flags(&(device->flags), device->id, device->agent_metadata); } return G_SOURCE_REMOVE; case EAGAIN: period_ms = pcmk__mainloop_timer_get_period(device->timer); if (period_ms < 160 * 1000) { mainloop_timer_set_period(device->timer, 2 * period_ms); } return G_SOURCE_CONTINUE; default: return G_SOURCE_REMOVE; } } /*! * \internal * \brief Call a command's action callback for an internal (not library) result * - * \param[in] cmd Command to report result for - * \param[in] rc Legacy return code to pass to callback + * \param[in] cmd Command to report result for + * \param[in] execution_status Execution status to use for result + * \param[in] exit_status Exit status to use for result + * \param[in] exit_reason Exit reason to use for result */ static void -report_internal_result(async_command_t *cmd, int rc) +report_internal_result(async_command_t *cmd, int exit_status, + int execution_status, const char *exit_reason) { - cmd->done_cb(0, rc, NULL, cmd); + pcmk__action_result_t result = { + // Ensure we don't pass garbage to free() + .exit_reason = NULL, + .action_stdout = NULL, + .action_stderr = NULL + }; + + pcmk__set_result(&result, exit_status, execution_status, exit_reason); + cmd->done_cb(0, &result, cmd); + pcmk__reset_result(&result); } static gboolean stonith_device_execute(stonith_device_t * device) { int exec_rc = 0; const char *action_str = NULL; const char *host_arg = NULL; async_command_t *cmd = NULL; stonith_action_t *action = NULL; int active_cmds = 0; int action_limit = 0; GList *gIter = NULL; GList *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, pcmk__plural_s(active_cmds)); 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 using %s was asked to run too early, " "waiting for start delay of %ds", pending_op->action, ((pending_op->victim == NULL)? "" : " targeting "), ((pending_op->victim == NULL)? "" : 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("No actions using %s are needed", device->id); return TRUE; } if (pcmk__str_any_of(device->agent, STONITH_WATCHDOG_AGENT, STONITH_WATCHDOG_AGENT_INTERNAL, NULL)) { if (pcmk__strcase_any_of(cmd->action, "reboot", "off", NULL)) { if (node_does_watchdog_fencing(stonith_our_uname)) { pcmk__panic(__func__); goto done; } } else { crm_info("Faking success for %s watchdog operation", cmd->action); - report_internal_result(cmd, pcmk_ok); + report_internal_result(cmd, CRM_EX_OK, PCMK_EXEC_DONE, NULL); goto done; } } #if SUPPORT_CIBSECRETS exec_rc = pcmk__substitute_secrets(device->id, device->params); if (exec_rc != pcmk_rc_ok) { if (pcmk__str_eq(cmd->action, "stop", pcmk__str_casei)) { crm_info("Proceeding with stop operation for %s " "despite being unable to load CIB secrets (%s)", device->id, pcmk_rc_str(exec_rc)); } else { crm_err("Considering %s unconfigured " "because unable to load CIB secrets: %s", device->id, pcmk_rc_str(exec_rc)); - report_internal_result(cmd, -EACCES); + report_internal_result(cmd, CRM_EX_ERROR, PCMK_EXEC_NO_SECRETS, + "Failed to get CIB secrets"); goto done; } } #endif action_str = cmd->action; if (pcmk__str_eq(cmd->action, "reboot", pcmk__str_casei) && !pcmk_is_set(device->flags, st_device_supports_reboot)) { crm_notice("Remapping 'reboot' action%s%s using %s to 'off' " "because agent '%s' does not support reboot", ((cmd->victim == NULL)? "" : " targeting "), ((cmd->victim == NULL)? "" : cmd->victim), device->id, device->agent); action_str = "off"; } if (pcmk_is_set(device->flags, st_device_supports_parameter_port)) { host_arg = "port"; } else if (pcmk_is_set(device->flags, st_device_supports_parameter_plug)) { host_arg = "plug"; } action = stonith_action_create(device->agent, action_str, cmd->victim, cmd->victim_nodeid, cmd->timeout, device->params, device->aliases, host_arg); /* 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) { cmd->activating_on = NULL; - report_internal_result(cmd, exec_rc); + cmd->done_cb(0, stonith__action_result(action), cmd); stonith__destroy_action(action); } 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; int requested_delay = cmd->start_delay; CRM_CHECK(cmd != NULL, return); CRM_CHECK(device != NULL, return); if (cmd->device) { free(cmd->device); } if (device->include_nodeid && cmd->victim) { crm_node_t *node = crm_get_peer(0, cmd->victim); cmd->victim_nodeid = node->id; } cmd->device = strdup(device->id); cmd->timeout = get_action_timeout(device, cmd->action, cmd->default_timeout); if (cmd->remote_op_id) { crm_debug("Scheduling '%s' action%s%s using %s for remote peer %s " "with op id %.8s and timeout %ds", cmd->action, cmd->victim ? " targeting " : "", cmd->victim ? cmd->victim : "", device->id, cmd->origin, cmd->remote_op_id, cmd->timeout); } else { crm_debug("Scheduling '%s' action%s%s using %s for %s with 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); // Value -1 means disable any static/random fencing delays if (requested_delay < 0) { return; } delay_max = get_action_delay_max(device, cmd->action); delay_base = get_action_delay_base(device, cmd->action, cmd->victim); if (delay_max == 0) { delay_max = delay_base; } if (delay_max < delay_base) { crm_warn(PCMK_STONITH_DELAY_BASE " (%ds) is larger than " PCMK_STONITH_DELAY_MAX " (%ds) for %s using %s " "(limiting to maximum delay)", delay_base, delay_max, cmd->action, device->id); delay_base = delay_max; } if (delay_max > 0) { // coverity[dont_call] We're not using rand() for security cmd->start_delay += ((delay_max != delay_base)?(rand() % (delay_max - delay_base)):0) + delay_base; } if (cmd->start_delay > 0) { crm_notice("Delaying '%s' action%s%s using %s for %ds " CRM_XS " timeout=%ds requested_delay=%ds base=%ds max=%ds", cmd->action, cmd->victim ? " targeting " : "", cmd->victim ? cmd->victim : "", device->id, cmd->start_delay, cmd->timeout, requested_delay, delay_base, delay_max); cmd->delay_id = g_timeout_add_seconds(cmd->start_delay, start_delay_helper, cmd); } } static void free_device(gpointer data) { GList *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); - report_internal_result(cmd, -ENODEV); + report_internal_result(cmd, CRM_EX_ERROR, PCMK_EXEC_NO_FENCE_DEVICE, + "Device was removed before action could be executed"); } g_list_free(device->pending_ops); g_list_free_full(device->targets, free); if (device->timer) { mainloop_timer_stop(device->timer); mainloop_timer_del(device->timer); } 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(void) { if (device_list != NULL) { g_hash_table_destroy(device_list); device_list = NULL; } } void init_device_list(void) { if (device_list == NULL) { device_list = pcmk__strkey_table(NULL, free_device); } } static GHashTable * build_port_aliases(const char *hostmap, GList ** targets) { char *name = NULL; int last = 0, lpc = 0, max = 0, added = 0; GHashTable *aliases = pcmk__strikey_table(free, free); if (hostmap == NULL) { return aliases; } max = strlen(hostmap); for (; lpc <= max; lpc++) { switch (hostmap[lpc]) { /* Skip escaped chars */ case '\\': lpc++; break; /* 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; int k = 0; value = calloc(1, 1 + lpc - last); memcpy(value, hostmap + last, lpc - last); for (int i = 0; value[i] != '\0'; i++) { if (value[i] != '\\') { value[k++] = value[i]; } } value[k] = '\0'; 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(void) { if (metadata_cache != NULL) { g_hash_table_destroy(metadata_cache); metadata_cache = NULL; } } static void init_metadata_cache(void) { if (metadata_cache == NULL) { metadata_cache = pcmk__strkey_table(free, free); } } int get_agent_metadata(const char *agent, xmlNode ** metadata) { char *buffer = NULL; if (metadata == NULL) { return EINVAL; } *metadata = NULL; if (pcmk__str_eq(agent, STONITH_WATCHDOG_AGENT_INTERNAL, pcmk__str_none)) { return pcmk_rc_ok; } init_metadata_cache(); buffer = g_hash_table_lookup(metadata_cache, agent); 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 EAGAIN; } 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 EAGAIN; } g_hash_table_replace(metadata_cache, strdup(agent), buffer); } *metadata = string2xml(buffer); return pcmk_rc_ok; } 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(pcmk__str_eq(action, "list", pcmk__str_casei)) { stonith__set_device_flags(device->flags, device->id, st_device_supports_list); } else if(pcmk__str_eq(action, "status", pcmk__str_casei)) { stonith__set_device_flags(device->flags, device->id, st_device_supports_status); } else if(pcmk__str_eq(action, "reboot", pcmk__str_casei)) { stonith__set_device_flags(device->flags, device->id, st_device_supports_reboot); } else if (pcmk__str_eq(action, "on", pcmk__str_casei)) { /* "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 const char * target_list_type(stonith_device_t * dev) { const char *check_type = NULL; check_type = g_hash_table_lookup(dev->params, PCMK_STONITH_HOST_CHECK); if (check_type == NULL) { if (g_hash_table_lookup(dev->params, PCMK_STONITH_HOST_LIST)) { check_type = "static-list"; } else if (g_hash_table_lookup(dev->params, PCMK_STONITH_HOST_MAP)) { check_type = "static-list"; } else if (pcmk_is_set(dev->flags, st_device_supports_list)) { check_type = "dynamic-list"; } else if (pcmk_is_set(dev->flags, st_device_supports_status)) { check_type = "status"; } else { check_type = "none"; } } return check_type; } 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, PCMK_STONITH_HOST_LIST); if (value) { device->targets = stonith__parse_targets(value); } value = g_hash_table_lookup(device->params, PCMK_STONITH_HOST_MAP); device->aliases = build_port_aliases(value, &(device->targets)); value = target_list_type(device); if (!pcmk__str_eq(value, "static-list", pcmk__str_casei) && device->targets) { /* Other than "static-list", dev-> targets is unnecessary. */ g_list_free_full(device->targets, free); device->targets = NULL; } switch (get_agent_metadata(device->agent, &device->agent_metadata)) { case pcmk_rc_ok: if (device->agent_metadata) { read_action_metadata(device); stonith__device_parameter_flags(&(device->flags), device->id, device->agent_metadata); } break; case EAGAIN: if (device->timer == NULL) { device->timer = mainloop_timer_add("get_agent_metadata", 10 * 1000, TRUE, get_agent_metadata_cb, device); } if (!mainloop_timer_running(device->timer)) { mainloop_timer_start(device->timer); } break; default: break; } 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 (pcmk__str_eq(value, "unfencing", pcmk__str_casei)) { device->automatic_unfencing = TRUE; } if (is_action_required("on", device)) { crm_info("Fencing device '%s' requires unfencing", device->id); } if (device->on_target_actions) { crm_info("Fencing device '%s' requires actions (%s) to be executed " "on target", 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 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) (int pid, int rc, const char *output, + void (*done_cb) (int pid, + const pcmk__action_result_t *result, void *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); } // Fence agent status commands use custom exit status codes enum fence_status_code { fence_status_invalid = -1, fence_status_active = 0, fence_status_unknown = 1, fence_status_inactive = 2, }; static void -status_search_cb(int pid, int rc, const char *output, void *user_data) +status_search_cb(int pid, const pcmk__action_result_t *result, void *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); - switch (rc) { + if (result->execution_status != PCMK_EXEC_DONE) { + crm_warn("Assuming %s cannot fence %s " + "because status could not be executed: %s%s%s%s", + dev->id, search->host, + pcmk_exec_status_str(result->execution_status), + ((result->exit_reason == NULL)? "" : " ("), + ((result->exit_reason == NULL)? "" : result->exit_reason), + ((result->exit_reason == NULL)? "" : ")")); + search_devices_record_result(search, dev->id, FALSE); + return; + } + + switch (result->exit_status) { case fence_status_unknown: crm_trace("%s reported it cannot fence %s", dev->id, search->host); break; case fence_status_active: case fence_status_inactive: crm_trace("%s reported it can fence %s", dev->id, search->host); can = TRUE; break; default: crm_warn("Assuming %s cannot fence %s " "(status returned unknown code %d)", - dev->id, search->host, rc); + dev->id, search->host, result->exit_status); break; } search_devices_record_result(search, dev->id, can); } static void -dynamic_list_search_cb(int pid, int rc, const char *output, void *user_data) +dynamic_list_search_cb(int pid, const pcmk__action_result_t *result, + void *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 (rc == CRM_EX_OK) { + if ((result->execution_status == PCMK_EXEC_DONE) + && (result->exit_status == CRM_EX_OK)) { crm_info("Refreshing target list for %s", dev->id); g_list_free_full(dev->targets, free); - dev->targets = stonith__parse_targets(output); + dev->targets = stonith__parse_targets(result->action_stdout); dev->targets_age = time(NULL); } else if (dev->targets != NULL) { - crm_info("Reusing most recent target list for %s " - "because list returned error code %d", - dev->id, rc); + if (result->execution_status == PCMK_EXEC_DONE) { + crm_info("Reusing most recent target list for %s " + "because list returned error code %d", + dev->id, result->exit_status); + } else { + crm_info("Reusing most recent target list for %s " + "because list could not be executed: %s%s%s%s", + dev->id, pcmk_exec_status_str(result->execution_status), + ((result->exit_reason == NULL)? "" : " ("), + ((result->exit_reason == NULL)? "" : result->exit_reason), + ((result->exit_reason == NULL)? "" : ")")); + } } else { // We have never successfully executed list - crm_warn("Assuming %s cannot fence %s " - "because list returned error code %d", - dev->id, search->host, rc); + if (result->execution_status == PCMK_EXEC_DONE) { + crm_warn("Assuming %s cannot fence %s " + "because list returned error code %d", + dev->id, search->host, result->exit_status); + } else { + crm_warn("Assuming %s cannot fence %s " + "because list could not be executed: %s%s%s%s", + dev->id, search->host, + pcmk_exec_status_str(result->execution_status), + ((result->exit_reason == NULL)? "" : " ("), + ((result->exit_reason == NULL)? "" : result->exit_reason), + ((result->exit_reason == NULL)? "" : ")")); + } /* Fall back to pcmk_host_check="status" if the user didn't explicitly * specify "dynamic-list". */ if (g_hash_table_lookup(dev->params, PCMK_STONITH_HOST_CHECK) == NULL) { crm_notice("Switching to pcmk_host_check='status' for %s", dev->id); g_hash_table_replace(dev->params, strdup(PCMK_STONITH_HOST_CHECK), strdup("status")); } } if (dev->targets) { const char *alias = g_hash_table_lookup(dev->aliases, search->host); if (!alias) { alias = search->host; } if (pcmk__str_in_list(alias, dev->targets, pcmk__str_casei)) { 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 || !pcmk__str_eq(other_value, value, pcmk__str_casei)) { 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 (!pcmk__str_eq(dup->agent, device->agent, pcmk__str_casei)) { 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); guint ndevices = 0; int rv = pcmk_ok; CRM_CHECK(device != NULL, return -ENOMEM); /* do we have a watchdog-device? */ if (pcmk__str_eq(device->id, STONITH_WATCHDOG_ID, pcmk__str_none) || pcmk__str_any_of(device->agent, STONITH_WATCHDOG_AGENT, STONITH_WATCHDOG_AGENT_INTERNAL, NULL)) do { if (stonith_watchdog_timeout_ms <= 0) { crm_err("Ignoring watchdog fence device without " "stonith-watchdog-timeout set."); rv = -ENODEV; /* fall through to cleanup & return */ } else if (!pcmk__str_any_of(device->agent, STONITH_WATCHDOG_AGENT, STONITH_WATCHDOG_AGENT_INTERNAL, NULL)) { crm_err("Ignoring watchdog fence device with unknown " "agent '%s' unequal '" STONITH_WATCHDOG_AGENT "'.", device->agent?device->agent:""); rv = -ENODEV; /* fall through to cleanup & return */ } else if (!pcmk__str_eq(device->id, STONITH_WATCHDOG_ID, pcmk__str_none)) { crm_err("Ignoring watchdog fence device " "named %s !='"STONITH_WATCHDOG_ID"'.", device->id?device->id:""); rv = -ENODEV; /* fall through to cleanup & return */ } else { if (pcmk__str_eq(device->agent, STONITH_WATCHDOG_AGENT, pcmk__str_none)) { /* this either has an empty list or the targets configured for watchdog-fencing */ g_list_free_full(stonith_watchdog_targets, free); stonith_watchdog_targets = device->targets; device->targets = NULL; } if (node_does_watchdog_fencing(stonith_our_uname)) { g_list_free_full(device->targets, free); device->targets = stonith__parse_targets(stonith_our_uname); g_hash_table_replace(device->params, strdup(PCMK_STONITH_HOST_LIST), strdup(stonith_our_uname)); /* proceed as with any other stonith-device */ break; } crm_debug("Skip registration of watchdog fence device on node not in host-list."); /* cleanup and fall through to more cleanup and return */ device->targets = NULL; stonith_device_remove(device->id, from_cib); } free_device(device); return rv; } while (0); dup = device_has_duplicate(device); if (dup) { ndevices = g_hash_table_size(device_list); crm_debug("Device '%s' already in device list (%d active device%s)", device->id, ndevices, pcmk__plural_s(ndevices)); free_device(device); device = dup; dup = g_hash_table_lookup(device_list, device->id); dup->dirty = FALSE; } 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 existing entry for %s from 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); ndevices = g_hash_table_size(device_list); crm_notice("Added '%s' to device list (%d active device%s)", device->id, ndevices, pcmk__plural_s(ndevices)); } 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); guint ndevices = 0; if (!device) { ndevices = g_hash_table_size(device_list); crm_info("Device '%s' not found (%d active device%s)", id, ndevices, pcmk__plural_s(ndevices)); 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); ndevices = g_hash_table_size(device_list); crm_info("Removed '%s' from device list (%d active device%s)", id, ndevices, pcmk__plural_s(ndevices)); } else { crm_trace("Not removing '%s' from device list (%d active) because " "still registered via:%s%s", 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(void) { if (topology != NULL) { g_hash_table_destroy(topology); topology = NULL; } } void init_topology_list(void) { if (topology == NULL) { topology = pcmk__strkey_table(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 (pcmk__str_eq(TYPE(msg), XML_TAG_FENCING_LEVEL, pcmk__str_casei)) { 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); { int nlevels = count_active_levels(tp); crm_info("Target %s has %d active fencing level%s", tp->target, nlevels, pcmk__plural_s(nlevels)); } 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) { guint nentries = g_hash_table_size(topology); crm_info("No fencing topology found for %s (%d active %s)", target, nentries, pcmk__plural_alt(nentries, "entry", "entries")); } else if (id == 0 && g_hash_table_remove(topology, target)) { guint nentries = g_hash_table_size(topology); crm_info("Removed all fencing topology entries related to %s " "(%d active %s remaining)", target, nentries, pcmk__plural_alt(nentries, "entry", "entries")); } else if (id > 0 && tp->levels[id] != NULL) { guint nlevels; g_list_free_full(tp->levels[id], free); tp->levels[id] = NULL; nlevels = count_active_levels(tp); crm_info("Removed level %d from fencing topology for %s " "(%d active level%s remaining)", id, target, nlevels, pcmk__plural_s(nlevels)); } 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 char * list_to_string(GList *list, const char *delim, gboolean terminate_with_delim) { int max = g_list_length(list); size_t delim_len = delim?strlen(delim):0; size_t alloc_size = 1 + (max?((max-1+(terminate_with_delim?1:0))*delim_len):0); char *rv; GList *gIter; for (gIter = list; gIter != NULL; gIter = gIter->next) { const char *value = (const char *) gIter->data; alloc_size += strlen(value); } rv = calloc(alloc_size, sizeof(char)); if (rv) { char *pos = rv; const char *lead_delim = ""; for (gIter = list; gIter != NULL; gIter = gIter->next) { const char *value = (const char *) gIter->data; pos = &pos[sprintf(pos, "%s%s", lead_delim, value)]; lead_delim = delim; } if (max && terminate_with_delim) { sprintf(pos, "%s", delim); } } return rv; } 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; } if (pcmk__str_eq(id, STONITH_WATCHDOG_ID, pcmk__str_none)) { if (stonith_watchdog_timeout_ms <= 0) { return -ENODEV; } else { if (pcmk__str_eq(action, "list", pcmk__str_casei)) { *output = list_to_string(stonith_watchdog_targets, "\n", TRUE); return pcmk_ok; } else if (pcmk__str_eq(action, "monitor", pcmk__str_casei)) { return pcmk_ok; } } } device = g_hash_table_lookup(device_list, id); if ((device == NULL) || (!device->api_registered && !strcmp(action, "monitor"))) { // Monitors may run only on "started" (API-registered) devices crm_info("Ignoring API '%s' action request because device %s not found", action, id); return -ENODEV; } cmd = create_async_command(msg); if (cmd == NULL) { return -EPROTO; } schedule_stonith_command(cmd, device); return -EINPROGRESS; } static void search_devices_record_result(struct device_search_s *search, const char *device, gboolean can_fence) { search->replies_received++; if (can_fence && device) { search->capable = g_list_append(search->capable, strdup(device)); } if (search->replies_needed == search->replies_received) { guint ndevices = g_list_length(search->capable); crm_debug("Search found %d device%s that can perform '%s' targeting %s", ndevices, pcmk__plural_s(ndevices), (search->action? search->action : "unknown action"), (search->host? search->host : "any node")); 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 = pcmk__str_eq(target, stonith_our_uname, pcmk__str_casei); if (device && action && device->on_target_actions && strstr(device->on_target_actions, action)) { if (!localhost_is_target) { crm_trace("Operation '%s' using %s can only be executed for " "local host, not %s", action, device->id, target); return FALSE; } } else if (localhost_is_target && !allow_suicide) { crm_trace("'%s' operation does not support self-fencing", action); return FALSE; } return TRUE; } static void can_fence_host_with_device(stonith_device_t * dev, struct device_search_s *search) { gboolean can = FALSE; const char *check_type = NULL; const char *host = search->host; const char *alias = NULL; CRM_LOG_ASSERT(dev != NULL); if (dev == NULL) { goto search_report_results; } else if (host == NULL) { can = TRUE; goto search_report_results; } /* Short-circuit query if this host is not allowed to perform the action */ if (pcmk__str_eq(search->action, "reboot", pcmk__str_casei)) { /* 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 (pcmk__str_eq(check_type, "none", pcmk__str_casei)) { can = TRUE; } else if (pcmk__str_eq(check_type, "static-list", pcmk__str_casei)) { /* 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 (pcmk__str_in_list(host, dev->targets, pcmk__str_casei)) { can = TRUE; } else if (g_hash_table_lookup(dev->params, PCMK_STONITH_HOST_MAP) && g_hash_table_lookup(dev->aliases, host)) { can = TRUE; } } else if (pcmk__str_eq(check_type, "dynamic-list", pcmk__str_casei)) { time_t now = time(NULL); if (dev->targets == NULL || dev->targets_age + 60 < now) { int device_timeout = get_action_timeout(dev, "list", search->per_device_timeout); if (device_timeout > search->per_device_timeout) { crm_notice("Since the pcmk_list_timeout(%ds) parameter of %s is larger than stonith-timeout(%ds), timeout may occur", device_timeout, dev->id, search->per_device_timeout); } 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(__func__, 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 (pcmk__str_in_list(alias, dev->targets, pcmk__str_casei)) { can = TRUE; } } else if (pcmk__str_eq(check_type, "status", pcmk__str_casei)) { int device_timeout = get_action_timeout(dev, check_type, search->per_device_timeout); if (device_timeout > search->per_device_timeout) { crm_notice("Since the pcmk_status_timeout(%ds) parameter of %s is larger than stonith-timeout(%ds), timeout may occur", device_timeout, dev->id, search->per_device_timeout); } 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(__func__, 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 " PCMK_STONITH_HOST_CHECK ": %s", check_type); check_type = "Invalid " PCMK_STONITH_HOST_CHECK; } if (pcmk__str_eq(host, alias, pcmk__str_casei)) { 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; guint ndevices = g_hash_table_size(device_list); if (ndevices == 0) { callback(NULL, user_data); return; } search = calloc(1, sizeof(struct device_search_s)); if (!search) { crm_crit("Cannot search for capable fence devices: %s", strerror(ENOMEM)); callback(NULL, user_data); return; } search->host = host ? strdup(host) : NULL; search->action = action ? strdup(action) : NULL; search->per_device_timeout = timeout; search->allow_suicide = suicide; search->callback = callback; search->user_data = user_data; /* We are guaranteed this many replies, even if a device is * unregistered while the search is in progress. */ search->replies_needed = ndevices; crm_debug("Searching %d device%s to see which can execute '%s' targeting %s", ndevices, pcmk__plural_s(ndevices), (search->action? search->action : "unknown action"), (search->host? search->host : "any node")); 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 * \param[in] target Fence target */ static void add_action_specific_attributes(xmlNode *xml, const char *action, stonith_device_t *device, const char *target) { 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 using %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 using %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 using %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, target); 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 using %s", action, delay_max, device->id); } else if ((delay_max == 0) && (delay_base > 0)) { crm_trace("Action '%s' has a static delay of %dms using %s", action, delay_base, device->id); } else if ((delay_max > 0) && (delay_base > 0)) { crm_trace("Action '%s' has a minimum delay of %dms and a randomly chosen " "maximum delay of %dms using %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' using %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, target); 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; GList *lpc = NULL; /* Pack the results into XML */ list = create_xml_node(NULL, __func__); 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 (!pcmk_is_set(device->flags, st_device_supports_reboot) && pcmk__str_eq(query->action, "reboot", pcmk__str_casei)) { 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, query->target); if (pcmk__str_eq(query->action, "reboot", pcmk__str_casei)) { /* 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, pcmk_is_set(query->call_options, st_opt_allow_suicide)); add_action_reply(dev, "off", device, query->target, pcmk_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 device%s for target '%s'", available_devices, pcmk__plural_s(available_devices), query->target); } else { crm_debug("%d device%s installed", available_devices, pcmk__plural_s(available_devices)); } if (list != NULL) { crm_log_xml_trace(list, "Add query results"); add_message_xml(query->reply, F_STONITH_CALLDATA, list); } stonith_send_reply(query->reply, query->call_options, query->remote_peer, query->client_id); free_xml(query->reply); free(query->remote_peer); free(query->client_id); free(query->target); free(query->action); free(query); free_xml(list); g_list_free_full(devices, free); } static void stonith_query(xmlNode * msg, const char *remote_peer, const char *client_id, int call_options) { struct st_query_data *query = NULL; const char *action = NULL; const char *target = NULL; int timeout = 0; xmlNode *dev = get_xpath_object("//@" F_STONITH_ACTION, msg, LOG_NEVER); crm_element_value_int(msg, F_STONITH_TIMEOUT, &timeout); if (dev) { const char *device = crm_element_value(dev, F_STONITH_DEVICE); target = crm_element_value(dev, F_STONITH_TARGET); action = crm_element_value(dev, F_STONITH_ACTION); if (device && pcmk__str_eq(device, "manual_ack", pcmk__str_casei)) { /* 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, pcmk_is_set(call_options, st_opt_allow_suicide), query, stonith_query_capable_device_cb); } /*! * \internal * \brief Log the result of an asynchronous command * * \param[in] cmd Command the result is for - * \param[in] rc Legacy return code corresponding to result + * \param[in] result Result of command * \param[in] pid Process ID of command, if available * \param[in] next Alternate device that will be tried if command failed - * \param[in] output Command output, if any * \param[in] op_merged Whether this command was merged with an earlier one */ static void -log_async_result(async_command_t *cmd, int rc, int pid, const char *next, - const char *output, gboolean op_merged) +log_async_result(async_command_t *cmd, const pcmk__action_result_t *result, + int pid, const char *next, bool op_merged) { int log_level = LOG_ERR; int output_log_level = LOG_NEVER; guint devices_remaining = g_list_length(cmd->device_next); GString *msg = g_string_sized_new(80); // Reasonable starting size - // Choose log levels appropriately - if (rc == 0) { // Success + // Choose log levels appropriately if we have a result + if ((result->execution_status == PCMK_EXEC_DONE) + && (result->exit_status == CRM_EX_OK)) { // Success log_level = (cmd->victim == NULL)? LOG_DEBUG : LOG_NOTICE; - if ((output != NULL) + if ((result->action_stdout != NULL) && !pcmk__str_eq(cmd->action, "metadata", pcmk__str_casei)) { output_log_level = LOG_DEBUG; } next = NULL; } else { // Failure log_level = (cmd->victim == NULL)? LOG_NOTICE : LOG_ERR; - if ((output != NULL) + if ((result->action_stdout != NULL) && !pcmk__str_eq(cmd->action, "metadata", pcmk__str_casei)) { output_log_level = LOG_WARNING; } } // Build the log message piece by piece g_string_printf(msg, "Operation '%s' ", cmd->action); if (pid != 0) { g_string_append_printf(msg, "[%d] ", pid); } if (cmd->victim != NULL) { g_string_append_printf(msg, "targeting %s ", cmd->victim); } g_string_append_printf(msg, "using %s ", cmd->device); - // Add result - g_string_append_printf(msg, "returned %d (%s)", rc, pcmk_strerror(rc)); + // Add exit status or execution status as appropriate + if (result->execution_status == PCMK_EXEC_DONE) { + g_string_append_printf(msg, "returned %d", result->exit_status); + } else { + g_string_append_printf(msg, "could not be executed: %s", + pcmk_exec_status_str(result->execution_status)); + } - // Add next device if appropriate + // Add exit reason and next device if appropriate + if (result->exit_reason != NULL) { + g_string_append_printf(msg, " (%s)", result->exit_reason); + } if (next != NULL) { g_string_append_printf(msg, ", retrying with %s", next); } if (devices_remaining > 0) { g_string_append_printf(msg, " (%u device%s remaining)", (unsigned int) devices_remaining, pcmk__plural_s(devices_remaining)); } g_string_append_printf(msg, " " CRM_XS " %scall %d from %s", (op_merged? "merged " : ""), cmd->id, cmd->client_name); // Log the result do_crm_log(log_level, "%s", msg->str); g_string_free(msg, TRUE); // Log the output (which may have multiple lines), if appropriate if (output_log_level != LOG_NEVER) { char *prefix = crm_strdup_printf("%s[%d]", cmd->device, pid); - crm_log_output(output_log_level, prefix, output); + crm_log_output(output_log_level, prefix, result->action_stdout); free(prefix); } } +/*! + * \internal + * \brief Reply to requester after asynchronous command completion + * + * \param[in] cmd Command that completed + * \param[in] result Result of command + * \param[in] pid Process ID of command, if available + * \param[in] merged If true, command was merged with another, not executed + */ static void -stonith_send_async_reply(async_command_t *cmd, const char *output, int rc, - int pid, bool merged) +send_async_reply(async_command_t *cmd, const pcmk__action_result_t *result, + int pid, bool merged) { xmlNode *reply = NULL; gboolean bcast = FALSE; - reply = stonith_construct_async_reply(cmd, output, NULL, rc); + CRM_CHECK((cmd != NULL) && (result != NULL), return); + + reply = construct_async_reply(cmd, result); // Only replies for certain actions are broadcast if (pcmk__str_any_of(cmd->action, "metadata", "monitor", "list", "status", NULL)) { crm_trace("Never broadcast '%s' replies", cmd->action); } else if (!stand_alone && pcmk__str_eq(cmd->origin, cmd->victim, pcmk__str_casei) && !pcmk__str_eq(cmd->action, "on", pcmk__str_casei)) { crm_trace("Broadcast '%s' reply for %s", cmd->action, cmd->victim); crm_xml_add(reply, F_SUBTYPE, "broadcast"); bcast = TRUE; } - log_async_result(cmd, rc, pid, NULL, output, merged); + log_async_result(cmd, result, pid, NULL, merged); crm_log_xml_trace(reply, "Reply"); if (merged) { crm_xml_add(reply, F_STONITH_MERGED, "true"); } if (bcast) { crm_xml_add(reply, F_STONITH_OPERATION, T_STONITH_NOTIFY); send_cluster_message(NULL, crm_msg_stonith_ng, reply, FALSE); } else if (cmd->origin) { crm_trace("Directed reply to %s", cmd->origin); send_cluster_message(crm_get_peer(0, cmd->origin), crm_msg_stonith_ng, reply, FALSE); } else { crm_trace("Directed local %ssync reply to %s", (cmd->options & st_opt_sync_call) ? "" : "a-", cmd->client_name); do_local_reply(reply, cmd->client, cmd->options & st_opt_sync_call, FALSE); } if (stand_alone) { /* Do notification with a clean data object */ xmlNode *notify_data = create_xml_node(NULL, T_STONITH_NOTIFY_FENCE); + int rc = pcmk_rc2legacy(stonith__result2rc(result)); crm_xml_add_int(notify_data, F_STONITH_RC, rc); crm_xml_add(notify_data, F_STONITH_TARGET, cmd->victim); crm_xml_add(notify_data, F_STONITH_OPERATION, cmd->op); crm_xml_add(notify_data, F_STONITH_DELEGATE, "localhost"); crm_xml_add(notify_data, F_STONITH_DEVICE, cmd->device); crm_xml_add(notify_data, F_STONITH_REMOTE_OP_ID, cmd->remote_op_id); crm_xml_add(notify_data, F_STONITH_ORIGIN, cmd->client); do_stonith_notify(0, T_STONITH_NOTIFY_FENCE, rc, notify_data); do_stonith_notify(0, T_STONITH_NOTIFY_HISTORY, 0, NULL); } free_xml(reply); } static void cancel_stonith_command(async_command_t * cmd) { stonith_device_t *device; CRM_CHECK(cmd != NULL, return); if (!cmd->device) { return; } device = g_hash_table_lookup(device_list, cmd->device); if (device) { crm_trace("Cancel scheduled '%s' action using %s", cmd->action, device->id); device->pending_ops = g_list_remove(device->pending_ops, cmd); } } static void -st_child_done(int pid, int rc, const char *output, void *user_data) +st_child_done(int pid, const pcmk__action_result_t *result, void *user_data) { stonith_device_t *device = NULL; stonith_device_t *next_device = NULL; async_command_t *cmd = user_data; GList *gIter = NULL; GList *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) && + if (!device->verified && (result->exit_status == CRM_EX_OK) && (pcmk__strcase_any_of(cmd->action, "list", "monitor", "status", NULL))) { device->verified = TRUE; } mainloop_set_trigger(device->work); } - if (rc == 0) { + if (result->exit_status == CRM_EX_OK) { GList *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)) { + } else if ((cmd->device_next != NULL) + && !is_action_required(cmd->action, device)) { /* 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_async_result(cmd, rc, pid, next_device->id, output, FALSE); + log_async_result(cmd, result, pid, next_device->id, false); schedule_stonith_command(cmd, next_device); /* Prevent cmd from being freed */ cmd = NULL; goto done; } - stonith_send_async_reply(cmd, output, rc, pid, false); + send_async_reply(cmd, result, pid, false); - if (rc != 0) { + if (result->exit_status != CRM_EX_OK) { 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 (pcmk__str_eq(cmd->client, cmd_other->client, pcmk__str_casei) || !pcmk__str_eq(cmd->victim, cmd_other->victim, pcmk__str_casei) || !pcmk__str_eq(cmd->action, cmd_other->action, pcmk__str_casei) || !pcmk__str_eq(cmd->device, cmd_other->device, pcmk__str_casei)) { 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 fencing action '%s' targeting %s originating from " "client %s with identical fencing 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, true); + send_async_reply(cmd_other, result, pid, true); 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; guint ndevices = g_list_length(devices); crm_info("Found %d matching device%s for target '%s'", ndevices, pcmk__plural_s(ndevices), cmd->victim); if (devices != NULL) { /* Order based on priority */ devices = g_list_sort(devices, sort_device_priority); device = g_hash_table_lookup(device_list, devices->data); - - if (device) { - cmd->device_list = devices; - cmd->device_next = devices->next; - devices = NULL; /* list owned by cmd now */ - } } - /* we have a device, schedule it for fencing. */ - if (device) { - schedule_stonith_command(cmd, device); - /* in progress */ - return; - } + if (device == NULL) { // No device found + pcmk__action_result_t result = { + // Ensure we don't pass garbage to free() + .exit_reason = NULL, + .action_stdout = NULL, + .action_stderr = NULL + }; - /* no device found! */ - stonith_send_async_reply(cmd, NULL, -ENODEV, 0, false); + pcmk__set_result(&result, CRM_EX_ERROR, PCMK_EXEC_NO_FENCE_DEVICE, + "No fence device configured for target"); + send_async_reply(cmd, &result, 0, false); + pcmk__reset_result(&result); + free_async_command(cmd); + g_list_free_full(devices, free); - free_async_command(cmd); - g_list_free_full(devices, free); + } else { // Device found, schedule it for fencing + cmd->device_list = devices; + cmd->device_next = devices->next; + schedule_stonith_command(cmd, device); + } } 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_node_t *node; pcmk__scan_min_int(host, &nodeid, 0); node = pcmk__search_known_node_cache(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) { xmlNode *reply = NULL; reply = create_xml_node(NULL, T_STONITH_REPLY); crm_xml_add(reply, "st_origin", __func__); 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); if (request == NULL) { /* Most likely, this is the result of a stonith operation that was * initiated before we came up. Unfortunately that means we lack enough * information to provide clients with a full result. * * @TODO Maybe synchronize this information at start-up? */ crm_warn("Missing request information for client notifications for " "operation with result %d (initiated before we came up?)", rc); } else { 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 result reply with%s reply output (rc=%d)", (data? "" : "out"), rc); for (int lpc = 0; lpc < PCMK__NELEM(names); lpc++) { name = names[lpc]; value = crm_element_value(request, name); crm_xml_add(reply, name, value); } if (data != NULL) { add_message_xml(reply, F_STONITH_CALLDATA, data); } } return reply; } +/*! + * \internal + * \brief Build an XML reply to an asynchronous fencing command + * + * \param[in] cmd Fencing command that reply is for + * \param[in] result Command result + */ static xmlNode * -stonith_construct_async_reply(async_command_t * cmd, const char *output, xmlNode * data, int rc) +construct_async_reply(async_command_t *cmd, const pcmk__action_result_t *result) { - xmlNode *reply = NULL; - - crm_trace("Creating a basic reply"); - reply = create_xml_node(NULL, T_STONITH_REPLY); + xmlNode *reply = create_xml_node(NULL, T_STONITH_REPLY); crm_xml_add(reply, "st_origin", __func__); 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); - } + crm_xml_add_int(reply, F_STONITH_RC, + pcmk_rc2legacy(stonith__result2rc(result))); + crm_xml_add(reply, "st_output", result->action_stdout); return reply; } bool fencing_peer_active(crm_node_t *peer) { if (peer == NULL) { return FALSE; } else if (peer->uname == NULL) { return FALSE; } else if (pcmk_is_set(peer->processes, crm_get_cluster_proc())) { return TRUE; } return FALSE; } void set_fencing_completed(remote_fencing_op_t * op) { #ifdef CLOCK_MONOTONIC struct timespec tv; clock_gettime(CLOCK_MONOTONIC, &tv); op->completed = tv.tv_sec; op->completed_nsec = tv.tv_nsec; #else op->completed = time(NULL); op->completed_nsec = 0L; #endif } /*! * \internal * \brief Look for alternate node needed if local node shouldn't fence target * * \param[in] target Node that must be fenced * * \return Name of an alternate node that should fence \p target if any, * or NULL otherwise */ 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) && pcmk__str_eq(target, stonith_our_uname, pcmk__str_casei)) { 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) && !pcmk__str_eq(entry->uname, target, pcmk__str_casei)) { alternate_host = entry->uname; break; } } if (alternate_host == NULL) { crm_err("No alternate host available to handle request " "for self-fencing with topology"); 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, pcmk_is_set(call_options, st_opt_sync_call), (remote_peer != NULL)); } } static void remove_relay_op(xmlNode * request) { xmlNode *dev = get_xpath_object("//@" F_STONITH_ACTION, request, LOG_TRACE); const char *relay_op_id = NULL; const char *op_id = NULL; const char *client_name = NULL; const char *target = NULL; remote_fencing_op_t *relay_op = NULL; if (dev) { target = crm_element_value(dev, F_STONITH_TARGET); } relay_op_id = crm_element_value(request, F_STONITH_REMOTE_OP_ID_RELAY); op_id = crm_element_value(request, F_STONITH_REMOTE_OP_ID); client_name = crm_element_value(request, F_STONITH_CLIENTNAME); /* Delete RELAY operation. */ if (relay_op_id && target && pcmk__str_eq(target, stonith_our_uname, pcmk__str_casei)) { relay_op = g_hash_table_lookup(stonith_remote_op_list, relay_op_id); if (relay_op) { GHashTableIter iter; remote_fencing_op_t *list_op = NULL; g_hash_table_iter_init(&iter, stonith_remote_op_list); /* If the operation to be deleted is registered as a duplicate, delete the registration. */ while (g_hash_table_iter_next(&iter, NULL, (void **)&list_op)) { GList *dup_iter = NULL; if (list_op != relay_op) { for (dup_iter = list_op->duplicates; dup_iter != NULL; dup_iter = dup_iter->next) { remote_fencing_op_t *other = dup_iter->data; if (other == relay_op) { other->duplicates = g_list_remove(other->duplicates, relay_op); break; } } } } crm_debug("Deleting relay op %s ('%s' targeting %s for %s), " "replaced by op %s ('%s' targeting %s for %s)", relay_op->id, relay_op->action, relay_op->target, relay_op->client_name, op_id, relay_op->action, target, client_name); g_hash_table_remove(stonith_remote_op_list, relay_op_id); } } } static int handle_request(pcmk__client_t *client, uint32_t id, uint32_t flags, xmlNode *request, const char *remote_peer) { int call_options = 0; int rc = -EOPNOTSUPP; xmlNode *data = NULL; xmlNode *reply = NULL; char *output = NULL; const char *op = crm_element_value(request, F_STONITH_OPERATION); const char *client_id = crm_element_value(request, F_STONITH_CLIENTID); /* IPC commands related to fencing configuration may be done only by * privileged users (i.e. root or hacluster), because all other users should * go through the CIB to have ACLs applied. * * If no client was given, this is a peer request, which is always allowed. */ bool allowed = (client == NULL) || pcmk_is_set(client->flags, pcmk__client_privileged); crm_element_value_int(request, F_STONITH_CALLOPTS, &call_options); if (pcmk_is_set(call_options, st_opt_sync_call)) { CRM_ASSERT(client == NULL || client->request_id == id); } if (pcmk__str_eq(op, CRM_OP_REGISTER, pcmk__str_none)) { xmlNode *reply = create_xml_node(NULL, "reply"); CRM_ASSERT(client); crm_xml_add(reply, F_STONITH_OPERATION, CRM_OP_REGISTER); crm_xml_add(reply, F_STONITH_CLIENTID, client->id); pcmk__ipc_send_xml(client, id, reply, flags); client->request_id = 0; free_xml(reply); return 0; } else if (pcmk__str_eq(op, STONITH_OP_EXEC, pcmk__str_none)) { rc = stonith_device_action(request, &output); } else if (pcmk__str_eq(op, STONITH_OP_TIMEOUT_UPDATE, pcmk__str_none)) { 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 (pcmk__str_eq(op, STONITH_OP_QUERY, pcmk__str_none)) { if (remote_peer) { create_remote_stonith_op(client_id, request, TRUE); /* Record it for the future notification */ } /* Delete the DC node RELAY operation. */ remove_relay_op(request); stonith_query(request, remote_peer, client_id, call_options); return 0; } else if (pcmk__str_eq(op, T_STONITH_NOTIFY, pcmk__str_none)) { const char *flag_name = NULL; CRM_ASSERT(client); flag_name = crm_element_value(request, F_STONITH_NOTIFY_ACTIVATE); if (flag_name) { crm_debug("Enabling %s callbacks for client %s", flag_name, pcmk__client_name(client)); pcmk__set_client_flags(client, get_stonith_flag(flag_name)); } flag_name = crm_element_value(request, F_STONITH_NOTIFY_DEACTIVATE); if (flag_name) { crm_debug("Disabling %s callbacks for client %s", flag_name, pcmk__client_name(client)); pcmk__clear_client_flags(client, get_stonith_flag(flag_name)); } pcmk__ipc_send_ack(client, id, flags, "ack", CRM_EX_OK); return 0; } else if (pcmk__str_eq(op, STONITH_OP_RELAY, pcmk__str_none)) { xmlNode *dev = get_xpath_object("//@" F_STONITH_TARGET, request, LOG_TRACE); crm_notice("Received forwarded fencing request from " "%s %s to fence (%s) peer %s", ((client == NULL)? "peer" : "client"), ((client == NULL)? remote_peer : pcmk__client_name(client)), 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 (pcmk__str_eq(op, STONITH_OP_FENCE, pcmk__str_none)) { 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 wants to fence (%s) %s using %s", pcmk__client_name(client), action, target, (device? device : "any device")); 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; remote_fencing_op_t *op = NULL; crm_notice("Forwarding self-fencing request to peer %s" "due to topology", alternate_host); if (client->id) { client_id = client->id; } else { client_id = crm_element_value(request, F_STONITH_CLIENTID); } /* Create an operation for RELAY and send the ID in the RELAY message. */ /* When a QUERY response is received, delete the RELAY operation to avoid the existence of duplicate operations. */ op = 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); crm_xml_add(request, F_STONITH_REMOTE_OP_ID, op->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 (pcmk__str_eq(op, STONITH_OP_FENCE_HISTORY, pcmk__str_none)) { 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 (pcmk__str_eq(op, STONITH_OP_DEVICE_ADD, pcmk__str_none)) { const char *device_id = NULL; if (allowed) { rc = stonith_device_register(request, &device_id, FALSE); } else { rc = -EACCES; } do_stonith_notify_device(call_options, op, rc, device_id); } else if (pcmk__str_eq(op, STONITH_OP_DEVICE_DEL, pcmk__str_none)) { xmlNode *dev = get_xpath_object("//" F_STONITH_DEVICE, request, LOG_ERR); const char *device_id = crm_element_value(dev, XML_ATTR_ID); if (allowed) { rc = stonith_device_remove(device_id, FALSE); } else { rc = -EACCES; } do_stonith_notify_device(call_options, op, rc, device_id); } else if (pcmk__str_eq(op, STONITH_OP_LEVEL_ADD, pcmk__str_none)) { char *device_id = NULL; if (allowed) { rc = stonith_level_register(request, &device_id); } else { rc = -EACCES; } do_stonith_notify_level(call_options, op, rc, device_id); free(device_id); } else if (pcmk__str_eq(op, STONITH_OP_LEVEL_DEL, pcmk__str_none)) { char *device_id = NULL; if (allowed) { rc = stonith_level_remove(request, &device_id); } else { rc = -EACCES; } do_stonith_notify_level(call_options, op, rc, device_id); } else if(pcmk__str_eq(op, CRM_OP_RM_NODE_CACHE, pcmk__str_casei)) { int node_id = 0; const char *name = NULL; crm_element_value_int(request, XML_ATTR_ID, &node_id); name = crm_element_value(request, XML_ATTR_UNAME); reap_crm_member(node_id, name); return pcmk_ok; } else { crm_err("Unknown IPC request %s from %s %s", op, ((client == NULL)? "peer" : "client"), ((client == NULL)? remote_peer : pcmk__client_name(client))); } done: if (rc == -EACCES) { crm_warn("Rejecting IPC request '%s' from unprivileged client %s", crm_str(op), pcmk__client_name(client)); } /* 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, pcmk_is_set(call_options, st_opt_sync_call), call_options, crm_element_value(request, F_STONITH_CALLOPTS)); if (pcmk_is_set(call_options, st_opt_sync_call)) { CRM_ASSERT(client == NULL || client->request_id == id); } reply = stonith_construct_reply(request, output, data, rc); stonith_send_reply(reply, call_options, remote_peer, client_id); } free(output); free_xml(data); free_xml(reply); return rc; } static void handle_reply(pcmk__client_t *client, xmlNode *request, const char *remote_peer) { const char *op = crm_element_value(request, F_STONITH_OPERATION); if (pcmk__str_eq(op, STONITH_OP_QUERY, pcmk__str_none)) { process_remote_stonith_query(request); } else if (pcmk__str_eq(op, T_STONITH_NOTIFY, pcmk__str_none)) { process_remote_stonith_exec(request); } else if (pcmk__str_eq(op, STONITH_OP_FENCE, pcmk__str_none)) { /* Reply to a complex fencing op */ process_remote_stonith_exec(request); } else { crm_err("Unknown %s reply from %s %s", op, ((client == NULL)? "peer" : "client"), ((client == NULL)? remote_peer : pcmk__client_name(client))); crm_log_xml_warn(request, "UnknownOp"); } } void stonith_command(pcmk__client_t *client, uint32_t id, uint32_t flags, xmlNode *request, const char *remote_peer) { int call_options = 0; int rc = 0; gboolean is_reply = FALSE; /* Copy op for reporting. The original might get freed by handle_reply() * before we use it in crm_debug(): * handle_reply() * |- process_remote_stonith_exec() * |-- remote_op_done() * |--- handle_local_reply_and_notify() * |---- crm_xml_add(...F_STONITH_OPERATION...) * |--- free_xml(op->request) */ char *op = crm_element_value_copy(request, F_STONITH_OPERATION); if (get_xpath_object("//" T_STONITH_REPLY, request, LOG_NEVER)) { is_reply = TRUE; } crm_element_value_int(request, F_STONITH_CALLOPTS, &call_options); crm_debug("Processing %s%s %u from %s %s with call options 0x%08x", op, (is_reply? " reply" : ""), id, ((client == NULL)? "peer" : "client"), ((client == NULL)? remote_peer : pcmk__client_name(client)), call_options); if (pcmk_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: %s (rc=%d)", op, (is_reply? " reply" : ""), ((client == NULL)? "peer" : "client"), ((client == NULL)? remote_peer : pcmk__client_name(client)), ((rc > 0)? "" : pcmk_strerror(rc)), rc); free(op); } diff --git a/daemons/fenced/fenced_remote.c b/daemons/fenced/fenced_remote.c index ffaf60018a..b09d2865e3 100644 --- a/daemons/fenced/fenced_remote.c +++ b/daemons/fenced/fenced_remote.c @@ -1,2211 +1,2211 @@ /* * Copyright 2009-2021 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU General Public License version 2 * or later (GPLv2+) WITHOUT ANY WARRANTY. */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #define TIMEOUT_MULTIPLY_FACTOR 1.2 /* When one fencer queries its peers for devices able to handle a fencing * request, each peer will reply with a list of such devices available to it. * Each reply will be parsed into a st_query_result_t, with each device's * information kept in a device_properties_t. */ typedef struct device_properties_s { /* Whether access to this device has been verified */ gboolean verified; /* The remaining members are indexed by the operation's "phase" */ /* Whether this device has been executed in each phase */ gboolean executed[st_phase_max]; /* Whether this device is disallowed from executing in each phase */ gboolean disallowed[st_phase_max]; /* Action-specific timeout for each phase */ int custom_action_timeout[st_phase_max]; /* Action-specific maximum random delay for each phase */ int delay_max[st_phase_max]; /* Action-specific base delay for each phase */ int delay_base[st_phase_max]; } device_properties_t; typedef struct st_query_result_s { /* Name of peer that sent this result */ char *host; /* Only try peers for non-topology based operations once */ gboolean tried; /* Number of entries in the devices table */ int ndevices; /* Devices available to this host that are capable of fencing the target */ GHashTable *devices; } st_query_result_t; GHashTable *stonith_remote_op_list = NULL; void call_remote_stonith(remote_fencing_op_t * op, st_query_result_t * peer, int rc); static void remote_op_done(remote_fencing_op_t * op, xmlNode * data, int rc, int dup); extern xmlNode *stonith_create_op(int call_id, const char *token, const char *op, xmlNode * data, int call_options); static void report_timeout_period(remote_fencing_op_t * op, int op_timeout); static int get_op_total_timeout(const remote_fencing_op_t *op, const st_query_result_t *chosen_peer); static gint sort_strings(gconstpointer a, gconstpointer b) { return strcmp(a, b); } static void free_remote_query(gpointer data) { if (data) { st_query_result_t *query = data; crm_trace("Free'ing query result from %s", query->host); g_hash_table_destroy(query->devices); free(query->host); free(query); } } void free_stonith_remote_op_list() { if (stonith_remote_op_list != NULL) { g_hash_table_destroy(stonith_remote_op_list); stonith_remote_op_list = NULL; } } struct peer_count_data { const remote_fencing_op_t *op; gboolean verified_only; int count; }; /*! * \internal * \brief Increment a counter if a device has not been executed yet * * \param[in] key Device ID (ignored) * \param[in] value Device properties * \param[in] user_data Peer count data */ static void count_peer_device(gpointer key, gpointer value, gpointer user_data) { device_properties_t *props = (device_properties_t*)value; struct peer_count_data *data = user_data; if (!props->executed[data->op->phase] && (!data->verified_only || props->verified)) { ++(data->count); } } /*! * \internal * \brief Check the number of available devices in a peer's query results * * \param[in] op Operation that results are for * \param[in] peer Peer to count * \param[in] verified_only Whether to count only verified devices * * \return Number of devices available to peer that were not already executed */ static int count_peer_devices(const remote_fencing_op_t *op, const st_query_result_t *peer, gboolean verified_only) { struct peer_count_data data; data.op = op; data.verified_only = verified_only; data.count = 0; if (peer) { g_hash_table_foreach(peer->devices, count_peer_device, &data); } return data.count; } /*! * \internal * \brief Search for a device in a query result * * \param[in] op Operation that result is for * \param[in] peer Query result for a peer * \param[in] device Device ID to search for * * \return Device properties if found, NULL otherwise */ static device_properties_t * find_peer_device(const remote_fencing_op_t *op, const st_query_result_t *peer, const char *device) { device_properties_t *props = g_hash_table_lookup(peer->devices, device); return (props && !props->executed[op->phase] && !props->disallowed[op->phase])? props : NULL; } /*! * \internal * \brief Find a device in a peer's device list and mark it as executed * * \param[in] op Operation that peer result is for * \param[in,out] peer Peer with results to search * \param[in] device ID of device to mark as done * \param[in] verified_devices_only Only consider verified devices * * \return TRUE if device was found and marked, FALSE otherwise */ static gboolean grab_peer_device(const remote_fencing_op_t *op, st_query_result_t *peer, const char *device, gboolean verified_devices_only) { device_properties_t *props = find_peer_device(op, peer, device); if ((props == NULL) || (verified_devices_only && !props->verified)) { return FALSE; } crm_trace("Removing %s from %s (%d remaining)", device, peer->host, count_peer_devices(op, peer, FALSE)); props->executed[op->phase] = TRUE; return TRUE; } static void clear_remote_op_timers(remote_fencing_op_t * op) { if (op->query_timer) { g_source_remove(op->query_timer); op->query_timer = 0; } if (op->op_timer_total) { g_source_remove(op->op_timer_total); op->op_timer_total = 0; } if (op->op_timer_one) { g_source_remove(op->op_timer_one); op->op_timer_one = 0; } } static void free_remote_op(gpointer data) { remote_fencing_op_t *op = data; crm_log_xml_debug(op->request, "Destroying"); clear_remote_op_timers(op); free(op->id); free(op->action); free(op->delegate); free(op->target); free(op->client_id); free(op->client_name); free(op->originator); if (op->query_results) { g_list_free_full(op->query_results, free_remote_query); } if (op->request) { free_xml(op->request); op->request = NULL; } if (op->devices_list) { g_list_free_full(op->devices_list, free); op->devices_list = NULL; } g_list_free_full(op->automatic_list, free); g_list_free(op->duplicates); free(op); } void init_stonith_remote_op_hash_table(GHashTable **table) { if (*table == NULL) { *table = pcmk__strkey_table(NULL, free_remote_op); } } /*! * \internal * \brief Return an operation's originally requested action (before any remap) * * \param[in] op Operation to check * * \return Operation's original action */ static const char * op_requested_action(const remote_fencing_op_t *op) { return ((op->phase > st_phase_requested)? "reboot" : op->action); } /*! * \internal * \brief Remap a "reboot" operation to the "off" phase * * \param[in,out] op Operation to remap */ static void op_phase_off(remote_fencing_op_t *op) { crm_info("Remapping multiple-device reboot targeting %s to 'off' " CRM_XS " id=%.8s", op->target, op->id); op->phase = st_phase_off; /* Happily, "off" and "on" are shorter than "reboot", so we can reuse the * memory allocation at each phase. */ strcpy(op->action, "off"); } /*! * \internal * \brief Advance a remapped reboot operation to the "on" phase * * \param[in,out] op Operation to remap */ static void op_phase_on(remote_fencing_op_t *op) { GList *iter = NULL; crm_info("Remapped 'off' targeting %s complete, " "remapping to 'on' for %s " CRM_XS " id=%.8s", op->target, op->client_name, op->id); op->phase = st_phase_on; strcpy(op->action, "on"); /* Skip devices with automatic unfencing, because the cluster will handle it * when the node rejoins. */ for (iter = op->automatic_list; iter != NULL; iter = iter->next) { GList *match = g_list_find_custom(op->devices_list, iter->data, sort_strings); if (match) { op->devices_list = g_list_remove(op->devices_list, match->data); } } g_list_free_full(op->automatic_list, free); op->automatic_list = NULL; /* Rewind device list pointer */ op->devices = op->devices_list; } /*! * \internal * \brief Reset a remapped reboot operation * * \param[in,out] op Operation to reset */ static void undo_op_remap(remote_fencing_op_t *op) { if (op->phase > 0) { crm_info("Undoing remap of reboot targeting %s for %s " CRM_XS " id=%.8s", op->target, op->client_name, op->id); op->phase = st_phase_requested; strcpy(op->action, "reboot"); } } static xmlNode * create_op_done_notify(remote_fencing_op_t * op, int rc) { xmlNode *notify_data = create_xml_node(NULL, T_STONITH_NOTIFY_FENCE); crm_xml_add_int(notify_data, "state", op->state); crm_xml_add_int(notify_data, F_STONITH_RC, rc); crm_xml_add(notify_data, F_STONITH_TARGET, op->target); crm_xml_add(notify_data, F_STONITH_ACTION, op->action); crm_xml_add(notify_data, F_STONITH_DELEGATE, op->delegate); crm_xml_add(notify_data, F_STONITH_REMOTE_OP_ID, op->id); crm_xml_add(notify_data, F_STONITH_ORIGIN, op->originator); crm_xml_add(notify_data, F_STONITH_CLIENTID, op->client_id); crm_xml_add(notify_data, F_STONITH_CLIENTNAME, op->client_name); return notify_data; } void stonith_bcast_result_to_peers(remote_fencing_op_t * op, int rc, gboolean op_merged) { static int count = 0; xmlNode *bcast = create_xml_node(NULL, T_STONITH_REPLY); xmlNode *notify_data = create_op_done_notify(op, rc); count++; crm_trace("Broadcasting result to peers"); crm_xml_add(bcast, F_TYPE, T_STONITH_NOTIFY); crm_xml_add(bcast, F_SUBTYPE, "broadcast"); crm_xml_add(bcast, F_STONITH_OPERATION, T_STONITH_NOTIFY); crm_xml_add_int(bcast, "count", count); if (op_merged) { crm_xml_add(bcast, F_STONITH_MERGED, "true"); } add_message_xml(bcast, F_STONITH_CALLDATA, notify_data); send_cluster_message(NULL, crm_msg_stonith_ng, bcast, FALSE); free_xml(notify_data); free_xml(bcast); return; } static void handle_local_reply_and_notify(remote_fencing_op_t * op, xmlNode * data, int rc) { xmlNode *notify_data = NULL; xmlNode *reply = NULL; if (op->notify_sent == TRUE) { /* nothing to do */ return; } /* Do notification with a clean data object */ notify_data = create_op_done_notify(op, rc); crm_xml_add_int(data, "state", op->state); crm_xml_add(data, F_STONITH_TARGET, op->target); crm_xml_add(data, F_STONITH_OPERATION, op->action); reply = stonith_construct_reply(op->request, NULL, data, rc); crm_xml_add(reply, F_STONITH_DELEGATE, op->delegate); /* Send fencing OP reply to local client that initiated fencing */ do_local_reply(reply, op->client_id, op->call_options & st_opt_sync_call, FALSE); /* bcast to all local clients that the fencing operation happend */ do_stonith_notify(0, T_STONITH_NOTIFY_FENCE, rc, notify_data); do_stonith_notify(0, T_STONITH_NOTIFY_HISTORY, 0, NULL); /* mark this op as having notify's already sent */ op->notify_sent = TRUE; free_xml(reply); free_xml(notify_data); } static void handle_duplicates(remote_fencing_op_t * op, xmlNode * data, int rc) { GList *iter = NULL; for (iter = op->duplicates; iter != NULL; iter = iter->next) { remote_fencing_op_t *other = iter->data; if (other->state == st_duplicate) { other->state = op->state; crm_debug("Performing duplicate notification for %s@%s: %s " CRM_XS " id=%.8s", other->client_name, other->originator, pcmk_strerror(rc), other->id); remote_op_done(other, data, rc, TRUE); } else { // Possible if (for example) it timed out already crm_err("Skipping duplicate notification for %s@%s " CRM_XS " state=%s id=%.8s", other->client_name, other->originator, stonith_op_state_str(other->state), other->id); } } } static char * delegate_from_xml(xmlNode *xml) { xmlNode *match = get_xpath_object("//@" F_STONITH_DELEGATE, xml, LOG_NEVER); if (match == NULL) { return crm_element_value_copy(xml, F_ORIG); } else { return crm_element_value_copy(match, F_STONITH_DELEGATE); } } /*! * \internal * \brief Finalize a remote operation. * * \description This function has two code paths. * * Path 1. This node is the owner of the operation and needs * to notify the cpg group via a broadcast as to the operation's * results. * * Path 2. The cpg broadcast is received. All nodes notify their local * stonith clients the operation results. * * So, The owner of the operation first notifies the cluster of the result, * and once that cpg notify is received back it notifies all the local clients. * * Nodes that are passive watchers of the operation will receive the * broadcast and only need to notify their local clients the operation finished. * * \param op, The fencing operation to finalize * \param data, The xml msg reply (if present) of the last delegated fencing * operation. * \param dup, Is this operation a duplicate, if so treat it a little differently * making sure the broadcast is not sent out. */ static void remote_op_done(remote_fencing_op_t * op, xmlNode * data, int rc, int dup) { int level = LOG_ERR; const char *subt = NULL; xmlNode *local_data = NULL; gboolean op_merged = FALSE; set_fencing_completed(op); clear_remote_op_timers(op); undo_op_remap(op); if (op->notify_sent == TRUE) { crm_err("Already sent notifications for '%s' targeting %s by %s for " "client %s@%s: %s " CRM_XS " rc=%d state=%s id=%.8s", op->action, op->target, (op->delegate? op->delegate : "unknown node"), op->client_name, op->originator, pcmk_strerror(rc), rc, stonith_op_state_str(op->state), op->id); goto remote_op_done_cleanup; } if (data == NULL) { data = create_xml_node(NULL, "remote-op"); local_data = data; } else if (op->delegate == NULL) { switch (rc) { case -ENODEV: case -EHOSTUNREACH: break; default: op->delegate = delegate_from_xml(data); break; } } if(dup) { op_merged = TRUE; } else if (crm_element_value(data, F_STONITH_MERGED)) { op_merged = TRUE; } /* Tell everyone the operation is done, we will continue * with doing the local notifications once we receive * the broadcast back. */ subt = crm_element_value(data, F_SUBTYPE); if (dup == FALSE && !pcmk__str_eq(subt, "broadcast", pcmk__str_casei)) { /* Defer notification until the bcast message arrives */ stonith_bcast_result_to_peers(op, rc, (op_merged? TRUE: FALSE)); goto remote_op_done_cleanup; } if (rc == pcmk_ok || dup) { level = LOG_NOTICE; } else if (!pcmk__str_eq(op->originator, stonith_our_uname, pcmk__str_casei)) { level = LOG_NOTICE; } do_crm_log(level, "Operation '%s'%s%s by %s for %s@%s%s: %s " CRM_XS " id=%.8s", op->action, (op->target? " targeting " : ""), (op->target? op->target : ""), (op->delegate? op->delegate : "unknown node"), op->client_name, op->originator, (op_merged? " (merged)" : ""), pcmk_strerror(rc), op->id); handle_local_reply_and_notify(op, data, rc); if (dup == FALSE) { handle_duplicates(op, data, rc); } /* Free non-essential parts of the record * Keep the record around so we can query the history */ if (op->query_results) { g_list_free_full(op->query_results, free_remote_query); op->query_results = NULL; } if (op->request) { free_xml(op->request); op->request = NULL; } remote_op_done_cleanup: free_xml(local_data); } static gboolean remote_op_watchdog_done(gpointer userdata) { remote_fencing_op_t *op = userdata; op->op_timer_one = 0; crm_notice("Self-fencing (%s) by %s for %s assumed complete " CRM_XS " id=%.8s", op->action, op->target, op->client_name, op->id); op->state = st_done; remote_op_done(op, NULL, pcmk_ok, FALSE); return FALSE; } static gboolean remote_op_timeout_one(gpointer userdata) { remote_fencing_op_t *op = userdata; op->op_timer_one = 0; crm_notice("Peer's '%s' action targeting %s for client %s timed out " CRM_XS " id=%.8s", op->action, op->target, op->client_name, op->id); call_remote_stonith(op, NULL, pcmk_ok); return FALSE; } static gboolean remote_op_timeout(gpointer userdata) { remote_fencing_op_t *op = userdata; op->op_timer_total = 0; if (op->state == st_done) { crm_debug("Action '%s' targeting %s for client %s already completed " CRM_XS " id=%.8s", op->action, op->target, op->client_name, op->id); return FALSE; } crm_debug("Action '%s' targeting %s for client %s timed out " CRM_XS " id=%.8s", op->action, op->target, op->client_name, op->id); if (op->phase == st_phase_on) { /* A remapped reboot operation timed out in the "on" phase, but the * "off" phase completed successfully, so quit trying any further * devices, and return success. */ remote_op_done(op, NULL, pcmk_ok, FALSE); return FALSE; } op->state = st_failed; remote_op_done(op, NULL, -ETIME, FALSE); return FALSE; } static gboolean remote_op_query_timeout(gpointer data) { remote_fencing_op_t *op = data; op->query_timer = 0; if (op->state == st_done) { crm_debug("Operation %.8s targeting %s already completed", op->id, op->target); } else if (op->state == st_exec) { crm_debug("Operation %.8s targeting %s already in progress", op->id, op->target); } else if (op->query_results) { crm_debug("Query %.8s targeting %s complete (state=%s)", op->id, op->target, stonith_op_state_str(op->state)); call_remote_stonith(op, NULL, pcmk_ok); } else { crm_debug("Query %.8s targeting %s timed out (state=%s)", op->id, op->target, stonith_op_state_str(op->state)); if (op->op_timer_total) { g_source_remove(op->op_timer_total); op->op_timer_total = 0; } remote_op_timeout(op); } return FALSE; } static gboolean topology_is_empty(stonith_topology_t *tp) { int i; if (tp == NULL) { return TRUE; } for (i = 0; i < ST_LEVEL_MAX; i++) { if (tp->levels[i] != NULL) { return FALSE; } } return TRUE; } /*! * \internal * \brief Add a device to an operation's automatic unfencing list * * \param[in,out] op Operation to modify * \param[in] device Device ID to add */ static void add_required_device(remote_fencing_op_t *op, const char *device) { GList *match = g_list_find_custom(op->automatic_list, device, sort_strings); if (!match) { op->automatic_list = g_list_prepend(op->automatic_list, strdup(device)); } } /*! * \internal * \brief Remove a device from the automatic unfencing list * * \param[in,out] op Operation to modify * \param[in] device Device ID to remove */ static void remove_required_device(remote_fencing_op_t *op, const char *device) { GList *match = g_list_find_custom(op->automatic_list, device, sort_strings); if (match) { op->automatic_list = g_list_remove(op->automatic_list, match->data); } } /* deep copy the device list */ static void set_op_device_list(remote_fencing_op_t * op, GList *devices) { GList *lpc = NULL; if (op->devices_list) { g_list_free_full(op->devices_list, free); op->devices_list = NULL; } for (lpc = devices; lpc != NULL; lpc = lpc->next) { op->devices_list = g_list_append(op->devices_list, strdup(lpc->data)); } op->devices = op->devices_list; } /*! * \internal * \brief Check whether a node matches a topology target * * \param[in] tp Topology table entry to check * \param[in] node Name of node to check * * \return TRUE if node matches topology target */ static gboolean topology_matches(const stonith_topology_t *tp, const char *node) { regex_t r_patt; CRM_CHECK(node && tp && tp->target, return FALSE); switch(tp->kind) { case 2: /* This level targets by attribute, so tp->target is a NAME=VALUE pair * of a permanent attribute applied to targeted nodes. The test below * relies on the locally cached copy of the CIB, so if fencing needs to * be done before the initial CIB is received or after a malformed CIB * is received, then the topology will be unable to be used. */ if (node_has_attr(node, tp->target_attribute, tp->target_value)) { crm_notice("Matched %s with %s by attribute", node, tp->target); return TRUE; } break; case 1: /* This level targets by name, so tp->target is a regular expression * matching names of nodes to be targeted. */ if (regcomp(&r_patt, tp->target_pattern, REG_EXTENDED|REG_NOSUB)) { crm_info("Bad regex '%s' for fencing level", tp->target); } else { int status = regexec(&r_patt, node, 0, NULL, 0); regfree(&r_patt); if (status == 0) { crm_notice("Matched %s with %s by name", node, tp->target); return TRUE; } } break; case 0: crm_trace("Testing %s against %s", node, tp->target); return pcmk__str_eq(tp->target, node, pcmk__str_casei); } crm_trace("No match for %s with %s", node, tp->target); return FALSE; } stonith_topology_t * find_topology_for_host(const char *host) { GHashTableIter tIter; stonith_topology_t *tp = g_hash_table_lookup(topology, host); if(tp != NULL) { crm_trace("Found %s for %s in %d entries", tp->target, host, g_hash_table_size(topology)); return tp; } g_hash_table_iter_init(&tIter, topology); while (g_hash_table_iter_next(&tIter, NULL, (gpointer *) & tp)) { if (topology_matches(tp, host)) { crm_trace("Found %s for %s in %d entries", tp->target, host, g_hash_table_size(topology)); return tp; } } crm_trace("No matches for %s in %d topology entries", host, g_hash_table_size(topology)); return NULL; } /*! * \internal * \brief Set fencing operation's device list to target's next topology level * * \param[in,out] op Remote fencing operation to modify * \param[in] empty_ok If true, an operation without a target (i.e. * queries) or a target without a topology will get a * pcmk_rc_ok return value instead of ENODEV * * \return Standard Pacemaker return value */ static int advance_topology_level(remote_fencing_op_t *op, bool empty_ok) { stonith_topology_t *tp = NULL; if (op->target) { tp = find_topology_for_host(op->target); } if (topology_is_empty(tp)) { return empty_ok? pcmk_rc_ok : ENODEV; } CRM_ASSERT(tp->levels != NULL); stonith__set_call_options(op->call_options, op->id, st_opt_topology); /* This is a new level, so undo any remapping left over from previous */ undo_op_remap(op); do { op->level++; } while (op->level < ST_LEVEL_MAX && tp->levels[op->level] == NULL); if (op->level < ST_LEVEL_MAX) { crm_trace("Attempting fencing level %d targeting %s (%d devices) " "for client %s@%s (id=%.8s)", op->level, op->target, g_list_length(tp->levels[op->level]), op->client_name, op->originator, op->id); set_op_device_list(op, tp->levels[op->level]); // The requested delay has been applied for the first fencing level if (op->level > 1 && op->delay > 0) { op->delay = 0; } if (g_list_next(op->devices_list) && pcmk__str_eq(op->action, "reboot", pcmk__str_casei)) { /* A reboot has been requested for a topology level with multiple * devices. Instead of rebooting the devices sequentially, we will * turn them all off, then turn them all on again. (Think about * switched power outlets for redundant power supplies.) */ op_phase_off(op); } return pcmk_rc_ok; } crm_notice("All fencing options targeting %s for client %s@%s failed " CRM_XS " id=%.8s", op->target, op->client_name, op->originator, op->id); return ENODEV; } /*! * \brief Check to see if this operation is a duplicate of another in flight * operation. If so merge this operation into the inflight operation, and mark * it as a duplicate. */ static void merge_duplicates(remote_fencing_op_t * op) { GHashTableIter iter; remote_fencing_op_t *other = NULL; time_t now = time(NULL); g_hash_table_iter_init(&iter, stonith_remote_op_list); while (g_hash_table_iter_next(&iter, NULL, (void **)&other)) { const char *other_action = op_requested_action(other); if (!strcmp(op->id, other->id)) { continue; // Don't compare against self } if (other->state > st_exec) { crm_trace("%.8s not duplicate of %.8s: not in progress", op->id, other->id); continue; } if (!pcmk__str_eq(op->target, other->target, pcmk__str_casei)) { crm_trace("%.8s not duplicate of %.8s: node %s vs. %s", op->id, other->id, op->target, other->target); continue; } if (!pcmk__str_eq(op->action, other_action, pcmk__str_casei)) { crm_trace("%.8s not duplicate of %.8s: action %s vs. %s", op->id, other->id, op->action, other_action); continue; } if (pcmk__str_eq(op->client_name, other->client_name, pcmk__str_casei)) { crm_trace("%.8s not duplicate of %.8s: same client %s", op->id, other->id, op->client_name); continue; } if (pcmk__str_eq(other->target, other->originator, pcmk__str_casei)) { crm_trace("%.8s not duplicate of %.8s: suicide for %s", op->id, other->id, other->target); continue; } if (!fencing_peer_active(crm_get_peer(0, other->originator))) { crm_notice("Failing action '%s' targeting %s originating from " "client %s@%s: Originator is dead " CRM_XS " id=%.8s", other->action, other->target, other->client_name, other->originator, other->id); crm_trace("%.8s not duplicate of %.8s: originator dead", op->id, other->id); other->state = st_failed; continue; } if ((other->total_timeout > 0) && (now > (other->total_timeout + other->created))) { crm_trace("%.8s not duplicate of %.8s: old (%ld vs. %ld + %d)", op->id, other->id, now, other->created, other->total_timeout); continue; } /* There is another in-flight request to fence the same host * Piggyback on that instead. If it fails, so do we. */ other->duplicates = g_list_append(other->duplicates, op); if (other->total_timeout == 0) { other->total_timeout = op->total_timeout = TIMEOUT_MULTIPLY_FACTOR * get_op_total_timeout(op, NULL); crm_trace("Best guess as to timeout used for %.8s: %d", other->id, other->total_timeout); } crm_notice("Merging fencing action '%s' targeting %s originating from " "client %s with identical request from %s@%s " CRM_XS " original=%.8s duplicate=%.8s total_timeout=%ds", op->action, op->target, op->client_name, other->client_name, other->originator, op->id, other->id, other->total_timeout); report_timeout_period(op, other->total_timeout); op->state = st_duplicate; } } static uint32_t fencing_active_peers(void) { uint32_t count = 0; crm_node_t *entry; GHashTableIter gIter; g_hash_table_iter_init(&gIter, crm_peer_cache); while (g_hash_table_iter_next(&gIter, NULL, (void **)&entry)) { if(fencing_peer_active(entry)) { count++; } } return count; } int stonith_manual_ack(xmlNode * msg, remote_fencing_op_t * op) { xmlNode *dev = get_xpath_object("//@" F_STONITH_TARGET, msg, LOG_ERR); op->state = st_done; set_fencing_completed(op); op->delegate = strdup("a human"); crm_notice("Injecting manual confirmation that %s is safely off/down", crm_element_value(dev, F_STONITH_TARGET)); remote_op_done(op, msg, pcmk_ok, FALSE); - /* Replies are sent via done_cb->stonith_send_async_reply()->do_local_reply() */ + // Replies are sent via done_cb -> send_async_reply() -> do_local_reply() return -EINPROGRESS; } /*! * \internal * \brief Create a new remote stonith operation * * \param[in] client ID of local stonith client that initiated the operation * \param[in] request The request from the client that started the operation * \param[in] peer TRUE if this operation is owned by another stonith peer * (an operation owned by one peer is stored on all peers, * but only the owner executes it; all nodes get the results * once the owner finishes execution) */ void * create_remote_stonith_op(const char *client, xmlNode * request, gboolean peer) { remote_fencing_op_t *op = NULL; xmlNode *dev = get_xpath_object("//@" F_STONITH_TARGET, request, LOG_NEVER); int call_options = 0; const char *operation = NULL; init_stonith_remote_op_hash_table(&stonith_remote_op_list); /* If this operation is owned by another node, check to make * sure we haven't already created this operation. */ if (peer && dev) { const char *op_id = crm_element_value(dev, F_STONITH_REMOTE_OP_ID); CRM_CHECK(op_id != NULL, return NULL); op = g_hash_table_lookup(stonith_remote_op_list, op_id); if (op) { crm_debug("Reusing existing remote fencing op %.8s for %s", op_id, ((client == NULL)? "unknown client" : client)); return op; } } op = calloc(1, sizeof(remote_fencing_op_t)); CRM_ASSERT(op != NULL); crm_element_value_int(request, F_STONITH_TIMEOUT, &(op->base_timeout)); // Value -1 means disable any static/random fencing delays crm_element_value_int(request, F_STONITH_DELAY, &(op->delay)); if (peer && dev) { op->id = crm_element_value_copy(dev, F_STONITH_REMOTE_OP_ID); } else { op->id = crm_generate_uuid(); } g_hash_table_replace(stonith_remote_op_list, op->id, op); op->state = st_query; op->replies_expected = fencing_active_peers(); op->action = crm_element_value_copy(dev, F_STONITH_ACTION); op->originator = crm_element_value_copy(dev, F_STONITH_ORIGIN); op->delegate = crm_element_value_copy(dev, F_STONITH_DELEGATE); /* May not be set */ op->created = time(NULL); if (op->originator == NULL) { /* Local or relayed request */ op->originator = strdup(stonith_our_uname); } CRM_LOG_ASSERT(client != NULL); if (client) { op->client_id = strdup(client); } /* For a RELAY operation, set fenced on the client. */ operation = crm_element_value(request, F_STONITH_OPERATION); if (pcmk__str_eq(operation, STONITH_OP_RELAY, pcmk__str_none)) { op->client_name = crm_strdup_printf("%s.%lu", crm_system_name, (unsigned long) getpid()); } else { op->client_name = crm_element_value_copy(request, F_STONITH_CLIENTNAME); } op->target = crm_element_value_copy(dev, F_STONITH_TARGET); op->request = copy_xml(request); /* TODO: Figure out how to avoid this */ crm_element_value_int(request, F_STONITH_CALLOPTS, &call_options); op->call_options = call_options; crm_element_value_int(request, F_STONITH_CALLID, &(op->client_callid)); crm_trace("%s new fencing op %s ('%s' targeting %s for client %s, " "base timeout %d, %u %s expected)", (peer && dev)? "Recorded" : "Generated", op->id, op->action, op->target, op->client_name, op->base_timeout, op->replies_expected, pcmk__plural_alt(op->replies_expected, "reply", "replies")); if (op->call_options & st_opt_cs_nodeid) { int nodeid; crm_node_t *node; pcmk__scan_min_int(op->target, &nodeid, 0); node = pcmk__search_known_node_cache(nodeid, NULL, CRM_GET_PEER_ANY); /* Ensure the conversion only happens once */ stonith__clear_call_options(op->call_options, op->id, st_opt_cs_nodeid); if (node && node->uname) { free(op->target); op->target = strdup(node->uname); } else { crm_warn("Could not expand nodeid '%s' into a host name", op->target); } } /* check to see if this is a duplicate operation of another in-flight operation */ merge_duplicates(op); if (op->state != st_duplicate) { /* kick history readers */ do_stonith_notify(0, T_STONITH_NOTIFY_HISTORY, 0, NULL); } /* safe to trim as long as that doesn't touch pending ops */ stonith_fence_history_trim(); return op; } remote_fencing_op_t * initiate_remote_stonith_op(pcmk__client_t *client, xmlNode *request, gboolean manual_ack) { int query_timeout = 0; xmlNode *query = NULL; const char *client_id = NULL; remote_fencing_op_t *op = NULL; const char *relay_op_id = NULL; const char *operation = NULL; if (client) { client_id = client->id; } else { client_id = crm_element_value(request, F_STONITH_CLIENTID); } CRM_LOG_ASSERT(client_id != NULL); op = create_remote_stonith_op(client_id, request, FALSE); op->owner = TRUE; if (manual_ack) { crm_notice("Processing manual confirmation of fencing targeting %s " CRM_XS " id=%.8s", op->target, op->id); return op; } CRM_CHECK(op->action, return NULL); if (advance_topology_level(op, true) != pcmk_rc_ok) { op->state = st_failed; } switch (op->state) { case st_failed: crm_warn("Could not request peer fencing (%s) targeting %s " CRM_XS " id=%.8s", op->action, op->target, op->id); remote_op_done(op, NULL, -EINVAL, FALSE); return op; case st_duplicate: crm_info("Requesting peer fencing (%s) targeting %s (duplicate) " CRM_XS " id=%.8s", op->action, op->target, op->id); return op; default: crm_notice("Requesting peer fencing (%s) targeting %s " CRM_XS " id=%.8s state=%s base_timeout=%d", op->action, op->target, op->id, stonith_op_state_str(op->state), op->base_timeout); } query = stonith_create_op(op->client_callid, op->id, STONITH_OP_QUERY, NULL, op->call_options); crm_xml_add(query, F_STONITH_REMOTE_OP_ID, op->id); crm_xml_add(query, F_STONITH_TARGET, op->target); crm_xml_add(query, F_STONITH_ACTION, op_requested_action(op)); crm_xml_add(query, F_STONITH_ORIGIN, op->originator); crm_xml_add(query, F_STONITH_CLIENTID, op->client_id); crm_xml_add(query, F_STONITH_CLIENTNAME, op->client_name); crm_xml_add_int(query, F_STONITH_TIMEOUT, op->base_timeout); /* In case of RELAY operation, RELAY information is added to the query to delete the original operation of RELAY. */ operation = crm_element_value(request, F_STONITH_OPERATION); if (pcmk__str_eq(operation, STONITH_OP_RELAY, pcmk__str_none)) { relay_op_id = crm_element_value(request, F_STONITH_REMOTE_OP_ID); if (relay_op_id) { crm_xml_add(query, F_STONITH_REMOTE_OP_ID_RELAY, relay_op_id); } } send_cluster_message(NULL, crm_msg_stonith_ng, query, FALSE); free_xml(query); query_timeout = op->base_timeout * TIMEOUT_MULTIPLY_FACTOR; op->query_timer = g_timeout_add((1000 * query_timeout), remote_op_query_timeout, op); return op; } enum find_best_peer_options { /*! Skip checking the target peer for capable fencing devices */ FIND_PEER_SKIP_TARGET = 0x0001, /*! Only check the target peer for capable fencing devices */ FIND_PEER_TARGET_ONLY = 0x0002, /*! Skip peers and devices that are not verified */ FIND_PEER_VERIFIED_ONLY = 0x0004, }; static st_query_result_t * find_best_peer(const char *device, remote_fencing_op_t * op, enum find_best_peer_options options) { GList *iter = NULL; gboolean verified_devices_only = (options & FIND_PEER_VERIFIED_ONLY) ? TRUE : FALSE; if (!device && pcmk_is_set(op->call_options, st_opt_topology)) { return NULL; } for (iter = op->query_results; iter != NULL; iter = iter->next) { st_query_result_t *peer = iter->data; crm_trace("Testing result from %s targeting %s with %d device%s: %d %x", peer->host, op->target, peer->ndevices, pcmk__plural_s(peer->ndevices), peer->tried, options); if ((options & FIND_PEER_SKIP_TARGET) && pcmk__str_eq(peer->host, op->target, pcmk__str_casei)) { continue; } if ((options & FIND_PEER_TARGET_ONLY) && !pcmk__str_eq(peer->host, op->target, pcmk__str_casei)) { continue; } if (pcmk_is_set(op->call_options, st_opt_topology)) { if (grab_peer_device(op, peer, device, verified_devices_only)) { return peer; } } else if ((peer->tried == FALSE) && count_peer_devices(op, peer, verified_devices_only)) { /* No topology: Use the current best peer */ crm_trace("Simple fencing"); return peer; } } return NULL; } static st_query_result_t * stonith_choose_peer(remote_fencing_op_t * op) { const char *device = NULL; st_query_result_t *peer = NULL; uint32_t active = fencing_active_peers(); do { if (op->devices) { device = op->devices->data; crm_trace("Checking for someone to fence (%s) %s using %s", op->action, op->target, device); } else { crm_trace("Checking for someone to fence (%s) %s", op->action, op->target); } /* Best choice is a peer other than the target with verified access */ peer = find_best_peer(device, op, FIND_PEER_SKIP_TARGET|FIND_PEER_VERIFIED_ONLY); if (peer) { crm_trace("Found verified peer %s for %s", peer->host, device?device:""); return peer; } if(op->query_timer != 0 && op->replies < QB_MIN(op->replies_expected, active)) { crm_trace("Waiting before looking for unverified devices to fence %s", op->target); return NULL; } /* If no other peer has verified access, next best is unverified access */ peer = find_best_peer(device, op, FIND_PEER_SKIP_TARGET); if (peer) { crm_trace("Found best unverified peer %s", peer->host); return peer; } /* If no other peer can do it, last option is self-fencing * (which is never allowed for the "on" phase of a remapped reboot) */ if (op->phase != st_phase_on) { peer = find_best_peer(device, op, FIND_PEER_TARGET_ONLY); if (peer) { crm_trace("%s will fence itself", peer->host); return peer; } } /* Try the next fencing level if there is one (unless we're in the "on" * phase of a remapped "reboot", because we ignore errors in that case) */ } while ((op->phase != st_phase_on) && pcmk_is_set(op->call_options, st_opt_topology) && (advance_topology_level(op, false) == pcmk_rc_ok)); crm_notice("Couldn't find anyone to fence (%s) %s using %s", op->action, op->target, (device? device : "any device")); return NULL; } static int get_device_timeout(const remote_fencing_op_t *op, const st_query_result_t *peer, const char *device) { device_properties_t *props; if (!peer || !device) { return op->base_timeout; } props = g_hash_table_lookup(peer->devices, device); if (!props) { return op->base_timeout; } return (props->custom_action_timeout[op->phase]? props->custom_action_timeout[op->phase] : op->base_timeout) + props->delay_max[op->phase]; } struct timeout_data { const remote_fencing_op_t *op; const st_query_result_t *peer; int total_timeout; }; /*! * \internal * \brief Add timeout to a total if device has not been executed yet * * \param[in] key GHashTable key (device ID) * \param[in] value GHashTable value (device properties) * \param[in] user_data Timeout data */ static void add_device_timeout(gpointer key, gpointer value, gpointer user_data) { const char *device_id = key; device_properties_t *props = value; struct timeout_data *timeout = user_data; if (!props->executed[timeout->op->phase] && !props->disallowed[timeout->op->phase]) { timeout->total_timeout += get_device_timeout(timeout->op, timeout->peer, device_id); } } static int get_peer_timeout(const remote_fencing_op_t *op, const st_query_result_t *peer) { struct timeout_data timeout; timeout.op = op; timeout.peer = peer; timeout.total_timeout = 0; g_hash_table_foreach(peer->devices, add_device_timeout, &timeout); return (timeout.total_timeout? timeout.total_timeout : op->base_timeout); } static int get_op_total_timeout(const remote_fencing_op_t *op, const st_query_result_t *chosen_peer) { int total_timeout = 0; stonith_topology_t *tp = find_topology_for_host(op->target); if (pcmk_is_set(op->call_options, st_opt_topology) && tp) { int i; GList *device_list = NULL; GList *iter = NULL; /* Yep, this looks scary, nested loops all over the place. * Here is what is going on. * Loop1: Iterate through fencing levels. * Loop2: If a fencing level has devices, loop through each device * Loop3: For each device in a fencing level, see what peer owns it * and what that peer has reported the timeout is for the device. */ for (i = 0; i < ST_LEVEL_MAX; i++) { if (!tp->levels[i]) { continue; } for (device_list = tp->levels[i]; device_list; device_list = device_list->next) { for (iter = op->query_results; iter != NULL; iter = iter->next) { const st_query_result_t *peer = iter->data; if (find_peer_device(op, peer, device_list->data)) { total_timeout += get_device_timeout(op, peer, device_list->data); break; } } /* End Loop3: match device with peer that owns device, find device's timeout period */ } /* End Loop2: iterate through devices at a specific level */ } /*End Loop1: iterate through fencing levels */ } else if (chosen_peer) { total_timeout = get_peer_timeout(op, chosen_peer); } else { total_timeout = op->base_timeout; } return total_timeout ? total_timeout : op->base_timeout; } static void report_timeout_period(remote_fencing_op_t * op, int op_timeout) { GList *iter = NULL; xmlNode *update = NULL; const char *client_node = NULL; const char *client_id = NULL; const char *call_id = NULL; if (op->call_options & st_opt_sync_call) { /* There is no reason to report the timeout for a synchronous call. It * is impossible to use the reported timeout to do anything when the client * is blocking for the response. This update is only important for * async calls that require a callback to report the results in. */ return; } else if (!op->request) { return; } crm_trace("Reporting timeout for %s (id=%.8s)", op->client_name, op->id); client_node = crm_element_value(op->request, F_STONITH_CLIENTNODE); call_id = crm_element_value(op->request, F_STONITH_CALLID); client_id = crm_element_value(op->request, F_STONITH_CLIENTID); if (!client_node || !call_id || !client_id) { return; } if (pcmk__str_eq(client_node, stonith_our_uname, pcmk__str_casei)) { /* The client is connected to this node, send the update direclty to them */ do_stonith_async_timeout_update(client_id, call_id, op_timeout); return; } /* The client is connected to another node, relay this update to them */ update = stonith_create_op(op->client_callid, op->id, STONITH_OP_TIMEOUT_UPDATE, NULL, 0); crm_xml_add(update, F_STONITH_REMOTE_OP_ID, op->id); crm_xml_add(update, F_STONITH_CLIENTID, client_id); crm_xml_add(update, F_STONITH_CALLID, call_id); crm_xml_add_int(update, F_STONITH_TIMEOUT, op_timeout); send_cluster_message(crm_get_peer(0, client_node), crm_msg_stonith_ng, update, FALSE); free_xml(update); for (iter = op->duplicates; iter != NULL; iter = iter->next) { remote_fencing_op_t *dup = iter->data; crm_trace("Reporting timeout for duplicate %.8s to client %s", dup->id, dup->client_name); report_timeout_period(iter->data, op_timeout); } } /*! * \internal * \brief Advance an operation to the next device in its topology * * \param[in,out] op Operation to advance * \param[in] device ID of device just completed * \param[in] msg XML reply that contained device result (if available) * \param[in] rc Return code of device's execution */ static void advance_topology_device_in_level(remote_fencing_op_t *op, const char *device, xmlNode *msg, int rc) { /* Advance to the next device at this topology level, if any */ if (op->devices) { op->devices = op->devices->next; } /* Handle automatic unfencing if an "on" action was requested */ if ((op->phase == st_phase_requested) && pcmk__str_eq(op->action, "on", pcmk__str_casei)) { /* If the device we just executed was required, it's not anymore */ remove_required_device(op, device); /* If there are no more devices at this topology level, run through any * remaining devices with automatic unfencing */ if (op->devices == NULL) { op->devices = op->automatic_list; } } if ((op->devices == NULL) && (op->phase == st_phase_off)) { /* We're done with this level and with required devices, but we had * remapped "reboot" to "off", so start over with "on". If any devices * need to be turned back on, op->devices will be non-NULL after this. */ op_phase_on(op); } if (op->devices) { /* Necessary devices remain, so execute the next one */ crm_trace("Next targeting %s on behalf of %s@%s (rc was %d)", op->target, op->client_name, op->originator, rc); // The requested delay has been applied for the first device if (op->delay > 0) { op->delay = 0; } call_remote_stonith(op, NULL, pcmk_ok); } else { /* We're done with all devices and phases, so finalize operation */ crm_trace("Marking complex fencing op targeting %s as complete", op->target); op->state = st_done; remote_op_done(op, msg, rc, FALSE); } } static gboolean check_watchdog_fencing_and_wait(remote_fencing_op_t * op) { if (node_does_watchdog_fencing(op->target)) { crm_notice("Waiting %lds for %s to self-fence (%s) for " "client %s " CRM_XS " id=%.8s", (stonith_watchdog_timeout_ms / 1000), op->target, op->action, op->client_name, op->id); op->op_timer_one = g_timeout_add(stonith_watchdog_timeout_ms, remote_op_watchdog_done, op); return TRUE; } else { crm_debug("Skipping fallback to watchdog-fencing as %s is " "not in host-list", op->target); } return FALSE; } void call_remote_stonith(remote_fencing_op_t * op, st_query_result_t * peer, int rc) { const char *device = NULL; int timeout = op->base_timeout; crm_trace("Action %.8s targeting %s for %s is %s", op->id, op->target, op->client_name, stonith_op_state_str(op->state)); if ((peer == NULL) && !pcmk_is_set(op->call_options, st_opt_topology)) { peer = stonith_choose_peer(op); } if (!op->op_timer_total) { int total_timeout = get_op_total_timeout(op, peer); op->total_timeout = TIMEOUT_MULTIPLY_FACTOR * total_timeout; op->op_timer_total = g_timeout_add(1000 * op->total_timeout, remote_op_timeout, op); report_timeout_period(op, op->total_timeout); crm_info("Total timeout set to %d for peer's fencing targeting %s for %s" CRM_XS "id=%.8s", total_timeout, op->target, op->client_name, op->id); } if (pcmk_is_set(op->call_options, st_opt_topology) && op->devices) { /* Ignore any peer preference, they might not have the device we need */ /* When using topology, stonith_choose_peer() removes the device from * further consideration, so be sure to calculate timeout beforehand */ peer = stonith_choose_peer(op); device = op->devices->data; timeout = get_device_timeout(op, peer, device); } if (peer) { int timeout_one = 0; xmlNode *remote_op = stonith_create_op(op->client_callid, op->id, STONITH_OP_FENCE, NULL, 0); crm_xml_add(remote_op, F_STONITH_REMOTE_OP_ID, op->id); crm_xml_add(remote_op, F_STONITH_TARGET, op->target); crm_xml_add(remote_op, F_STONITH_ACTION, op->action); crm_xml_add(remote_op, F_STONITH_ORIGIN, op->originator); crm_xml_add(remote_op, F_STONITH_CLIENTID, op->client_id); crm_xml_add(remote_op, F_STONITH_CLIENTNAME, op->client_name); crm_xml_add_int(remote_op, F_STONITH_TIMEOUT, timeout); crm_xml_add_int(remote_op, F_STONITH_CALLOPTS, op->call_options); crm_xml_add_int(remote_op, F_STONITH_DELAY, op->delay); if (device) { timeout_one = TIMEOUT_MULTIPLY_FACTOR * get_device_timeout(op, peer, device); crm_notice("Requesting that %s perform '%s' action targeting %s " "using %s " CRM_XS " for client %s (%ds)", peer->host, op->action, op->target, device, op->client_name, timeout_one); crm_xml_add(remote_op, F_STONITH_DEVICE, device); } else { timeout_one = TIMEOUT_MULTIPLY_FACTOR * get_peer_timeout(op, peer); crm_notice("Requesting that %s perform '%s' action targeting %s " CRM_XS " for client %s (%ds, %lds)", peer->host, op->action, op->target, op->client_name, timeout_one, stonith_watchdog_timeout_ms); } op->state = st_exec; if (op->op_timer_one) { g_source_remove(op->op_timer_one); } if (!(stonith_watchdog_timeout_ms > 0 && ( (pcmk__str_eq(device, STONITH_WATCHDOG_ID, pcmk__str_none)) || (pcmk__str_eq(peer->host, op->target, pcmk__str_casei) && !pcmk__str_eq(op->action, "on", pcmk__str_casei))) && check_watchdog_fencing_and_wait(op))) { /* Some thoughts about self-fencing cases reaching this point: - Actually check in check_watchdog_fencing_and_wait shouldn't fail if STONITH_WATCHDOG_ID is chosen as fencing-device and it being present implies watchdog-fencing is enabled anyway - If watchdog-fencing is disabled either in general or for a specific target - detected in check_watchdog_fencing_and_wait - for some other kind of self-fencing we can't expect a success answer but timeout is fine if the node doesn't come back in between - Delicate might be the case where we have watchdog-fencing enabled for a node but the watchdog-fencing-device isn't explicitly chosen for suicide. Local pe-execution in sbd may detect the node as unclean and lead to timely suicide. Otherwise the selection of stonith-watchdog-timeout at least is questionable. */ op->op_timer_one = g_timeout_add((1000 * timeout_one), remote_op_timeout_one, op); } send_cluster_message(crm_get_peer(0, peer->host), crm_msg_stonith_ng, remote_op, FALSE); peer->tried = TRUE; free_xml(remote_op); return; } else if (op->phase == st_phase_on) { /* A remapped "on" cannot be executed, but the node was already * turned off successfully, so ignore the error and continue. */ crm_warn("Ignoring %s 'on' failure (no capable peers) targeting %s " "after successful 'off'", device, op->target); advance_topology_device_in_level(op, device, NULL, pcmk_ok); return; } else if (op->owner == FALSE) { crm_err("Fencing (%s) targeting %s for client %s is not ours to control", op->action, op->target, op->client_name); } else if (op->query_timer == 0) { /* We've exhausted all available peers */ crm_info("No remaining peers capable of fencing (%s) %s for client %s " CRM_XS " state=%s", op->action, op->target, op->client_name, stonith_op_state_str(op->state)); CRM_LOG_ASSERT(op->state < st_done); remote_op_timeout(op); } else if(op->replies >= op->replies_expected || op->replies >= fencing_active_peers()) { // int rc = -EHOSTUNREACH; /* if the operation never left the query state, * but we have all the expected replies, then no devices * are available to execute the fencing operation. */ if(stonith_watchdog_timeout_ms > 0 && pcmk__str_eq(device, STONITH_WATCHDOG_ID, pcmk__str_null_matches)) { if (check_watchdog_fencing_and_wait(op)) { return; } } if (op->state == st_query) { crm_info("No peers (out of %d) have devices capable of fencing " "(%s) %s for client %s " CRM_XS " state=%s", op->replies, op->action, op->target, op->client_name, stonith_op_state_str(op->state)); rc = -ENODEV; } else { if (pcmk_is_set(op->call_options, st_opt_topology)) { rc = -EHOSTUNREACH; } crm_info("No peers (out of %d) are capable of fencing (%s) %s " "for client %s " CRM_XS " state=%s", op->replies, op->action, op->target, op->client_name, stonith_op_state_str(op->state)); } op->state = st_failed; remote_op_done(op, NULL, rc, FALSE); } else { crm_info("Waiting for additional peers capable of fencing (%s) %s%s%s " "for client %s " CRM_XS " id=%.8s", op->action, op->target, (device? " using " : ""), (device? device : ""), op->client_name, op->id); } } /*! * \internal * \brief Comparison function for sorting query results * * \param[in] a GList item to compare * \param[in] b GList item to compare * * \return Per the glib documentation, "a negative integer if the first value * comes before the second, 0 if they are equal, or a positive integer * if the first value comes after the second." */ static gint sort_peers(gconstpointer a, gconstpointer b) { const st_query_result_t *peer_a = a; const st_query_result_t *peer_b = b; return (peer_b->ndevices - peer_a->ndevices); } /*! * \internal * \brief Determine if all the devices in the topology are found or not */ static gboolean all_topology_devices_found(remote_fencing_op_t * op) { GList *device = NULL; GList *iter = NULL; device_properties_t *match = NULL; stonith_topology_t *tp = NULL; gboolean skip_target = FALSE; int i; tp = find_topology_for_host(op->target); if (!tp) { return FALSE; } if (pcmk__strcase_any_of(op->action, "off", "reboot", NULL)) { /* Don't count the devices on the target node if we are killing * the target node. */ skip_target = TRUE; } for (i = 0; i < ST_LEVEL_MAX; i++) { for (device = tp->levels[i]; device; device = device->next) { match = NULL; for (iter = op->query_results; iter && !match; iter = iter->next) { st_query_result_t *peer = iter->data; if (skip_target && pcmk__str_eq(peer->host, op->target, pcmk__str_casei)) { continue; } match = find_peer_device(op, peer, device->data); } if (!match) { return FALSE; } } } return TRUE; } /*! * \internal * \brief Parse action-specific device properties from XML * * \param[in] msg XML element containing the properties * \param[in] peer Name of peer that sent XML (for logs) * \param[in] device Device ID (for logs) * \param[in] action Action the properties relate to (for logs) * \param[in] phase Phase the properties relate to * \param[in,out] props Device properties to update */ static void parse_action_specific(xmlNode *xml, const char *peer, const char *device, const char *action, remote_fencing_op_t *op, enum st_remap_phase phase, device_properties_t *props) { props->custom_action_timeout[phase] = 0; crm_element_value_int(xml, F_STONITH_ACTION_TIMEOUT, &props->custom_action_timeout[phase]); if (props->custom_action_timeout[phase]) { crm_trace("Peer %s with device %s returned %s action timeout %d", peer, device, action, props->custom_action_timeout[phase]); } props->delay_max[phase] = 0; crm_element_value_int(xml, F_STONITH_DELAY_MAX, &props->delay_max[phase]); if (props->delay_max[phase]) { crm_trace("Peer %s with device %s returned maximum of random delay %d for %s", peer, device, props->delay_max[phase], action); } props->delay_base[phase] = 0; crm_element_value_int(xml, F_STONITH_DELAY_BASE, &props->delay_base[phase]); if (props->delay_base[phase]) { crm_trace("Peer %s with device %s returned base delay %d for %s", peer, device, props->delay_base[phase], action); } /* Handle devices with automatic unfencing */ if (pcmk__str_eq(action, "on", pcmk__str_casei)) { int required = 0; crm_element_value_int(xml, F_STONITH_DEVICE_REQUIRED, &required); if (required) { crm_trace("Peer %s requires device %s to execute for action %s", peer, device, action); add_required_device(op, device); } } /* If a reboot is remapped to off+on, it's possible that a node is allowed * to perform one action but not another. */ if (crm_is_true(crm_element_value(xml, F_STONITH_ACTION_DISALLOWED))) { props->disallowed[phase] = TRUE; crm_trace("Peer %s is disallowed from executing %s for device %s", peer, action, device); } } /*! * \internal * \brief Parse one device's properties from peer's XML query reply * * \param[in] xml XML node containing device properties * \param[in,out] op Operation that query and reply relate to * \param[in,out] result Peer's results * \param[in] device ID of device being parsed */ static void add_device_properties(xmlNode *xml, remote_fencing_op_t *op, st_query_result_t *result, const char *device) { xmlNode *child; int verified = 0; device_properties_t *props = calloc(1, sizeof(device_properties_t)); /* Add a new entry to this result's devices list */ CRM_ASSERT(props != NULL); g_hash_table_insert(result->devices, strdup(device), props); /* Peers with verified (monitored) access will be preferred */ crm_element_value_int(xml, F_STONITH_DEVICE_VERIFIED, &verified); if (verified) { crm_trace("Peer %s has confirmed a verified device %s", result->host, device); props->verified = TRUE; } /* Parse action-specific device properties */ parse_action_specific(xml, result->host, device, op_requested_action(op), op, st_phase_requested, props); for (child = pcmk__xml_first_child(xml); child != NULL; child = pcmk__xml_next(child)) { /* Replies for "reboot" operations will include the action-specific * values for "off" and "on" in child elements, just in case the reboot * winds up getting remapped. */ if (pcmk__str_eq(ID(child), "off", pcmk__str_casei)) { parse_action_specific(child, result->host, device, "off", op, st_phase_off, props); } else if (pcmk__str_eq(ID(child), "on", pcmk__str_casei)) { parse_action_specific(child, result->host, device, "on", op, st_phase_on, props); } } } /*! * \internal * \brief Parse a peer's XML query reply and add it to operation's results * * \param[in,out] op Operation that query and reply relate to * \param[in] host Name of peer that sent this reply * \param[in] ndevices Number of devices expected in reply * \param[in] xml XML node containing device list * * \return Newly allocated result structure with parsed reply */ static st_query_result_t * add_result(remote_fencing_op_t *op, const char *host, int ndevices, xmlNode *xml) { st_query_result_t *result = calloc(1, sizeof(st_query_result_t)); xmlNode *child; // cppcheck seems not to understand the abort logic in CRM_CHECK // cppcheck-suppress memleak CRM_CHECK(result != NULL, return NULL); result->host = strdup(host); result->devices = pcmk__strkey_table(free, free); /* Each child element describes one capable device available to the peer */ for (child = pcmk__xml_first_child(xml); child != NULL; child = pcmk__xml_next(child)) { const char *device = ID(child); if (device) { add_device_properties(child, op, result, device); } } result->ndevices = g_hash_table_size(result->devices); CRM_CHECK(ndevices == result->ndevices, crm_err("Query claimed to have %d device%s but %d found", ndevices, pcmk__plural_s(ndevices), result->ndevices)); op->query_results = g_list_insert_sorted(op->query_results, result, sort_peers); return result; } /*! * \internal * \brief Handle a peer's reply to our fencing query * * Parse a query result from XML and store it in the remote operation * table, and when enough replies have been received, issue a fencing request. * * \param[in] msg XML reply received * * \return pcmk_ok on success, -errno on error * * \note See initiate_remote_stonith_op() for how the XML query was initially * formed, and stonith_query() for how the peer formed its XML reply. */ int process_remote_stonith_query(xmlNode * msg) { int ndevices = 0; gboolean host_is_target = FALSE; gboolean have_all_replies = FALSE; const char *id = NULL; const char *host = NULL; remote_fencing_op_t *op = NULL; st_query_result_t *result = NULL; uint32_t replies_expected; xmlNode *dev = get_xpath_object("//@" F_STONITH_REMOTE_OP_ID, msg, LOG_ERR); CRM_CHECK(dev != NULL, return -EPROTO); id = crm_element_value(dev, F_STONITH_REMOTE_OP_ID); CRM_CHECK(id != NULL, return -EPROTO); dev = get_xpath_object("//@" F_STONITH_AVAILABLE_DEVICES, msg, LOG_ERR); CRM_CHECK(dev != NULL, return -EPROTO); crm_element_value_int(dev, F_STONITH_AVAILABLE_DEVICES, &ndevices); op = g_hash_table_lookup(stonith_remote_op_list, id); if (op == NULL) { crm_debug("Received query reply for unknown or expired operation %s", id); return -EOPNOTSUPP; } replies_expected = fencing_active_peers(); if (op->replies_expected < replies_expected) { replies_expected = op->replies_expected; } if ((++op->replies >= replies_expected) && (op->state == st_query)) { have_all_replies = TRUE; } host = crm_element_value(msg, F_ORIG); host_is_target = pcmk__str_eq(host, op->target, pcmk__str_casei); crm_info("Query result %d of %d from %s for %s/%s (%d device%s) %s", op->replies, replies_expected, host, op->target, op->action, ndevices, pcmk__plural_s(ndevices), id); if (ndevices > 0) { result = add_result(op, host, ndevices, dev); } if (pcmk_is_set(op->call_options, st_opt_topology)) { /* If we start the fencing before all the topology results are in, * it is possible fencing levels will be skipped because of the missing * query results. */ if (op->state == st_query && all_topology_devices_found(op)) { /* All the query results are in for the topology, start the fencing ops. */ crm_trace("All topology devices found"); call_remote_stonith(op, result, pcmk_ok); } else if (have_all_replies) { crm_info("All topology query replies have arrived, continuing (%d expected/%d received) ", replies_expected, op->replies); call_remote_stonith(op, NULL, pcmk_ok); } } else if (op->state == st_query) { int nverified = count_peer_devices(op, result, TRUE); /* We have a result for a non-topology fencing op that looks promising, * go ahead and start fencing before query timeout */ if (result && (host_is_target == FALSE) && nverified) { /* we have a verified device living on a peer that is not the target */ crm_trace("Found %d verified device%s", nverified, pcmk__plural_s(nverified)); call_remote_stonith(op, result, pcmk_ok); } else if (have_all_replies) { crm_info("All query replies have arrived, continuing (%d expected/%d received) ", replies_expected, op->replies); call_remote_stonith(op, NULL, pcmk_ok); } else { crm_trace("Waiting for more peer results before launching fencing operation"); } } else if (result && (op->state == st_done)) { crm_info("Discarding query result from %s (%d device%s): " "Operation is %s", result->host, result->ndevices, pcmk__plural_s(result->ndevices), stonith_op_state_str(op->state)); } return pcmk_ok; } /*! * \internal * \brief Handle a peer's reply to a fencing request * * Parse a fencing reply from XML, and either finalize the operation * or attempt another device as appropriate. * * \param[in] msg XML reply received * * \return pcmk_ok on success, -errno on error */ int process_remote_stonith_exec(xmlNode * msg) { int rc = 0; const char *id = NULL; const char *device = NULL; remote_fencing_op_t *op = NULL; xmlNode *dev = get_xpath_object("//@" F_STONITH_REMOTE_OP_ID, msg, LOG_ERR); CRM_CHECK(dev != NULL, return -EPROTO); id = crm_element_value(dev, F_STONITH_REMOTE_OP_ID); CRM_CHECK(id != NULL, return -EPROTO); dev = get_xpath_object("//@" F_STONITH_RC, msg, LOG_ERR); CRM_CHECK(dev != NULL, return -EPROTO); crm_element_value_int(dev, F_STONITH_RC, &rc); device = crm_element_value(dev, F_STONITH_DEVICE); if (stonith_remote_op_list) { op = g_hash_table_lookup(stonith_remote_op_list, id); } if (op == NULL && rc == pcmk_ok) { /* Record successful fencing operations */ const char *client_id = crm_element_value(dev, F_STONITH_CLIENTID); op = create_remote_stonith_op(client_id, dev, TRUE); } if (op == NULL) { /* Could be for an event that began before we started */ /* TODO: Record the op for later querying */ crm_info("Received peer result of unknown or expired operation %s", id); return -EOPNOTSUPP; } if (op->devices && device && !pcmk__str_eq(op->devices->data, device, pcmk__str_casei)) { crm_err("Received outdated reply for device %s (instead of %s) to " "fence (%s) %s. Operation already timed out at peer level.", device, (const char *) op->devices->data, op->action, op->target); return rc; } if (pcmk__str_eq(crm_element_value(msg, F_SUBTYPE), "broadcast", pcmk__str_casei)) { crm_debug("Finalizing action '%s' targeting %s on behalf of %s@%s: %s " CRM_XS " rc=%d id=%.8s", op->action, op->target, op->client_name, op->originator, pcmk_strerror(rc), rc, op->id); if (rc == pcmk_ok) { op->state = st_done; } else { op->state = st_failed; } remote_op_done(op, msg, rc, FALSE); return pcmk_ok; } else if (!pcmk__str_eq(op->originator, stonith_our_uname, pcmk__str_casei)) { /* If this isn't a remote level broadcast, and we are not the * originator of the operation, we should not be receiving this msg. */ crm_err("Received non-broadcast fencing result for operation %.8s " "we do not own (device %s targeting %s)", op->id, device, op->target); return rc; } if (pcmk_is_set(op->call_options, st_opt_topology)) { const char *device = crm_element_value(msg, F_STONITH_DEVICE); crm_notice("Action '%s' targeting %s using %s on behalf of %s@%s: %s " CRM_XS " rc=%d", op->action, op->target, device, op->client_name, op->originator, pcmk_strerror(rc), rc); /* We own the op, and it is complete. broadcast the result to all nodes * and notify our local clients. */ if (op->state == st_done) { remote_op_done(op, msg, rc, FALSE); return rc; } if ((op->phase == 2) && (rc != pcmk_ok)) { /* A remapped "on" failed, but the node was already turned off * successfully, so ignore the error and continue. */ crm_warn("Ignoring %s 'on' failure (exit code %d) targeting %s " "after successful 'off'", device, rc, op->target); rc = pcmk_ok; } if (rc == pcmk_ok) { /* An operation completed successfully. Try another device if * necessary, otherwise mark the operation as done. */ advance_topology_device_in_level(op, device, msg, rc); return rc; } else { /* This device failed, time to try another topology level. If no other * levels are available, mark this operation as failed and report results. */ if (advance_topology_level(op, false) != pcmk_rc_ok) { op->state = st_failed; remote_op_done(op, msg, rc, FALSE); return rc; } } } else if (rc == pcmk_ok && op->devices == NULL) { crm_trace("All done for %s", op->target); op->state = st_done; remote_op_done(op, msg, rc, FALSE); return rc; } else if (rc == -ETIME && op->devices == NULL) { /* If the operation timed out don't bother retrying other peers. */ op->state = st_failed; remote_op_done(op, msg, rc, FALSE); return rc; } else { /* fall-through and attempt other fencing action using another peer */ } /* Retry on failure */ crm_trace("Next for %s on behalf of %s@%s (rc was %d)", op->target, op->originator, op->client_name, rc); call_remote_stonith(op, NULL, rc); return rc; } gboolean stonith_check_fence_tolerance(int tolerance, const char *target, const char *action) { GHashTableIter iter; time_t now = time(NULL); remote_fencing_op_t *rop = NULL; if (tolerance <= 0 || !stonith_remote_op_list || target == NULL || action == NULL) { return FALSE; } g_hash_table_iter_init(&iter, stonith_remote_op_list); while (g_hash_table_iter_next(&iter, NULL, (void **)&rop)) { if (strcmp(rop->target, target) != 0) { continue; } else if (rop->state != st_done) { continue; /* We don't have to worry about remapped reboots here * because if state is done, any remapping has been undone */ } else if (strcmp(rop->action, action) != 0) { continue; } else if ((rop->completed + tolerance) < now) { continue; } crm_notice("Target %s was fenced (%s) less than %ds ago by %s on behalf of %s", target, action, tolerance, rop->delegate, rop->originator); return TRUE; } return FALSE; } diff --git a/include/crm/fencing/internal.h b/include/crm/fencing/internal.h index fa9059e6f6..6a7e4232c0 100644 --- a/include/crm/fencing/internal.h +++ b/include/crm/fencing/internal.h @@ -1,201 +1,202 @@ /* * Copyright 2011-2021 the Pacemaker project contributors * * The version control history for this file may have further details. * * This source code is licensed under the GNU Lesser General Public License * version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY. */ #ifndef STONITH_NG_INTERNAL__H # define STONITH_NG_INTERNAL__H # include # include # include # include # include enum st_device_flags { st_device_supports_list = 0x0001, st_device_supports_status = 0x0002, st_device_supports_reboot = 0x0004, st_device_supports_parameter_plug = 0x0008, st_device_supports_parameter_port = 0x0010, }; #define stonith__set_device_flags(device_flags, device_id, flags_to_set) do { \ device_flags = pcmk__set_flags_as(__func__, __LINE__, LOG_TRACE, \ "Fence device", device_id, \ (device_flags), (flags_to_set), \ #flags_to_set); \ } while (0) #define stonith__set_call_options(st_call_opts, call_for, flags_to_set) do { \ st_call_opts = pcmk__set_flags_as(__func__, __LINE__, LOG_TRACE, \ "Fencer call", (call_for), \ (st_call_opts), (flags_to_set), \ #flags_to_set); \ } while (0) #define stonith__clear_call_options(st_call_opts, call_for, flags_to_clear) do { \ st_call_opts = pcmk__clear_flags_as(__func__, __LINE__, LOG_TRACE, \ "Fencer call", (call_for), \ (st_call_opts), (flags_to_clear), \ #flags_to_clear); \ } while (0) struct stonith_action_s; typedef struct stonith_action_s stonith_action_t; stonith_action_t *stonith_action_create(const char *agent, const char *_action, const char *victim, uint32_t victim_nodeid, int timeout, GHashTable * device_args, GHashTable * port_map, const char * host_arg); void stonith__destroy_action(stonith_action_t *action); -void stonith__action_result(stonith_action_t *action, int *rc, char **output, - char **error_output); +pcmk__action_result_t *stonith__action_result(stonith_action_t *action); +int stonith__result2rc(const pcmk__action_result_t *result); int stonith_action_execute_async(stonith_action_t * action, void *userdata, - void (*done) (int pid, int rc, const char *output, + void (*done) (int pid, + const pcmk__action_result_t *result, void *user_data), void (*fork_cb) (int pid, void *user_data)); xmlNode *create_level_registration_xml(const char *node, const char *pattern, const char *attr, const char *value, int level, stonith_key_value_t *device_list); xmlNode *create_device_registration_xml(const char *id, enum stonith_namespace namespace, const char *agent, stonith_key_value_t *params, const char *rsc_provides); void stonith__register_messages(pcmk__output_t *out); GList *stonith__parse_targets(const char *hosts); gboolean stonith__later_succeeded(stonith_history_t *event, stonith_history_t *top_history); stonith_history_t *stonith__sort_history(stonith_history_t *history); void stonith__device_parameter_flags(uint32_t *device_flags, const char *device_name, xmlNode *metadata); # define ST_LEVEL_MAX 10 # define F_STONITH_CLIENTID "st_clientid" # define F_STONITH_CALLOPTS "st_callopt" # define F_STONITH_CALLID "st_callid" # define F_STONITH_CALLDATA "st_calldata" # define F_STONITH_OPERATION "st_op" # define F_STONITH_TARGET "st_target" # define F_STONITH_REMOTE_OP_ID "st_remote_op" # define F_STONITH_REMOTE_OP_ID_RELAY "st_remote_op_relay" # define F_STONITH_RC "st_rc" /*! Timeout period per a device execution */ # define F_STONITH_TIMEOUT "st_timeout" # define F_STONITH_TOLERANCE "st_tolerance" # define F_STONITH_DELAY "st_delay" /*! Action specific timeout period returned in query of fencing devices. */ # define F_STONITH_ACTION_TIMEOUT "st_action_timeout" /*! Host in query result is not allowed to run this action */ # define F_STONITH_ACTION_DISALLOWED "st_action_disallowed" /*! Maximum of random fencing delay for a device */ # define F_STONITH_DELAY_MAX "st_delay_max" /*! Base delay used for a fencing delay */ # define F_STONITH_DELAY_BASE "st_delay_base" /*! Has this device been verified using a monitor type * operation (monitor, list, status) */ # define F_STONITH_DEVICE_VERIFIED "st_monitor_verified" /*! device is required for this action */ # define F_STONITH_DEVICE_REQUIRED "st_required" /*! number of available devices in query result */ # define F_STONITH_AVAILABLE_DEVICES "st-available-devices" # define F_STONITH_CALLBACK_TOKEN "st_async_id" # define F_STONITH_CLIENTNAME "st_clientname" # define F_STONITH_CLIENTNODE "st_clientnode" # define F_STONITH_NOTIFY_ACTIVATE "st_notify_activate" # define F_STONITH_NOTIFY_DEACTIVATE "st_notify_deactivate" # define F_STONITH_DELEGATE "st_delegate" /*! The node initiating the stonith operation. If an operation * is relayed, this is the last node the operation lands on. When * in standalone mode, origin is the client's id that originated the * operation. */ # define F_STONITH_ORIGIN "st_origin" # define F_STONITH_HISTORY_LIST "st_history" # define F_STONITH_DATE "st_date" # define F_STONITH_DATE_NSEC "st_date_nsec" # define F_STONITH_STATE "st_state" # define F_STONITH_ACTIVE "st_active" # define F_STONITH_DIFFERENTIAL "st_differential" # define F_STONITH_DEVICE "st_device_id" # define F_STONITH_ACTION "st_device_action" # define F_STONITH_MERGED "st_op_merged" # define T_STONITH_NG "stonith-ng" # define T_STONITH_REPLY "st-reply" /*! For async operations, an event from the server containing * the total amount of time the server is allowing for the operation * to take place is returned to the client. */ # define T_STONITH_TIMEOUT_VALUE "st-async-timeout-value" # define T_STONITH_NOTIFY "st_notify" # define STONITH_ATTR_ACTION_OP "action" # define STONITH_OP_EXEC "st_execute" # define STONITH_OP_TIMEOUT_UPDATE "st_timeout_update" # define STONITH_OP_QUERY "st_query" # define STONITH_OP_FENCE "st_fence" # define STONITH_OP_RELAY "st_relay" # define STONITH_OP_DEVICE_ADD "st_device_register" # define STONITH_OP_DEVICE_DEL "st_device_remove" # define STONITH_OP_FENCE_HISTORY "st_fence_history" # define STONITH_OP_LEVEL_ADD "st_level_add" # define STONITH_OP_LEVEL_DEL "st_level_remove" # define STONITH_WATCHDOG_AGENT "fence_watchdog" /* Don't change 2 below as it would break rolling upgrade */ # define STONITH_WATCHDOG_AGENT_INTERNAL "#watchdog" # define STONITH_WATCHDOG_ID "watchdog" /* Exported for crm_mon to reference */ int stonith__failed_history(pcmk__output_t *out, va_list args); int stonith__history(pcmk__output_t *out, va_list args); int stonith__full_history(pcmk__output_t *out, va_list args); int stonith__pending_actions(pcmk__output_t *out, va_list args); stonith_history_t *stonith__first_matching_event(stonith_history_t *history, bool (*matching_fn)(stonith_history_t *, void *), void *user_data); bool stonith__event_state_pending(stonith_history_t *history, void *user_data); bool stonith__event_state_eq(stonith_history_t *history, void *user_data); bool stonith__event_state_neq(stonith_history_t *history, void *user_data); /*! * \internal * \brief Is a fencing operation in pending state? * * \param[in] state State as enum op_state value * * \return A boolean */ static inline bool stonith__op_state_pending(enum op_state state) { return state != st_failed && state != st_done; } gboolean stonith__watchdog_fencing_enabled_for_node(const char *node); gboolean stonith__watchdog_fencing_enabled_for_node_api(stonith_t *st, const char *node); #endif diff --git a/lib/fencing/st_client.c b/lib/fencing/st_client.c index 08adb51c6c..2fbff7f247 100644 --- a/lib/fencing/st_client.c +++ b/lib/fencing/st_client.c @@ -1,2778 +1,2782 @@ /* * Copyright 2004-2021 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 #include +#include #include #include "fencing_private.h" 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) (int pid, int status, const char *output, void *user_data); + void (*done_cb) (int pid, const pcmk__action_result_t *result, + void *user_data); void (*fork_cb) (int pid, void *user_data); svc_action_t *svc_action; /*! internal timing information */ time_t initial_start_time; int tries; int remaining_timeout; int max_retries; int pid; - int rc; - char *output; - char *error; + pcmk__action_result_t result; }; 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 (pcmk__str_eq(namespace_s, "any", pcmk__str_null_matches)) { 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 (pcmk__str_eq(namespace_s, "internal", pcmk__str_casei)) { 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; } +/*! + * \internal + * \brief Set an action's result based on services library result + * + * \param[in] action Fence action to set result for + * \param[in] svc_action Service action to get result from + */ +static void +set_result_from_svc_action(stonith_action_t *action, svc_action_t *svc_action) +{ + pcmk__set_result(&(action->result), svc_action->rc, svc_action->status, + services__exit_reason(svc_action)); + pcmk__set_result_output(&(action->result), + services__grab_stdout(svc_action), + services__grab_stderr(svc_action)); +} + gboolean stonith__watchdog_fencing_enabled_for_node_api(stonith_t *st, const char *node) { gboolean rv = FALSE; stonith_t *stonith_api = st?st:stonith_api_new(); char *list = NULL; if(stonith_api) { if (stonith_api->state == stonith_disconnected) { int rc = stonith_api->cmds->connect(stonith_api, "stonith-api", NULL); if (rc != pcmk_ok) { crm_err("Failed connecting to Stonith-API for watchdog-fencing-query."); } } if (stonith_api->state != stonith_disconnected) { /* caveat!!! * this might fail when when stonithd is just updating the device-list * probably something we should fix as well for other api-calls */ int rc = stonith_api->cmds->list(stonith_api, st_opt_sync_call, STONITH_WATCHDOG_ID, &list, 0); if ((rc != pcmk_ok) || (list == NULL)) { /* due to the race described above it can happen that * we drop in here - so as not to make remote nodes * panic on that answer */ crm_warn("watchdog-fencing-query failed"); } else if (list[0] == '\0') { rv = TRUE; } else { GList *targets = stonith__parse_targets(list); rv = pcmk__str_in_list(node, targets, pcmk__str_casei); g_list_free_full(targets, free); } free(list); if (!st) { /* if we're provided the api we still might have done the * connection - but let's assume the caller won't bother */ stonith_api->cmds->disconnect(stonith_api); } } if (!st) { stonith_api_delete(stonith_api); } } else { crm_err("Stonith-API for watchdog-fencing-query couldn't be created."); } crm_trace("Pacemaker assumes node %s %sto do watchdog-fencing.", node, rv?"":"not "); return rv; } gboolean stonith__watchdog_fencing_enabled_for_node(const char *node) { return stonith__watchdog_fencing_enabled_for_node_api(NULL, node); } 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) { + /* The services library has already logged the output at info or debug + * level, so just raise to warning for stderr. + */ + if (action->result.action_stderr != NULL) { /* 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); + crm_log_output(LOG_WARNING, prefix, action->result.action_stderr); 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, __func__); 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, __func__); 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, __func__); 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) { size_t 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, __func__); 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); } // cppcheck seems not to understand the abort logic behind pcmk__realloc // cppcheck-suppress memleak for (; device_list; device_list = device_list->next) { pcmk__add_separated_word(&list, &len, device_list->value, ","); } 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_config_arg(gpointer key, gpointer value, gpointer user_data) { /* The fencer will filter "action" out when it registers the device, * but ignore it here in case any external API users don't. * * Also filter out parameters handled directly by Pacemaker. */ if (!pcmk__str_eq(key, STONITH_ATTR_ACTION_OP, pcmk__str_casei) && !pcmk_stonith_param(key) && (strstr(key, CRM_META) == NULL) && !pcmk__str_eq(key, "crm_feature_set", pcmk__str_casei)) { crm_trace("Passing %s=%s with fence action", (const char *) key, (const char *) (value? value : "")); g_hash_table_insert((GHashTable *) user_data, strdup(key), strdup(value? value : "")); } } static GHashTable * make_args(const char *agent, const char *action, const char *victim, uint32_t victim_nodeid, GHashTable * device_args, GHashTable * port_map, const char *host_arg) { GHashTable *arg_list = NULL; const char *value = NULL; CRM_CHECK(action != NULL, return NULL); arg_list = pcmk__strkey_table(free, free); // Add action to arguments (using an alias if requested) if (device_args) { char buffer[512]; snprintf(buffer, sizeof(buffer), "pcmk_%s_action", action); value = g_hash_table_lookup(device_args, buffer); if (value) { crm_debug("Substituting '%s' for fence action %s targeting %s", value, action, victim); action = value; } } g_hash_table_insert(arg_list, strdup(STONITH_ATTR_ACTION_OP), strdup(action)); /* If this is a fencing operation against another node, add more standard * arguments. */ if (victim && device_args) { const char *param = NULL; /* Always pass the target's name, per * https://github.com/ClusterLabs/fence-agents/blob/master/doc/FenceAgentAPI.md */ g_hash_table_insert(arg_list, strdup("nodename"), strdup(victim)); // If the target's node ID was specified, pass it, too if (victim_nodeid) { char *nodeid = crm_strdup_printf("%" PRIu32, victim_nodeid); // cts-fencing looks for this log message crm_info("Passing '%s' as nodeid with fence action '%s' targeting %s", nodeid, action, victim); g_hash_table_insert(arg_list, strdup("nodeid"), nodeid); } // Check whether target must be specified in some other way param = g_hash_table_lookup(device_args, PCMK_STONITH_HOST_ARGUMENT); if (!pcmk__str_eq(agent, "fence_legacy", pcmk__str_none) && !pcmk__str_eq(param, "none", pcmk__str_casei)) { if (param == NULL) { /* Use the caller's default for pcmk_host_argument, or "port" if * none was given */ param = (host_arg == NULL)? "port" : host_arg; } value = g_hash_table_lookup(device_args, param); if (pcmk__str_eq(value, "dynamic", pcmk__str_casei|pcmk__str_null_matches)) { /* If the host argument was "dynamic" or not explicitly specified, * add it with the target */ const char *alias = NULL; if (port_map) { alias = g_hash_table_lookup(port_map, victim); } if (alias == NULL) { alias = victim; } crm_debug("Passing %s='%s' with fence action %s targeting %s", param, alias, action, victim); g_hash_table_insert(arg_list, strdup(param), strdup(alias)); } } } 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); + pcmk__reset_result(&(action->result)); 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) + * \param[in] action Executed action * - * \note If output or error_output is not NULL, the caller is responsible for - * freeing the memory. + * \return Pointer to action's result (or NULL if \p action is NULL) */ -void -stonith__action_result(stonith_action_t *action, int *rc, char **output, - char **error_output) +pcmk__action_result_t * +stonith__action_result(stonith_action_t *action) { - 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 - } - } + return (action == NULL)? NULL : &(action->result); } #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, const char *host_arg) { stonith_action_t *action; action = calloc(1, sizeof(stonith_action_t)); action->args = make_args(agent, _action, victim, victim_nodeid, device_args, port_map, host_arg); 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; + pcmk__set_result(&(action->result), PCMK_OCF_UNKNOWN, PCMK_EXEC_UNKNOWN, + "Initialization bug in fencing library"); + 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)) { + } else if ((action->result.execution_status != PCMK_EXEC_TIMEOUT) + && (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; +/*! + * \internal + * \brief Map a fencing action result to a standard return code + * + * \param[in] result Fencing action result to map + * + * \return Standard Pacemaker return code that best corresponds to \p result + */ +int +stonith__result2rc(const pcmk__action_result_t *result) +{ + switch (result->execution_status) { + case PCMK_EXEC_CANCELLED: return ECANCELED; + case PCMK_EXEC_TIMEOUT: return ETIME; + case PCMK_EXEC_NOT_INSTALLED: return ENOENT; + case PCMK_EXEC_NOT_SUPPORTED: return EOPNOTSUPP; + case PCMK_EXEC_NOT_CONNECTED: return ENOTCONN; + case PCMK_EXEC_NO_FENCE_DEVICE: return ENODEV; + case PCMK_EXEC_NO_SECRETS: return EACCES; + default: break; + } - if (svc_action->status == PCMK_EXEC_TIMEOUT) { - rv = -ETIME; + if (result->exit_status == CRM_EX_OK) { + return pcmk_rc_ok; + } - } else if (svc_action->rc != PCMK_OCF_OK) { - /* Try to provide a useful error code based on the fence agent's - * error output. - */ - if (svc_action->stderr_data == NULL) { - rv = -ENODATA; + // Try to provide useful error code based on result's error output - } else if (strstr(svc_action->stderr_data, "imed out")) { - /* Some agents have their own internal timeouts */ - rv = -ETIME; + if (result->action_stderr == NULL) { + return ENODATA; - } else if (strstr(svc_action->stderr_data, "Unrecognised action")) { - rv = -EOPNOTSUPP; + } else if (strcasestr(result->action_stderr, "timed out") + || strcasestr(result->action_stderr, "timeout")) { + return ETIME; - } else { - rv = -pcmk_err_generic; - } + } else if (strcasestr(result->action_stderr, "unrecognised action") + || strcasestr(result->action_stderr, "unrecognized action") + || strcasestr(result->action_stderr, "unsupported action")) { + return EOPNOTSUPP; } - return rv; + + // Oh well, we tried + return pcmk_rc_error; } 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; + set_result_from_svc_action(action, svc_action); 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)) { + if ((action->result.exit_status != CRM_EX_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->done_cb(action->pid, &(action->result), 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 == NULL) || (action->action == NULL) || (action->args == NULL) + CRM_CHECK(action != NULL, return -EINVAL); + + if ((action->action == NULL) || (action->args == NULL) || (action->agent == NULL)) { - return -EPROTO; + pcmk__set_result(&(action->result), PCMK_OCF_UNKNOWN_ERROR, + PCMK_EXEC_ERROR_FATAL, "Bug in fencing library"); + return -EINVAL; } 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; } buffer = crm_strdup_printf(PCMK__FENCE_BINDIR "/%s", basename(action->agent)); svc_action = services_action_create_generic(buffer, NULL); free(buffer); if (svc_action->rc != PCMK_OCF_UNKNOWN) { + set_result_from_svc_action(action, svc_action); services_action_free(svc_action); return -E2BIG; } 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; svc_action->flags = pcmk__set_flags_as(__func__, __LINE__, LOG_TRACE, "Action", svc_action->id, svc_action->flags, SVC_ACTION_NON_BLOCKED, "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; + pcmk__reset_result(&(action->result)); sleep(1); } if (action->async) { /* async */ if (services_action_async_fork_notify(svc_action, &stonith_action_async_done, &stonith_action_async_forked)) { + pcmk__set_result(&(action->result), PCMK_OCF_UNKNOWN, + PCMK_EXEC_PENDING, NULL); return pcmk_ok; } } else if (services_action_sync(svc_action)) { // sync success rc = pcmk_ok; - 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 { // sync failure - action->rc = -ECONNABORTED; - rc = action->rc; + rc = -ECONNABORTED; } + set_result_from_svc_action(action, svc_action); svc_action->params = NULL; services_action_free(svc_action); 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) (int pid, int rc, const char *output, + void (*done) (int pid, + const pcmk__action_result_t *result, void *user_data), void (*fork_cb) (int pid, void *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, __func__); 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, __func__); 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_with_delay(stonith_t * stonith, int call_options, const char *node, const char *action, int timeout, int tolerance, int delay) { int rc = 0; xmlNode *data = NULL; data = create_xml_node(NULL, __func__); 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); crm_xml_add_int(data, F_STONITH_DELAY, delay); rc = stonith_send_command(stonith, STONITH_OP_FENCE, data, NULL, call_options, timeout); free_xml(data); return rc; } static int stonith_api_fence(stonith_t * stonith, int call_options, const char *node, const char *action, int timeout, int tolerance) { return stonith_api_fence_with_delay(stonith, call_options, node, action, timeout, tolerance, 0); } static int stonith_api_confirm(stonith_t * stonith, int call_options, const char *target) { stonith__set_call_options(call_options, target, st_opt_manual_ack); return stonith_api_fence(stonith, call_options, 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, __func__); crm_xml_add(data, F_STONITH_TARGET, node); } stonith__set_call_options(call_options, node, st_opt_sync_call); rc = stonith_send_command(stonith, STONITH_OP_FENCE_HISTORY, data, &output, call_options, timeout); free_xml(data); if (rc == 0) { xmlNode *op = NULL; xmlNode *reply = get_xpath_object("//" F_STONITH_HISTORY_LIST, output, LOG_NEVER); for (op = pcmk__xml_first_child(reply); op != NULL; op = pcmk__xml_next(op)) { stonith_history_t *kvp; long long completed; long long completed_nsec = 0L; 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_ll(op, F_STONITH_DATE, &completed); kvp->completed = (time_t) completed; crm_element_value_ll(op, F_STONITH_DATE_NSEC, &completed_nsec); kvp->completed_nsec = completed_nsec; crm_element_value_int(op, F_STONITH_STATE, &kvp->state); if (last) { last->next = kvp; } else { *history = kvp; } last = kvp; } } free_xml(output); return rc; } void stonith_history_free(stonith_history_t *history) { stonith_history_t *hp, *hp_old; for (hp = history; hp; hp_old = hp, hp = hp->next, free(hp_old)) { free(hp->target); free(hp->action); free(hp->origin); free(hp->delegate); free(hp->client); } } static gint stonithlib_GCompareFunc(gconstpointer a, gconstpointer b) { int rc = 0; const stonith_notify_client_t *a_client = a; const stonith_notify_client_t *b_client = b; if (a_client->delete || b_client->delete) { /* make entries marked for deletion not findable */ return -1; } CRM_CHECK(a_client->event != NULL && b_client->event != NULL, return 0); rc = strcmp(a_client->event, b_client->event); if (rc == 0) { if (a_client->notify == NULL || b_client->notify == NULL) { return 0; } else if (a_client->notify == b_client->notify) { return 0; } else if (((long)a_client->notify) < ((long)b_client->notify)) { crm_err("callbacks for %s are not equal: %p vs. %p", a_client->event, a_client->notify, b_client->notify); return -1; } crm_err("callbacks for %s are not equal: %p vs. %p", a_client->event, a_client->notify, b_client->notify); return 1; } return rc; } xmlNode * stonith_create_op(int call_id, const char *token, const char *op, xmlNode * data, int call_options) { xmlNode *op_msg = create_xml_node(NULL, "stonith_command"); CRM_CHECK(op_msg != NULL, return NULL); CRM_CHECK(token != NULL, return NULL); crm_xml_add(op_msg, F_XML_TAGNAME, "stonith_command"); crm_xml_add(op_msg, F_TYPE, T_STONITH_NG); crm_xml_add(op_msg, F_STONITH_CALLBACK_TOKEN, token); crm_xml_add(op_msg, F_STONITH_OPERATION, op); crm_xml_add_int(op_msg, F_STONITH_CALLID, call_id); crm_trace("Sending call options: %.8lx, %d", (long)call_options, call_options); crm_xml_add_int(op_msg, F_STONITH_CALLOPTS, call_options); if (data != NULL) { add_message_xml(op_msg, F_STONITH_CALLDATA, data); } return op_msg; } static void stonith_destroy_op_callback(gpointer data) { stonith_callback_client_t *blob = data; if (blob->timer && blob->timer->ref > 0) { g_source_remove(blob->timer->ref); } free(blob->timer); free(blob); } static int stonith_api_signoff(stonith_t * stonith) { stonith_private_t *native = stonith->st_private; crm_debug("Disconnecting from the fencer"); if (native->source != NULL) { /* Attached to mainloop */ mainloop_del_ipc_client(native->source); native->source = NULL; native->ipc = NULL; } else if (native->ipc) { /* Not attached to mainloop */ crm_ipc_t *ipc = native->ipc; native->ipc = NULL; crm_ipc_close(ipc); crm_ipc_destroy(ipc); } free(native->token); native->token = NULL; stonith->state = stonith_disconnected; return pcmk_ok; } static int stonith_api_del_callback(stonith_t * stonith, int call_id, bool all_callbacks) { stonith_private_t *private = stonith->st_private; if (all_callbacks) { private->op_callback = NULL; g_hash_table_destroy(private->stonith_op_callback_table); private->stonith_op_callback_table = pcmk__intkey_table(stonith_destroy_op_callback); } else if (call_id == 0) { private->op_callback = NULL; } else { pcmk__intkey_table_remove(private->stonith_op_callback_table, 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 = pcmk__intkey_table_lookup(private->stonith_op_callback_table, 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 = pcmk__intkey_table_lookup(private->stonith_op_callback_table, 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 (pcmk__str_eq(type, T_STONITH_NG, pcmk__str_casei)) { stonith_perform_callback(st, blob.xml, 0, 0); } else if (pcmk__str_eq(type, T_STONITH_NOTIFY, pcmk__str_casei)) { foreach_notify_entry(private, stonith_send_notification, &blob); } else if (pcmk__str_eq(type, T_STONITH_TIMEOUT_VALUE, pcmk__str_casei)) { int call_id = 0; int timeout = 0; crm_element_value_int(blob.xml, F_STONITH_TIMEOUT, &timeout); crm_element_value_int(blob.xml, F_STONITH_CALLID, &call_id); update_callback_timeout(call_id, timeout, st); } else { crm_err("Unknown message type: %s", type); crm_log_xml_warn(blob.xml, "BadReply"); } free_xml(blob.xml); return 1; } static int stonith_api_signon(stonith_t * stonith, const char *name, int *stonith_fd) { int rc = pcmk_ok; stonith_private_t *native = NULL; const char *display_name = name? name : "client"; struct ipc_client_callbacks st_callbacks = { .dispatch = stonith_dispatch_internal, .destroy = stonith_connection_destroy }; CRM_CHECK(stonith != NULL, return -EINVAL); native = stonith->st_private; CRM_ASSERT(native != NULL); crm_debug("Attempting fencer connection by %s with%s mainloop", display_name, (stonith_fd? "out" : "")); stonith->state = stonith_connected_command; if (stonith_fd) { /* No mainloop */ native->ipc = crm_ipc_new("stonith-ng", 0); if (native->ipc && crm_ipc_connect(native->ipc)) { *stonith_fd = crm_ipc_get_fd(native->ipc); } else if (native->ipc) { crm_ipc_close(native->ipc); crm_ipc_destroy(native->ipc); native->ipc = NULL; } } else { /* With mainloop */ native->source = mainloop_add_ipc_client("stonith-ng", G_PRIORITY_MEDIUM, 0, stonith, &st_callbacks); native->ipc = mainloop_get_ipc_client(native->source); } if (native->ipc == NULL) { rc = -ENOTCONN; } else { xmlNode *reply = NULL; xmlNode *hello = create_xml_node(NULL, "stonith_command"); crm_xml_add(hello, F_TYPE, T_STONITH_NG); crm_xml_add(hello, F_STONITH_OPERATION, CRM_OP_REGISTER); crm_xml_add(hello, F_STONITH_CLIENTNAME, name); rc = crm_ipc_send(native->ipc, hello, crm_ipc_client_response, -1, &reply); if (rc < 0) { crm_debug("Couldn't register with the fencer: %s " CRM_XS " rc=%d", pcmk_strerror(rc), rc); rc = -ECOMM; } else if (reply == NULL) { crm_debug("Couldn't register with the fencer: no reply"); rc = -EPROTO; } else { const char *msg_type = crm_element_value(reply, F_STONITH_OPERATION); native->token = crm_element_value_copy(reply, F_STONITH_CLIENTID); if (!pcmk__str_eq(msg_type, CRM_OP_REGISTER, pcmk__str_casei)) { 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 = PCMK__IPC_TIMEOUT; #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, __func__); 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); } pcmk__intkey_table_insert(private->stonith_op_callback_table, 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 (pcmk__str_eq(ntype, T_STONITH_NOTIFY_FENCE, pcmk__str_casei)) { 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 (!pcmk__str_eq(entry->event, event, pcmk__str_casei)) { crm_trace("Skipping callback - event mismatch %p/%s vs. %s", entry, entry->event, event); return; } st_event = xml_to_event(blob->xml); crm_trace("Invoking callback for %p/%s event...", entry, event); entry->notify(blob->stonith, st_event); crm_trace("Callback invoked..."); event_free(st_event); } /*! * \internal * \brief Create and send an API request * * \param[in] stonith Stonith connection * \param[in] op API operation to request * \param[in] data Data to attach to request * \param[out] output_data If not NULL, will be set to reply if synchronous * \param[in] call_options Bitmask of stonith_call_options to use * \param[in] timeout Error if not completed within this many seconds * * \return pcmk_ok (for synchronous requests) or positive call ID * (for asynchronous requests) on success, -errno otherwise */ static int stonith_send_command(stonith_t * stonith, const char *op, xmlNode * data, xmlNode ** output_data, int call_options, int timeout) { int rc = 0; int reply_id = -1; xmlNode *op_msg = NULL; xmlNode *op_reply = NULL; stonith_private_t *native = NULL; CRM_ASSERT(stonith && stonith->st_private && op); native = stonith->st_private; if (output_data != NULL) { *output_data = NULL; } if ((stonith->state == stonith_disconnected) || (native->token == NULL)) { return -ENOTCONN; } /* Increment the call ID, which must be positive to avoid conflicting with * error codes. This shouldn't be a problem unless the client mucked with * it or the counter wrapped around. */ stonith->call_id++; if (stonith->call_id < 1) { stonith->call_id = 1; } op_msg = stonith_create_op(stonith->call_id, native->token, op, data, call_options); if (op_msg == NULL) { return -EINVAL; } crm_xml_add_int(op_msg, F_STONITH_TIMEOUT, timeout); crm_trace("Sending %s message to fencer with timeout %ds", op, timeout); if (data) { const char *delay_s = crm_element_value(data, F_STONITH_DELAY); if (delay_s) { crm_xml_add(op_msg, F_STONITH_DELAY, delay_s); } } { enum crm_ipc_flags ipc_flags = crm_ipc_flags_none; if (call_options & st_opt_sync_call) { pcmk__set_ipc_flags(ipc_flags, "stonith command", 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"; const char *host_arg = NULL; GHashTable *params_table = pcmk__strkey_table(free, free); // Convert parameter list to a hash table for (; params; params = params->next) { if (pcmk__str_eq(params->key, PCMK_STONITH_HOST_ARGUMENT, pcmk__str_casei)) { host_arg = params->value; } if (!pcmk_stonith_param(params->key)) { g_hash_table_insert(params_table, strdup(params->key), strdup(params->value)); } } #if SUPPORT_CIBSECRETS rc = pcmk__substitute_secrets(rsc_id, params_table); if (rc != pcmk_rc_ok) { crm_warn("Could not replace secret parameters for validation of %s: %s", agent, pcmk_rc_str(rc)); // rc is standard return value, don't return it in this function } #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, host_arg, 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 = pcmk__intkey_table(stonith_destroy_op_callback); private->notify_list = NULL; private->notify_refcnt = 0; private->notify_deletes = FALSE; new_stonith->call_id = 1; new_stonith->state = stonith_disconnected; new_stonith->cmds = calloc(1, sizeof(stonith_api_operations_t)); if (new_stonith->cmds == NULL) { free(new_stonith->st_private); free(new_stonith); return NULL; } /* *INDENT-OFF* */ new_stonith->cmds->free = stonith_api_free; new_stonith->cmds->connect = stonith_api_signon; new_stonith->cmds->disconnect = stonith_api_signoff; new_stonith->cmds->list = stonith_api_list; new_stonith->cmds->monitor = stonith_api_monitor; new_stonith->cmds->status = stonith_api_status; new_stonith->cmds->fence = stonith_api_fence; new_stonith->cmds->fence_with_delay = stonith_api_fence_with_delay; new_stonith->cmds->confirm = stonith_api_confirm; new_stonith->cmds->history = stonith_api_history; new_stonith->cmds->list_agents = stonith_api_device_list; new_stonith->cmds->metadata = stonith_api_device_metadata; new_stonith->cmds->query = stonith_api_query; new_stonith->cmds->remove_device = stonith_api_remove_device; new_stonith->cmds->register_device = stonith_api_register_device; new_stonith->cmds->remove_level = stonith_api_remove_level; new_stonith->cmds->remove_level_full = stonith_api_remove_level_full; new_stonith->cmds->register_level = stonith_api_register_level; new_stonith->cmds->register_level_full = stonith_api_register_level_full; new_stonith->cmds->remove_callback = stonith_api_del_callback; new_stonith->cmds->register_callback = stonith_api_add_callback; new_stonith->cmds->remove_notification = stonith_api_del_notification; new_stonith->cmds->register_notification = stonith_api_add_notification; new_stonith->cmds->validate = stonith_api_validate; /* *INDENT-ON* */ return new_stonith; } /*! * \brief Make a blocking connection attempt to the fencer * * \param[in,out] st Fencer API object * \param[in] name Client name to use with fencer * \param[in] max_attempts Return error if this many attempts fail * * \return pcmk_ok on success, result of last attempt otherwise */ int stonith_api_connect_retry(stonith_t *st, const char *name, int max_attempts) { int rc = -EINVAL; // if max_attempts is not positive for (int attempt = 1; attempt <= max_attempts; attempt++) { rc = st->cmds->connect(st, name, NULL); if (rc == pcmk_ok) { return pcmk_ok; } else if (attempt < max_attempts) { crm_notice("Fencer connection attempt %d of %d failed (retrying in 2s): %s " CRM_XS " rc=%d", attempt, max_attempts, pcmk_strerror(rc), rc); sleep(2); } } crm_notice("Could not connect to fencer: %s " CRM_XS " rc=%d", pcmk_strerror(rc), rc); return rc; } stonith_key_value_t * stonith_key_value_add(stonith_key_value_t * head, const char *key, const char *value) { stonith_key_value_t *p, *end; p = calloc(1, sizeof(stonith_key_value_t)); if (key) { p->key = strdup(key); } if (value) { p->value = strdup(value); } end = head; while (end && end->next) { end = end->next; } if (end) { end->next = p; } else { head = p; } return head; } void stonith_key_value_freeall(stonith_key_value_t * head, int keys, int values) { stonith_key_value_t *p; while (head) { p = head->next; if (keys) { free(head->key); } if (values) { free(head->value); } free(head); head = p; } } #define api_log_open() openlog("stonith-api", LOG_CONS | LOG_NDELAY | LOG_PID, LOG_DAEMON) #define api_log(level, fmt, args...) syslog(level, "%s: "fmt, __func__, 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 = (uname == NULL)? pcmk__itoa(nodeid) : strdup(uname); int opts = 0; stonith__set_call_options(opts, name, st_opt_sync_call|st_opt_allow_suicide); if ((uname == NULL) && (nodeid > 0)) { stonith__set_call_options(opts, name, st_opt_cs_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; int opts = 0; char *name = (uname == NULL)? pcmk__itoa(nodeid) : strdup(uname); stonith__set_call_options(opts, name, st_opt_sync_call); if ((uname == NULL) && (nodeid > 0)) { stonith__set_call_options(opts, name, st_opt_cs_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 (pcmk__str_eq(dIter->value, agent, pcmk__str_none)) { 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 * \param[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 (pcmk__strcase_any_of(entry, "on", "off", NULL)) { /* 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) && pcmk__str_eq(event->target, prev_hp->target, pcmk__str_casei) && pcmk__str_eq(event->action, prev_hp->action, pcmk__str_casei) && pcmk__str_eq(event->delegate, prev_hp->delegate, pcmk__str_casei) && ((event->completed < prev_hp->completed) || ((event->completed == prev_hp->completed) && (event->completed_nsec < prev_hp->completed_nsec)))) { 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->completed == new->completed) && (hp->completed_nsec > new->completed_nsec))) { hp->next = new; new = hp; } else { np = new; do { if ((!np->next) || (hp->completed > np->next->completed) || ((hp->completed == np->next->completed) && (hp->completed_nsec > np->next->completed_nsec))) { 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; } /*! * \brief Return string equivalent of an operation state value * * \param[in] state Fencing operation state value * * \return Human-friendly string equivalent of state */ const char * stonith_op_state_str(enum op_state state) { switch (state) { case st_query: return "querying"; case st_exec: return "executing"; case st_done: return "completed"; case st_duplicate: return "duplicate"; case st_failed: return "failed"; } return "unknown"; } stonith_history_t * stonith__first_matching_event(stonith_history_t *history, bool (*matching_fn)(stonith_history_t *, void *), void *user_data) { for (stonith_history_t *hp = history; hp; hp = hp->next) { if (matching_fn(hp, user_data)) { return hp; } } return NULL; } bool stonith__event_state_pending(stonith_history_t *history, void *user_data) { return history->state != st_failed && history->state != st_done; } bool stonith__event_state_eq(stonith_history_t *history, void *user_data) { return history->state == GPOINTER_TO_INT(user_data); } bool stonith__event_state_neq(stonith_history_t *history, void *user_data) { return history->state != GPOINTER_TO_INT(user_data); } void stonith__device_parameter_flags(uint32_t *device_flags, const char *device_name, xmlNode *metadata) { xmlXPathObjectPtr xpath = NULL; int max = 0; int lpc = 0; CRM_CHECK((device_flags != NULL) && (metadata != NULL), return); xpath = xpath_search(metadata, "//parameter"); max = numXpathResults(xpath); if (max <= 0) { freeXpathObject(xpath); return; } for (lpc = 0; lpc < max; lpc++) { const char *parameter = NULL; xmlNode *match = getXpathResult(xpath, lpc); CRM_LOG_ASSERT(match != NULL); if (match == NULL) { continue; } parameter = crm_element_value(match, "name"); if (pcmk__str_eq(parameter, "plug", pcmk__str_casei)) { stonith__set_device_flags(*device_flags, device_name, st_device_supports_parameter_plug); } else if (pcmk__str_eq(parameter, "port", pcmk__str_casei)) { stonith__set_device_flags(*device_flags, device_name, st_device_supports_parameter_port); } } freeXpathObject(xpath); } // Deprecated functions kept only for backward API compatibility // LCOV_EXCL_START const char *get_stonith_provider(const char *agent, const char *provider); const char * get_stonith_provider(const char *agent, const char *provider) { return stonith_namespace2text(stonith_get_namespace(agent, provider)); } // LCOV_EXCL_STOP // End deprecated API diff --git a/lib/fencing/st_rhcs.c b/lib/fencing/st_rhcs.c index 89a2625bd3..6c8cbedc77 100644 --- a/lib/fencing/st_rhcs.c +++ b/lib/fencing/st_rhcs.c @@ -1,296 +1,318 @@ /* - * Copyright 2004-2020 the Pacemaker project contributors + * Copyright 2004-2021 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 "fencing_private.h" #define RH_STONITH_PREFIX "fence_" /*! * \internal * \brief Add available RHCS-compatible agents to a list * * \param[in,out] List to add to * * \return Number of agents added */ int stonith__list_rhcs_agents(stonith_key_value_t **devices) { // Essentially: ls -1 @sbin_dir@/fence_* int count = 0, i; struct dirent **namelist; const int file_num = scandir(PCMK__FENCE_BINDIR, &namelist, 0, alphasort); #if _POSIX_C_SOURCE < 200809L && !(defined(O_SEARCH) || defined(O_PATH)) char buffer[FILENAME_MAX + 1]; #elif defined(O_SEARCH) const int dirfd = open(PCMK__FENCE_BINDIR, O_SEARCH); #else const int dirfd = open(PCMK__FENCE_BINDIR, O_PATH); #endif for (i = 0; i < file_num; i++) { struct stat prop; if (pcmk__starts_with(namelist[i]->d_name, RH_STONITH_PREFIX)) { #if _POSIX_C_SOURCE < 200809L && !(defined(O_SEARCH) || defined(O_PATH)) snprintf(buffer, sizeof(buffer), "%s/%s", PCMK__FENCE_BINDIR, namelist[i]->d_name); if (stat(buffer, &prop) == 0 && S_ISREG(prop.st_mode)) { #else if (dirfd == -1) { if (i == 0) { crm_notice("Problem with listing %s directory" CRM_XS "errno=%d", RH_STONITH_PREFIX, errno); } free(namelist[i]); continue; } /* note: we can possibly prevent following symlinks here, which may be a good idea, but fall on the nose when these agents are moved elsewhere & linked back */ if (fstatat(dirfd, namelist[i]->d_name, &prop, 0) == 0 && S_ISREG(prop.st_mode)) { #endif *devices = stonith_key_value_add(*devices, NULL, namelist[i]->d_name); count++; } } free(namelist[i]); } if (file_num > 0) { free(namelist); } #if _POSIX_C_SOURCE >= 200809L || defined(O_SEARCH) || defined(O_PATH) if (dirfd >= 0) { close(dirfd); } #endif return count; } static void stonith_rhcs_parameter_not_required(xmlNode *metadata, const char *parameter) { char *xpath = NULL; xmlXPathObject *xpathObj = NULL; CRM_CHECK(metadata != NULL, return); CRM_CHECK(parameter != NULL, return); xpath = crm_strdup_printf("//parameter[@name='%s']", parameter); /* Fudge metadata so that the parameter isn't required in config * Pacemaker handles and adds it */ xpathObj = xpath_search(metadata, xpath); if (numXpathResults(xpathObj) > 0) { xmlNode *tmp = getXpathResult(xpathObj, 0); crm_xml_add(tmp, "required", "0"); } freeXpathObject(xpathObj); free(xpath); } /*! * \brief Execute RHCS-compatible agent's meta-data action * * \param[in] agent Agent to execute * \param[in] timeout Action timeout * \param[out] metadata Where to store output xmlNode (or NULL to ignore) * * \todo timeout is currently ignored; shouldn't we use it? */ static int stonith__rhcs_get_metadata(const char *agent, int timeout, xmlNode **metadata) { - char *buffer = NULL; xmlNode *xml = NULL; xmlNode *actions = NULL; xmlXPathObject *xpathObj = NULL; + pcmk__action_result_t *result = NULL; stonith_action_t *action = stonith_action_create(agent, "metadata", NULL, 0, 5, NULL, NULL, NULL); int rc = stonith__execute(action); if (rc < 0) { crm_warn("Could not execute metadata action for %s: %s " CRM_XS " rc=%d", agent, pcmk_strerror(rc), rc); stonith__destroy_action(action); return rc; } - stonith__action_result(action, &rc, &buffer, NULL); - stonith__destroy_action(action); - if (rc < 0) { - crm_warn("Metadata action for %s failed: %s " CRM_XS "rc=%d", - agent, pcmk_strerror(rc), rc); - free(buffer); + result = stonith__action_result(action); + + if (result->execution_status != PCMK_EXEC_DONE) { + crm_warn("Could not execute metadata action for %s: %s", + agent, pcmk_exec_status_str(result->execution_status)); + rc = pcmk_rc2legacy(stonith__result2rc(result)); + stonith__destroy_action(action); return rc; } - if (buffer == NULL) { + if (result->exit_status != CRM_EX_OK) { + crm_warn("Metadata action for %s returned error code %d", + agent, result->exit_status); + rc = pcmk_rc2legacy(stonith__result2rc(result)); + stonith__destroy_action(action); + return rc; + } + + if (result->action_stdout == NULL) { crm_warn("Metadata action for %s returned no data", agent); + stonith__destroy_action(action); return -ENODATA; } - xml = string2xml(buffer); - free(buffer); - buffer = NULL; + xml = string2xml(result->action_stdout); + stonith__destroy_action(action); + if (xml == NULL) { crm_warn("Metadata for %s is invalid", agent); return -pcmk_err_schema_validation; } xpathObj = xpath_search(xml, "//actions"); if (numXpathResults(xpathObj) > 0) { actions = getXpathResult(xpathObj, 0); } freeXpathObject(xpathObj); // Add start and stop (implemented by pacemaker, not agent) to meta-data xpathObj = xpath_search(xml, "//action[@name='stop']"); if (numXpathResults(xpathObj) <= 0) { xmlNode *tmp = NULL; tmp = create_xml_node(actions, "action"); crm_xml_add(tmp, "name", "stop"); crm_xml_add(tmp, "timeout", CRM_DEFAULT_OP_TIMEOUT_S); tmp = create_xml_node(actions, "action"); crm_xml_add(tmp, "name", "start"); crm_xml_add(tmp, "timeout", CRM_DEFAULT_OP_TIMEOUT_S); } freeXpathObject(xpathObj); // Fudge metadata so parameters are not required in config (pacemaker adds them) stonith_rhcs_parameter_not_required(xml, "action"); stonith_rhcs_parameter_not_required(xml, "plug"); stonith_rhcs_parameter_not_required(xml, "port"); if (metadata) { *metadata = xml; } else { free_xml(xml); } return pcmk_ok; } /*! * \brief Execute RHCS-compatible agent's meta-data action * * \param[in] agent Agent to execute * \param[in] timeout Action timeout * \param[out] output Where to store action output (or NULL to ignore) * * \todo timeout is currently ignored; shouldn't we use it? */ int stonith__rhcs_metadata(const char *agent, int timeout, char **output) { char *buffer = NULL; xmlNode *xml = NULL; int rc = stonith__rhcs_get_metadata(agent, timeout, &xml); if (rc != pcmk_ok) { free_xml(xml); return rc; } buffer = dump_xml_formatted_with_text(xml); free_xml(xml); if (buffer == NULL) { return -pcmk_err_schema_validation; } if (output) { *output = buffer; } else { free(buffer); } return pcmk_ok; } bool stonith__agent_is_rhcs(const char *agent) { struct stat prop; char *buffer = crm_strdup_printf(PCMK__FENCE_BINDIR "/%s", agent); int rc = stat(buffer, &prop); free(buffer); return (rc >= 0) && S_ISREG(prop.st_mode); } int stonith__rhcs_validate(stonith_t *st, int call_options, const char *target, const char *agent, GHashTable *params, const char * host_arg, int timeout, char **output, char **error_output) { int rc = pcmk_ok; int remaining_timeout = timeout; xmlNode *metadata = NULL; stonith_action_t *action = NULL; if (host_arg == NULL) { time_t start_time = time(NULL); rc = stonith__rhcs_get_metadata(agent, remaining_timeout, &metadata); if (rc == pcmk_ok) { uint32_t device_flags = 0; stonith__device_parameter_flags(&device_flags, agent, metadata); if (pcmk_is_set(device_flags, st_device_supports_parameter_port)) { host_arg = "port"; } else if (pcmk_is_set(device_flags, st_device_supports_parameter_plug)) { host_arg = "plug"; } } free_xml(metadata); remaining_timeout -= time(NULL) - start_time; if (rc == -ETIME || remaining_timeout <= 0 ) { return -ETIME; } } else if (pcmk__str_eq(host_arg, "none", pcmk__str_casei)) { host_arg = NULL; } action = stonith_action_create(agent, "validate-all", target, 0, remaining_timeout, params, NULL, host_arg); rc = stonith__execute(action); if (rc == pcmk_ok) { - stonith__action_result(action, &rc, output, error_output); + pcmk__action_result_t *result = stonith__action_result(action); + + rc = pcmk_rc2legacy(stonith__result2rc(result)); + + // Take ownership of output so stonith__destroy_action() doesn't free it + if (output != NULL) { + *output = result->action_stdout; + result->action_stdout = NULL; + } + if (error_output != NULL) { + *error_output = result->action_stderr; + result->action_stderr = NULL; + } } stonith__destroy_action(action); return rc; } diff --git a/lib/services/services.c b/lib/services/services.c index 86a0a213cc..cf8bbc70e1 100644 --- a/lib/services/services.c +++ b/lib/services/services.c @@ -1,1349 +1,1349 @@ /* * Copyright 2010-2021 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 #ifndef _GNU_SOURCE # define _GNU_SOURCE #endif #include #include #include #include #include #include #include #include #include #include #include #include #include #include "services_private.h" #include "services_ocf.h" #include "services_lsb.h" #if SUPPORT_UPSTART # include #endif #if SUPPORT_SYSTEMD # include #endif #if SUPPORT_NAGIOS # include #endif /* TODO: Develop a rollover strategy */ static int operations = 0; static GHashTable *recurring_actions = NULL; /* ops waiting to run async because of conflicting active * pending ops */ static GList *blocked_ops = NULL; /* ops currently active (in-flight) */ static GList *inflight_ops = NULL; static void handle_blocked_ops(void); /*! * \brief Find first service class that can provide a specified agent * * \param[in] agent Name of agent to search for * * \return Service class if found, NULL otherwise * * \note The priority is LSB, then systemd, then upstart. It would be preferable * to put systemd first, but LSB merely requires a file existence check, * while systemd requires contacting D-Bus. */ const char * resources_find_service_class(const char *agent) { if (services__lsb_agent_exists(agent)) { return PCMK_RESOURCE_CLASS_LSB; } #if SUPPORT_SYSTEMD if (systemd_unit_exists(agent)) { return PCMK_RESOURCE_CLASS_SYSTEMD; } #endif #if SUPPORT_UPSTART if (upstart_job_exists(agent)) { return PCMK_RESOURCE_CLASS_UPSTART; } #endif return NULL; } static inline void init_recurring_actions(void) { if (recurring_actions == NULL) { recurring_actions = pcmk__strkey_table(NULL, NULL); } } /*! * \internal * \brief Check whether op is in-flight systemd or upstart op * * \param[in] op Operation to check * * \return TRUE if op is in-flight systemd or upstart op */ static inline gboolean inflight_systemd_or_upstart(svc_action_t *op) { return pcmk__strcase_any_of(op->standard, PCMK_RESOURCE_CLASS_SYSTEMD, PCMK_RESOURCE_CLASS_UPSTART, NULL) && g_list_find(inflight_ops, op) != NULL; } /*! * \internal * \brief Expand "service" alias to an actual resource class * * \param[in] rsc Resource name (for logging only) * \param[in] standard Resource class as configured * \param[in] agent Agent name to look for * * \return Newly allocated string with actual resource class * * \note The caller is responsible for calling free() on the result. */ static char * expand_resource_class(const char *rsc, const char *standard, const char *agent) { char *expanded_class = NULL; if (strcasecmp(standard, PCMK_RESOURCE_CLASS_SERVICE) == 0) { const char *found_class = resources_find_service_class(agent); if (found_class) { crm_debug("Found %s agent %s for %s", found_class, agent, rsc); expanded_class = strdup(found_class); } else { crm_info("Assuming resource class lsb for agent %s for %s", agent, rsc); expanded_class = strdup(PCMK_RESOURCE_CLASS_LSB); } } else { expanded_class = strdup(standard); } CRM_ASSERT(expanded_class); return expanded_class; } /*! * \internal * \brief Create a simple svc_action_t instance * * \return Newly allocated instance (or NULL if not enough memory) */ static svc_action_t * new_action(void) { svc_action_t *op = calloc(1, sizeof(svc_action_t)); if (op == NULL) { return NULL; } op->opaque = calloc(1, sizeof(svc_action_private_t)); if (op->opaque == NULL) { free(op); return NULL; } // Initialize result services__set_result(op, PCMK_OCF_UNKNOWN, PCMK_EXEC_UNKNOWN, NULL); return op; } static bool required_argument_missing(uint32_t ra_caps, const char *name, const char *standard, const char *provider, const char *agent, const char *action) { if (pcmk__str_empty(name)) { crm_info("Cannot create operation without resource name (bug?)"); return true; } if (pcmk__str_empty(standard)) { crm_info("Cannot create operation for %s without resource class (bug?)", name); return true; } if (pcmk_is_set(ra_caps, pcmk_ra_cap_provider) && pcmk__str_empty(provider)) { crm_info("Cannot create operation for %s resource %s " "without provider (bug?)", standard, name); return true; } if (pcmk__str_empty(agent)) { crm_info("Cannot create operation for %s without agent name (bug?)", name); return true; } if (pcmk__str_empty(action)) { crm_info("Cannot create operation for %s without action name (bug?)", name); return true; } return false; } // \return Standard Pacemaker return code (pcmk_rc_ok or ENOMEM) static int copy_action_arguments(svc_action_t *op, uint32_t ra_caps, const char *name, const char *standard, const char *provider, const char *agent, const char *action) { op->rsc = strdup(name); if (op->rsc == NULL) { return ENOMEM; } op->agent = strdup(agent); if (op->agent == NULL) { return ENOMEM; } op->standard = expand_resource_class(name, standard, agent); if (op->standard == NULL) { return ENOMEM; } if (pcmk_is_set(ra_caps, pcmk_ra_cap_status) && pcmk__str_eq(action, "monitor", pcmk__str_casei)) { action = "status"; } op->action = strdup(action); if (op->action == NULL) { return ENOMEM; } if (pcmk_is_set(ra_caps, pcmk_ra_cap_provider)) { op->provider = strdup(provider); if (op->provider == NULL) { return ENOMEM; } } return pcmk_rc_ok; } svc_action_t * services__create_resource_action(const char *name, const char *standard, const char *provider, const char *agent, const char *action, guint interval_ms, int timeout, GHashTable *params, enum svc_action_flags flags) { svc_action_t *op = NULL; uint32_t ra_caps = pcmk_get_ra_caps(standard); int rc = pcmk_rc_ok; op = new_action(); if (op == NULL) { crm_crit("Cannot prepare action: %s", strerror(ENOMEM)); if (params != NULL) { g_hash_table_destroy(params); } return NULL; } op->interval_ms = interval_ms; op->timeout = timeout; op->flags = flags; op->sequence = ++operations; // Take ownership of params if (pcmk_is_set(ra_caps, pcmk_ra_cap_params)) { op->params = params; } else if (params != NULL) { g_hash_table_destroy(params); params = NULL; } if (required_argument_missing(ra_caps, name, standard, provider, agent, action)) { services__set_result(op, services__generic_error(op), PCMK_EXEC_ERROR_FATAL, "Required agent or action information missing"); return op; } op->id = pcmk__op_key(name, action, interval_ms); if (copy_action_arguments(op, ra_caps, name, standard, provider, agent, action) != pcmk_rc_ok) { crm_crit("Cannot prepare %s action for %s: %s", action, name, strerror(ENOMEM)); services__handle_exec_error(op, ENOMEM); return op; } if (strcasecmp(op->standard, PCMK_RESOURCE_CLASS_OCF) == 0) { rc = services__ocf_prepare(op); } else if (strcasecmp(op->standard, PCMK_RESOURCE_CLASS_LSB) == 0) { rc = services__lsb_prepare(op); #if SUPPORT_SYSTEMD } else if (strcasecmp(op->standard, PCMK_RESOURCE_CLASS_SYSTEMD) == 0) { rc = services__systemd_prepare(op); #endif #if SUPPORT_UPSTART } else if (strcasecmp(op->standard, PCMK_RESOURCE_CLASS_UPSTART) == 0) { rc = services__upstart_prepare(op); #endif #if SUPPORT_NAGIOS } else if (strcasecmp(op->standard, PCMK_RESOURCE_CLASS_NAGIOS) == 0) { rc = services__nagios_prepare(op); #endif } else { - crm_err("Unknown resource standard: %s", op->standard); + crm_info("Unknown resource standard: %s", op->standard); rc = ENOENT; } if (rc != pcmk_rc_ok) { - crm_err("Cannot prepare %s operation for %s: %s", - action, name, strerror(rc)); + crm_info("Cannot prepare %s operation for %s: %s", + action, name, strerror(rc)); services__handle_exec_error(op, rc); } return op; } svc_action_t * resources_action_create(const char *name, const char *standard, const char *provider, const char *agent, const char *action, guint interval_ms, int timeout, GHashTable *params, enum svc_action_flags flags) { svc_action_t *op = services__create_resource_action(name, standard, provider, agent, action, interval_ms, timeout, params, flags); if (op == NULL || op->rc != 0) { services_action_free(op); return NULL; } else { // Preserve public API backward compatibility op->rc = PCMK_OCF_OK; op->status = PCMK_EXEC_DONE; return op; } } svc_action_t * services_action_create_generic(const char *exec, const char *args[]) { svc_action_t *op = new_action(); CRM_ASSERT(op != NULL); op->opaque->exec = strdup(exec); op->opaque->args[0] = strdup(exec); if ((op->opaque->exec == NULL) || (op->opaque->args[0] == NULL)) { crm_crit("Cannot prepare action for '%s': %s", exec, strerror(ENOMEM)); services__set_result(op, PCMK_OCF_UNKNOWN_ERROR, PCMK_EXEC_ERROR, strerror(ENOMEM)); return op; } if (args == NULL) { return op; } for (int cur_arg = 1; args[cur_arg - 1] != NULL; cur_arg++) { if (cur_arg == PCMK__NELEM(op->opaque->args)) { crm_info("Cannot prepare action for '%s': Too many arguments", exec); services__set_result(op, PCMK_OCF_UNKNOWN_ERROR, PCMK_EXEC_ERROR_HARD, "Too many arguments"); break; } op->opaque->args[cur_arg] = strdup(args[cur_arg - 1]); if (op->opaque->args[cur_arg] == NULL) { crm_crit("Cannot prepare action for '%s': %s", exec, strerror(ENOMEM)); services__set_result(op, PCMK_OCF_UNKNOWN_ERROR, PCMK_EXEC_ERROR, strerror(ENOMEM)); break; } } return op; } /*! * \brief Create an alert agent action * * \param[in] id Alert ID * \param[in] exec Path to alert agent executable * \param[in] timeout Action timeout * \param[in] params Parameters to use with action * \param[in] sequence Action sequence number * \param[in] cb_data Data to pass to callback function * * \return New action on success, NULL on error * \note It is the caller's responsibility to free cb_data. * The caller should not free params explicitly. */ svc_action_t * services_alert_create(const char *id, const char *exec, int timeout, GHashTable *params, int sequence, void *cb_data) { svc_action_t *action = services_action_create_generic(exec, NULL); action->timeout = timeout; action->id = strdup(id); action->params = params; action->sequence = sequence; action->cb_data = cb_data; return action; } /*! * \brief Set the user and group that an action will execute as * * \param[in,out] action Action to modify * \param[in] user Name of user to execute action as * \param[in] group Name of group to execute action as * * \return pcmk_ok on success, -errno otherwise * * \note This will have no effect unless the process executing the action runs * as root, and the action is not a systemd or upstart action. * We could implement this for systemd by adding User= and Group= to * [Service] in the override file, but that seems more likely to cause * problems than be useful. */ int services_action_user(svc_action_t *op, const char *user) { CRM_CHECK((op != NULL) && (user != NULL), return -EINVAL); return crm_user_lookup(user, &(op->opaque->uid), &(op->opaque->gid)); } /*! * \brief Execute an alert agent action * * \param[in] action Action to execute * \param[in] cb Function to call when action completes * * \return TRUE if the library will free action, FALSE otherwise * * \note If this function returns FALSE, it is the caller's responsibility to * free the action with services_action_free(). */ gboolean services_alert_async(svc_action_t *action, void (*cb)(svc_action_t *op)) { action->synchronous = false; action->opaque->callback = cb; return services__execute_file(action) == pcmk_rc_ok; } #if SUPPORT_DBUS /*! * \internal * \brief Update operation's pending DBus call, unreferencing old one if needed * * \param[in,out] op Operation to modify * \param[in] pending Pending call to set */ void services_set_op_pending(svc_action_t *op, DBusPendingCall *pending) { if (op->opaque->pending && (op->opaque->pending != pending)) { if (pending) { crm_info("Lost pending %s DBus call (%p)", op->id, op->opaque->pending); } else { crm_trace("Done with pending %s DBus call (%p)", op->id, op->opaque->pending); } dbus_pending_call_unref(op->opaque->pending); } op->opaque->pending = pending; if (pending) { crm_trace("Updated pending %s DBus call (%p)", op->id, pending); } else { crm_trace("Cleared pending %s DBus call", op->id); } } #endif void services_action_cleanup(svc_action_t * op) { if ((op == NULL) || (op->opaque == NULL)) { return; } #if SUPPORT_DBUS if(op->opaque->timerid != 0) { crm_trace("Removing timer for call %s to %s", op->action, op->rsc); g_source_remove(op->opaque->timerid); op->opaque->timerid = 0; } if(op->opaque->pending) { if (dbus_pending_call_get_completed(op->opaque->pending)) { // This should never be the case crm_warn("Result of %s op %s was unhandled", op->standard, op->id); } else { crm_debug("Will ignore any result of canceled %s op %s", op->standard, op->id); } dbus_pending_call_cancel(op->opaque->pending); services_set_op_pending(op, NULL); } #endif if (op->opaque->stderr_gsource) { mainloop_del_fd(op->opaque->stderr_gsource); op->opaque->stderr_gsource = NULL; } if (op->opaque->stdout_gsource) { mainloop_del_fd(op->opaque->stdout_gsource); op->opaque->stdout_gsource = NULL; } } /*! * \internal * \brief Map an actual resource action result to a standard OCF result * * \param[in] standard Agent standard (must not be "service") * \param[in] action Action that result is for * \param[in] exit_status Actual agent exit status * * \return Standard OCF result */ enum ocf_exitcode services_result2ocf(const char *standard, const char *action, int exit_status) { if (pcmk__str_eq(standard, PCMK_RESOURCE_CLASS_OCF, pcmk__str_casei)) { return services__ocf2ocf(exit_status); #if SUPPORT_SYSTEMD } else if (pcmk__str_eq(standard, PCMK_RESOURCE_CLASS_SYSTEMD, pcmk__str_casei)) { return services__systemd2ocf(exit_status); #endif #if SUPPORT_UPSTART } else if (pcmk__str_eq(standard, PCMK_RESOURCE_CLASS_UPSTART, pcmk__str_casei)) { return services__upstart2ocf(exit_status); #endif #if SUPPORT_NAGIOS } else if (pcmk__str_eq(standard, PCMK_RESOURCE_CLASS_NAGIOS, pcmk__str_casei)) { return services__nagios2ocf(exit_status); #endif } else if (pcmk__str_eq(standard, PCMK_RESOURCE_CLASS_LSB, pcmk__str_casei)) { return services__lsb2ocf(action, exit_status); } else { crm_warn("Treating result from unknown standard '%s' as OCF", ((standard == NULL)? "unspecified" : standard)); return services__ocf2ocf(exit_status); } } void services_action_free(svc_action_t * op) { unsigned int i; if (op == NULL) { return; } /* The operation should be removed from all tracking lists by this point. * If it's not, we have a bug somewhere, so bail. That may lead to a * memory leak, but it's better than a use-after-free segmentation fault. */ CRM_CHECK(g_list_find(inflight_ops, op) == NULL, return); CRM_CHECK(g_list_find(blocked_ops, op) == NULL, return); CRM_CHECK((recurring_actions == NULL) || (g_hash_table_lookup(recurring_actions, op->id) == NULL), return); services_action_cleanup(op); if (op->opaque->repeat_timer) { g_source_remove(op->opaque->repeat_timer); op->opaque->repeat_timer = 0; } free(op->id); free(op->opaque->exec); for (i = 0; i < PCMK__NELEM(op->opaque->args); i++) { free(op->opaque->args[i]); } free(op->opaque->exit_reason); free(op->opaque); free(op->rsc); free(op->action); free(op->standard); free(op->agent); free(op->provider); free(op->stdout_data); free(op->stderr_data); if (op->params) { g_hash_table_destroy(op->params); op->params = NULL; } free(op); } gboolean cancel_recurring_action(svc_action_t * op) { crm_info("Cancelling %s operation %s", op->standard, op->id); if (recurring_actions) { g_hash_table_remove(recurring_actions, op->id); } if (op->opaque->repeat_timer) { g_source_remove(op->opaque->repeat_timer); op->opaque->repeat_timer = 0; } return TRUE; } /*! * \brief Cancel a recurring action * * \param[in] name Name of resource that operation is for * \param[in] action Name of operation to cancel * \param[in] interval_ms Interval of operation to cancel * * \return TRUE if action was successfully cancelled, FALSE otherwise */ gboolean services_action_cancel(const char *name, const char *action, guint interval_ms) { gboolean cancelled = FALSE; char *id = pcmk__op_key(name, action, interval_ms); svc_action_t *op = NULL; /* We can only cancel a recurring action */ init_recurring_actions(); op = g_hash_table_lookup(recurring_actions, id); if (op == NULL) { goto done; } // Tell services__finalize_async_op() not to reschedule the operation op->cancel = TRUE; /* Stop tracking it as a recurring operation, and stop its repeat timer */ cancel_recurring_action(op); /* If the op has a PID, it's an in-flight child process, so kill it. * * Whether the kill succeeds or fails, the main loop will send the op to * async_action_complete() (and thus services__finalize_async_op()) when the * process goes away. */ if (op->pid != 0) { crm_info("Terminating in-flight op %s[%d] early because it was cancelled", id, op->pid); cancelled = mainloop_child_kill(op->pid); if (cancelled == FALSE) { crm_err("Termination of %s[%d] failed", id, op->pid); } goto done; } #if SUPPORT_DBUS // In-flight systemd and upstart ops don't have a pid if (inflight_systemd_or_upstart(op)) { inflight_ops = g_list_remove(inflight_ops, op); /* This will cause any result that comes in later to be discarded, so we * don't call the callback and free the operation twice. */ services_action_cleanup(op); } #endif /* The rest of this is essentially equivalent to * services__finalize_async_op(), minus the handle_blocked_ops() call. */ // Report operation as cancelled services__set_cancelled(op); if (op->opaque->callback) { op->opaque->callback(op); } blocked_ops = g_list_remove(blocked_ops, op); services_action_free(op); cancelled = TRUE; // @TODO Initiate handle_blocked_ops() asynchronously done: free(id); return cancelled; } gboolean services_action_kick(const char *name, const char *action, guint interval_ms) { svc_action_t * op = NULL; char *id = pcmk__op_key(name, action, interval_ms); init_recurring_actions(); op = g_hash_table_lookup(recurring_actions, id); free(id); if (op == NULL) { return FALSE; } if (op->pid || inflight_systemd_or_upstart(op)) { return TRUE; } else { if (op->opaque->repeat_timer) { g_source_remove(op->opaque->repeat_timer); op->opaque->repeat_timer = 0; } recurring_action_timer(op); return TRUE; } } /*! * \internal * \brief Add a new recurring operation, checking for duplicates * * \param[in] op Operation to add * * \return TRUE if duplicate found (and reschedule), FALSE otherwise */ static gboolean handle_duplicate_recurring(svc_action_t * op) { svc_action_t * dup = NULL; /* check for duplicates */ dup = g_hash_table_lookup(recurring_actions, op->id); if (dup && (dup != op)) { /* update user data */ if (op->opaque->callback) { dup->opaque->callback = op->opaque->callback; dup->cb_data = op->cb_data; op->cb_data = NULL; } /* immediately execute the next interval */ if (dup->pid != 0) { if (op->opaque->repeat_timer) { g_source_remove(op->opaque->repeat_timer); op->opaque->repeat_timer = 0; } recurring_action_timer(dup); } /* free the duplicate */ services_action_free(op); return TRUE; } return FALSE; } /*! * \internal * \brief Execute an action appropriately according to its standard * * \param[in] op Action to execute * * \return Standard Pacemaker return code * \retval EBUSY Recurring operation could not be initiated * \retval pcmk_rc_error Synchronous action failed * \retval pcmk_rc_ok Synchronous action succeeded, or asynchronous action * should not be freed (because it already was or is * pending) * * \note If the return value for an asynchronous action is not pcmk_rc_ok, the * caller is responsible for freeing the action. */ static int execute_action(svc_action_t *op) { #if SUPPORT_UPSTART if (pcmk__str_eq(op->standard, PCMK_RESOURCE_CLASS_UPSTART, pcmk__str_casei)) { return services__execute_upstart(op); } #endif #if SUPPORT_SYSTEMD if (pcmk__str_eq(op->standard, PCMK_RESOURCE_CLASS_SYSTEMD, pcmk__str_casei)) { return services__execute_systemd(op); } #endif return services__execute_file(op); } void services_add_inflight_op(svc_action_t * op) { if (op == NULL) { return; } CRM_ASSERT(op->synchronous == FALSE); /* keep track of ops that are in-flight to avoid collisions in the same namespace */ if (op->rsc) { inflight_ops = g_list_append(inflight_ops, op); } } /*! * \internal * \brief Stop tracking an operation that completed * * \param[in] op Operation to stop tracking */ void services_untrack_op(svc_action_t *op) { /* Op is no longer in-flight or blocked */ inflight_ops = g_list_remove(inflight_ops, op); blocked_ops = g_list_remove(blocked_ops, op); /* Op is no longer blocking other ops, so check if any need to run */ handle_blocked_ops(); } gboolean services_action_async_fork_notify(svc_action_t * op, void (*action_callback) (svc_action_t *), void (*action_fork_callback) (svc_action_t *)) { op->synchronous = false; if (action_callback) { op->opaque->callback = action_callback; } if (action_fork_callback) { op->opaque->fork_callback = action_fork_callback; } if (op->interval_ms > 0) { init_recurring_actions(); if (handle_duplicate_recurring(op) == TRUE) { /* entry rescheduled, dup freed */ /* exit early */ return TRUE; } g_hash_table_replace(recurring_actions, op->id, op); } if (!pcmk_is_set(op->flags, SVC_ACTION_NON_BLOCKED) && op->rsc && is_op_blocked(op->rsc)) { blocked_ops = g_list_append(blocked_ops, op); return TRUE; } return execute_action(op) == pcmk_rc_ok; } gboolean services_action_async(svc_action_t * op, void (*action_callback) (svc_action_t *)) { return services_action_async_fork_notify(op, action_callback, NULL); } static gboolean processing_blocked_ops = FALSE; gboolean is_op_blocked(const char *rsc) { GList *gIter = NULL; svc_action_t *op = NULL; for (gIter = inflight_ops; gIter != NULL; gIter = gIter->next) { op = gIter->data; if (pcmk__str_eq(op->rsc, rsc, pcmk__str_casei)) { return TRUE; } } return FALSE; } static void handle_blocked_ops(void) { GList *executed_ops = NULL; GList *gIter = NULL; svc_action_t *op = NULL; if (processing_blocked_ops) { /* avoid nested calling of this function */ return; } processing_blocked_ops = TRUE; /* n^2 operation here, but blocked ops are incredibly rare. this list * will be empty 99% of the time. */ for (gIter = blocked_ops; gIter != NULL; gIter = gIter->next) { op = gIter->data; if (is_op_blocked(op->rsc)) { continue; } executed_ops = g_list_append(executed_ops, op); if (execute_action(op) != pcmk_rc_ok) { /* this can cause this function to be called recursively * which is why we have processing_blocked_ops static variable */ services__finalize_async_op(op); } } for (gIter = executed_ops; gIter != NULL; gIter = gIter->next) { op = gIter->data; blocked_ops = g_list_remove(blocked_ops, op); } g_list_free(executed_ops); processing_blocked_ops = FALSE; } /*! * \internal * \brief Execute a meta-data action appropriately to standard * * \param[in] op Meta-data action to execute * * \return Standard Pacemaker return code */ static int execute_metadata_action(svc_action_t *op) { const char *class = op->standard; if (op->agent == NULL) { - crm_err("meta-data requested without specifying agent"); + crm_info("Meta-data requested without specifying agent"); services__set_result(op, services__generic_error(op), PCMK_EXEC_ERROR_FATAL, "Agent not specified"); return EINVAL; } if (class == NULL) { - crm_err("meta-data requested for agent %s without specifying class", + crm_info("Meta-data requested for agent %s without specifying class", op->agent); services__set_result(op, services__generic_error(op), PCMK_EXEC_ERROR_FATAL, "Agent standard not specified"); return EINVAL; } if (!strcmp(class, PCMK_RESOURCE_CLASS_SERVICE)) { class = resources_find_service_class(op->agent); } if (class == NULL) { - crm_err("meta-data requested for %s, but could not determine class", - op->agent); + crm_info("Meta-data requested for %s, but could not determine class", + op->agent); services__set_result(op, services__generic_error(op), PCMK_EXEC_ERROR_HARD, "Agent standard could not be determined"); return EINVAL; } if (pcmk__str_eq(class, PCMK_RESOURCE_CLASS_LSB, pcmk__str_casei)) { return pcmk_legacy2rc(services__get_lsb_metadata(op->agent, &op->stdout_data)); } #if SUPPORT_NAGIOS if (pcmk__str_eq(class, PCMK_RESOURCE_CLASS_NAGIOS, pcmk__str_casei)) { return pcmk_legacy2rc(services__get_nagios_metadata(op->agent, &op->stdout_data)); } #endif return execute_action(op); } gboolean services_action_sync(svc_action_t * op) { gboolean rc = TRUE; if (op == NULL) { crm_trace("No operation to execute"); return FALSE; } op->synchronous = true; if (pcmk__str_eq(op->action, "meta-data", pcmk__str_casei)) { /* Synchronous meta-data operations are handled specially. Since most * resource classes don't provide any meta-data, it has to be * synthesized from available information about the agent. * * services_action_async() doesn't treat meta-data actions specially, so * it will result in an error for classes that don't support the action. */ rc = (execute_metadata_action(op) == pcmk_rc_ok); } else { rc = (execute_action(op) == pcmk_rc_ok); } crm_trace(" > " PCMK__OP_FMT ": %s = %d", op->rsc, op->action, op->interval_ms, op->opaque->exec, op->rc); if (op->stdout_data) { crm_trace(" > stdout: %s", op->stdout_data); } if (op->stderr_data) { crm_trace(" > stderr: %s", op->stderr_data); } return rc; } GList * get_directory_list(const char *root, gboolean files, gboolean executable) { return services_os_get_directory_list(root, files, executable); } GList * resources_list_standards(void) { GList *standards = NULL; standards = g_list_append(standards, strdup(PCMK_RESOURCE_CLASS_OCF)); standards = g_list_append(standards, strdup(PCMK_RESOURCE_CLASS_LSB)); standards = g_list_append(standards, strdup(PCMK_RESOURCE_CLASS_SERVICE)); #if SUPPORT_SYSTEMD { GList *agents = systemd_unit_listall(); if (agents != NULL) { standards = g_list_append(standards, strdup(PCMK_RESOURCE_CLASS_SYSTEMD)); g_list_free_full(agents, free); } } #endif #if SUPPORT_UPSTART { GList *agents = upstart_job_listall(); if (agents != NULL) { standards = g_list_append(standards, strdup(PCMK_RESOURCE_CLASS_UPSTART)); g_list_free_full(agents, free); } } #endif #if SUPPORT_NAGIOS { GList *agents = services__list_nagios_agents(); if (agents != NULL) { standards = g_list_append(standards, strdup(PCMK_RESOURCE_CLASS_NAGIOS)); g_list_free_full(agents, free); } } #endif return standards; } GList * resources_list_providers(const char *standard) { if (pcmk_is_set(pcmk_get_ra_caps(standard), pcmk_ra_cap_provider)) { return resources_os_list_ocf_providers(); } return NULL; } GList * resources_list_agents(const char *standard, const char *provider) { if ((standard == NULL) || (strcasecmp(standard, PCMK_RESOURCE_CLASS_SERVICE) == 0)) { GList *tmp1; GList *tmp2; GList *result = services__list_lsb_agents(); if (standard == NULL) { tmp1 = result; tmp2 = resources_os_list_ocf_agents(NULL); if (tmp2) { result = g_list_concat(tmp1, tmp2); } } #if SUPPORT_SYSTEMD tmp1 = result; tmp2 = systemd_unit_listall(); if (tmp2) { result = g_list_concat(tmp1, tmp2); } #endif #if SUPPORT_UPSTART tmp1 = result; tmp2 = upstart_job_listall(); if (tmp2) { result = g_list_concat(tmp1, tmp2); } #endif return result; } else if (strcasecmp(standard, PCMK_RESOURCE_CLASS_OCF) == 0) { return resources_os_list_ocf_agents(provider); } else if (strcasecmp(standard, PCMK_RESOURCE_CLASS_LSB) == 0) { return services__list_lsb_agents(); #if SUPPORT_SYSTEMD } else if (strcasecmp(standard, PCMK_RESOURCE_CLASS_SYSTEMD) == 0) { return systemd_unit_listall(); #endif #if SUPPORT_UPSTART } else if (strcasecmp(standard, PCMK_RESOURCE_CLASS_UPSTART) == 0) { return upstart_job_listall(); #endif #if SUPPORT_NAGIOS } else if (strcasecmp(standard, PCMK_RESOURCE_CLASS_NAGIOS) == 0) { return services__list_nagios_agents(); #endif } return NULL; } gboolean resources_agent_exists(const char *standard, const char *provider, const char *agent) { GList *standards = NULL; GList *providers = NULL; GList *iter = NULL; gboolean rc = FALSE; gboolean has_providers = FALSE; standards = resources_list_standards(); for (iter = standards; iter != NULL; iter = iter->next) { if (pcmk__str_eq(iter->data, standard, pcmk__str_none)) { rc = TRUE; break; } } if (rc == FALSE) { goto done; } rc = FALSE; has_providers = pcmk_is_set(pcmk_get_ra_caps(standard), pcmk_ra_cap_provider); if (has_providers == TRUE && provider != NULL) { providers = resources_list_providers(standard); for (iter = providers; iter != NULL; iter = iter->next) { if (pcmk__str_eq(iter->data, provider, pcmk__str_none)) { rc = TRUE; break; } } } else if (has_providers == FALSE && provider == NULL) { rc = TRUE; } if (rc == FALSE) { goto done; } if (pcmk__str_eq(standard, PCMK_RESOURCE_CLASS_SERVICE, pcmk__str_casei)) { if (services__lsb_agent_exists(agent)) { rc = TRUE; #if SUPPORT_SYSTEMD } else if (systemd_unit_exists(agent)) { rc = TRUE; #endif #if SUPPORT_UPSTART } else if (upstart_job_exists(agent)) { rc = TRUE; #endif } else { rc = FALSE; } } else if (pcmk__str_eq(standard, PCMK_RESOURCE_CLASS_OCF, pcmk__str_casei)) { rc = services__ocf_agent_exists(provider, agent); } else if (pcmk__str_eq(standard, PCMK_RESOURCE_CLASS_LSB, pcmk__str_casei)) { rc = services__lsb_agent_exists(agent); #if SUPPORT_SYSTEMD } else if (pcmk__str_eq(standard, PCMK_RESOURCE_CLASS_SYSTEMD, pcmk__str_casei)) { rc = systemd_unit_exists(agent); #endif #if SUPPORT_UPSTART } else if (pcmk__str_eq(standard, PCMK_RESOURCE_CLASS_UPSTART, pcmk__str_casei)) { rc = upstart_job_exists(agent); #endif #if SUPPORT_NAGIOS } else if (pcmk__str_eq(standard, PCMK_RESOURCE_CLASS_NAGIOS, pcmk__str_casei)) { rc = services__nagios_agent_exists(agent); #endif } else { rc = FALSE; } done: g_list_free(standards); g_list_free(providers); return rc; } /*! * \internal * \brief Set the result of an action * * \param[out] action Where to set action result * \param[in] agent_status Exit status to set * \param[in] exec_status Execution status to set * \param[in] reason Human-friendly description of event to set */ void services__set_result(svc_action_t *action, int agent_status, enum pcmk_exec_status exec_status, const char *reason) { if (action == NULL) { return; } action->rc = agent_status; action->status = exec_status; if (!pcmk__str_eq(action->opaque->exit_reason, reason, pcmk__str_none)) { free(action->opaque->exit_reason); action->opaque->exit_reason = (reason == NULL)? NULL : strdup(reason); } } /*! * \internal * \brief Set the result of an action to cancelled * * \param[out] action Where to set action result * * \note This sets execution status but leaves the exit status unchanged */ void services__set_cancelled(svc_action_t *action) { if (action != NULL) { action->status = PCMK_EXEC_CANCELLED; free(action->opaque->exit_reason); action->opaque->exit_reason = NULL; } } /*! * \internal * \brief Get the exit reason of an action * * \param[in] action Action to check * * \return Action's exit reason (or NULL if none) */ const char * services__exit_reason(svc_action_t *action) { return action->opaque->exit_reason; } /*! * \internal * \brief Steal stdout from an action * * \param[in] action Action whose stdout is desired * * \return Action's stdout (which may be NULL) * \note Upon return, \p action will no longer track the output, so it is the * caller's responsibility to free the return value. */ char * services__grab_stdout(svc_action_t *action) { char *output = action->stdout_data; action->stdout_data = NULL; return output; } /*! * \internal * \brief Steal stderr from an action * * \param[in] action Action whose stderr is desired * * \return Action's stderr (which may be NULL) * \note Upon return, \p action will no longer track the output, so it is the * caller's responsibility to free the return value. */ char * services__grab_stderr(svc_action_t *action) { char *output = action->stderr_data; action->stderr_data = NULL; return output; } diff --git a/lib/services/services_linux.c b/lib/services/services_linux.c index b2ff27a0d7..9a4c6cf80d 100644 --- a/lib/services/services_linux.c +++ b/lib/services/services_linux.c @@ -1,1416 +1,1420 @@ /* * Copyright 2010-2021 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 #ifndef _GNU_SOURCE # define _GNU_SOURCE #endif #include #include #include #include #include #include #include #include #include #include #include "crm/crm.h" #include "crm/common/mainloop.h" #include "crm/services.h" #include "crm/services_internal.h" #include "services_private.h" static void close_pipe(int fildes[]); /* We have two alternative ways of handling SIGCHLD when synchronously waiting * for spawned processes to complete. Both rely on polling a file descriptor to * discover SIGCHLD events. * * If sys/signalfd.h is available (e.g. on Linux), we call signalfd() to * generate the file descriptor. Otherwise, we use the "self-pipe trick" * (opening a pipe and writing a byte to it when SIGCHLD is received). */ #ifdef HAVE_SYS_SIGNALFD_H // signalfd() implementation #include // Everything needed to manage SIGCHLD handling struct sigchld_data_s { sigset_t mask; // Signals to block now (including SIGCHLD) sigset_t old_mask; // Previous set of blocked signals }; // Initialize SIGCHLD data and prepare for use static bool sigchld_setup(struct sigchld_data_s *data) { sigemptyset(&(data->mask)); sigaddset(&(data->mask), SIGCHLD); sigemptyset(&(data->old_mask)); // Block SIGCHLD (saving previous set of blocked signals to restore later) if (sigprocmask(SIG_BLOCK, &(data->mask), &(data->old_mask)) < 0) { - crm_err("Wait for child process completion failed: %s " - CRM_XS " source=sigprocmask", pcmk_strerror(errno)); + crm_info("Wait for child process completion failed: %s " + CRM_XS " source=sigprocmask", pcmk_strerror(errno)); return false; } return true; } // Get a file descriptor suitable for polling for SIGCHLD events static int sigchld_open(struct sigchld_data_s *data) { int fd; CRM_CHECK(data != NULL, return -1); fd = signalfd(-1, &(data->mask), SFD_NONBLOCK); if (fd < 0) { - crm_err("Wait for child process completion failed: %s " - CRM_XS " source=signalfd", pcmk_strerror(errno)); + crm_info("Wait for child process completion failed: %s " + CRM_XS " source=signalfd", pcmk_strerror(errno)); } return fd; } // Close a file descriptor returned by sigchld_open() static void sigchld_close(int fd) { if (fd > 0) { close(fd); } } // Return true if SIGCHLD was received from polled fd static bool sigchld_received(int fd) { struct signalfd_siginfo fdsi; ssize_t s; if (fd < 0) { return false; } s = read(fd, &fdsi, sizeof(struct signalfd_siginfo)); if (s != sizeof(struct signalfd_siginfo)) { - crm_err("Wait for child process completion failed: %s " - CRM_XS " source=read", pcmk_strerror(errno)); + crm_info("Wait for child process completion failed: %s " + CRM_XS " source=read", pcmk_strerror(errno)); } else if (fdsi.ssi_signo == SIGCHLD) { return true; } return false; } // Do anything needed after done waiting for SIGCHLD static void sigchld_cleanup(struct sigchld_data_s *data) { // Restore the original set of blocked signals if ((sigismember(&(data->old_mask), SIGCHLD) == 0) && (sigprocmask(SIG_UNBLOCK, &(data->mask), NULL) < 0)) { crm_warn("Could not clean up after child process completion: %s", pcmk_strerror(errno)); } } #else // HAVE_SYS_SIGNALFD_H not defined // Self-pipe implementation (see above for function descriptions) struct sigchld_data_s { int pipe_fd[2]; // Pipe file descriptors struct sigaction sa; // Signal handling info (with SIGCHLD) struct sigaction old_sa; // Previous signal handling info }; // We need a global to use in the signal handler volatile struct sigchld_data_s *last_sigchld_data = NULL; static void sigchld_handler() { // We received a SIGCHLD, so trigger pipe polling if ((last_sigchld_data != NULL) && (last_sigchld_data->pipe_fd[1] >= 0) && (write(last_sigchld_data->pipe_fd[1], "", 1) == -1)) { - crm_err("Wait for child process completion failed: %s " - CRM_XS " source=write", pcmk_strerror(errno)); + crm_info("Wait for child process completion failed: %s " + CRM_XS " source=write", pcmk_strerror(errno)); } } static bool sigchld_setup(struct sigchld_data_s *data) { int rc; data->pipe_fd[0] = data->pipe_fd[1] = -1; if (pipe(data->pipe_fd) == -1) { - crm_err("Wait for child process completion failed: %s " - CRM_XS " source=pipe", pcmk_strerror(errno)); + crm_info("Wait for child process completion failed: %s " + CRM_XS " source=pipe", pcmk_strerror(errno)); return false; } rc = pcmk__set_nonblocking(data->pipe_fd[0]); if (rc != pcmk_rc_ok) { - crm_warn("Could not set pipe input non-blocking: %s " CRM_XS " rc=%d", + crm_info("Could not set pipe input non-blocking: %s " CRM_XS " rc=%d", pcmk_rc_str(rc), rc); } rc = pcmk__set_nonblocking(data->pipe_fd[1]); if (rc != pcmk_rc_ok) { - crm_warn("Could not set pipe output non-blocking: %s " CRM_XS " rc=%d", + crm_info("Could not set pipe output non-blocking: %s " CRM_XS " rc=%d", pcmk_rc_str(rc), rc); } // Set SIGCHLD handler data->sa.sa_handler = sigchld_handler; data->sa.sa_flags = 0; sigemptyset(&(data->sa.sa_mask)); if (sigaction(SIGCHLD, &(data->sa), &(data->old_sa)) < 0) { - crm_err("Wait for child process completion failed: %s " - CRM_XS " source=sigaction", pcmk_strerror(errno)); + crm_info("Wait for child process completion failed: %s " + CRM_XS " source=sigaction", pcmk_strerror(errno)); } // Remember data for use in signal handler last_sigchld_data = data; return true; } static int sigchld_open(struct sigchld_data_s *data) { CRM_CHECK(data != NULL, return -1); return data->pipe_fd[0]; } static void sigchld_close(int fd) { // Pipe will be closed in sigchld_cleanup() return; } static bool sigchld_received(int fd) { char ch; if (fd < 0) { return false; } // Clear out the self-pipe while (read(fd, &ch, 1) == 1) /*omit*/; return true; } static void sigchld_cleanup(struct sigchld_data_s *data) { // Restore the previous SIGCHLD handler if (sigaction(SIGCHLD, &(data->old_sa), NULL) < 0) { crm_warn("Could not clean up after child process completion: %s", pcmk_strerror(errno)); } close_pipe(data->pipe_fd); } #endif /*! * \internal * \brief Close the two file descriptors of a pipe * * \param[in] fildes Array of file descriptors opened by pipe() */ static void close_pipe(int fildes[]) { if (fildes[0] >= 0) { close(fildes[0]); fildes[0] = -1; } if (fildes[1] >= 0) { close(fildes[1]); fildes[1] = -1; } } static gboolean svc_read_output(int fd, svc_action_t * op, bool is_stderr) { char *data = NULL; int rc = 0, len = 0; char buf[500]; static const size_t buf_read_len = sizeof(buf) - 1; if (fd < 0) { crm_trace("No fd for %s", op->id); return FALSE; } if (is_stderr && op->stderr_data) { len = strlen(op->stderr_data); data = op->stderr_data; crm_trace("Reading %s stderr into offset %d", op->id, len); } else if (is_stderr == FALSE && op->stdout_data) { len = strlen(op->stdout_data); data = op->stdout_data; crm_trace("Reading %s stdout into offset %d", op->id, len); } else { crm_trace("Reading %s %s into offset %d", op->id, is_stderr?"stderr":"stdout", len); } do { rc = read(fd, buf, buf_read_len); if (rc > 0) { buf[rc] = 0; crm_trace("Got %d chars: %.80s", rc, buf); data = pcmk__realloc(data, len + rc + 1); len += sprintf(data + len, "%s", buf); } else if (errno != EINTR) { /* error or EOF * Cleanup happens in pipe_done() */ rc = FALSE; break; } } while (rc == buf_read_len || rc < 0); if (is_stderr) { op->stderr_data = data; } else { op->stdout_data = data; } return rc; } static int dispatch_stdout(gpointer userdata) { svc_action_t *op = (svc_action_t *) userdata; return svc_read_output(op->opaque->stdout_fd, op, FALSE); } static int dispatch_stderr(gpointer userdata) { svc_action_t *op = (svc_action_t *) userdata; return svc_read_output(op->opaque->stderr_fd, op, TRUE); } static void pipe_out_done(gpointer user_data) { svc_action_t *op = (svc_action_t *) user_data; crm_trace("%p", op); op->opaque->stdout_gsource = NULL; if (op->opaque->stdout_fd > STDOUT_FILENO) { close(op->opaque->stdout_fd); } op->opaque->stdout_fd = -1; } static void pipe_err_done(gpointer user_data) { svc_action_t *op = (svc_action_t *) user_data; op->opaque->stderr_gsource = NULL; if (op->opaque->stderr_fd > STDERR_FILENO) { close(op->opaque->stderr_fd); } op->opaque->stderr_fd = -1; } static struct mainloop_fd_callbacks stdout_callbacks = { .dispatch = dispatch_stdout, .destroy = pipe_out_done, }; static struct mainloop_fd_callbacks stderr_callbacks = { .dispatch = dispatch_stderr, .destroy = pipe_err_done, }; static void set_ocf_env(const char *key, const char *value, gpointer user_data) { if (setenv(key, value, 1) != 0) { crm_perror(LOG_ERR, "setenv failed for key:%s and value:%s", key, value); } } static void set_ocf_env_with_prefix(gpointer key, gpointer value, gpointer user_data) { char buffer[500]; snprintf(buffer, sizeof(buffer), strcmp(key, "OCF_CHECK_LEVEL") != 0 ? "OCF_RESKEY_%s" : "%s", (char *)key); set_ocf_env(buffer, value, user_data); } static void set_alert_env(gpointer key, gpointer value, gpointer user_data) { int rc; if (value != NULL) { rc = setenv(key, value, 1); } else { rc = unsetenv(key); } if (rc < 0) { crm_perror(LOG_ERR, "setenv %s=%s", (char*)key, (value? (char*)value : "")); } else { crm_trace("setenv %s=%s", (char*)key, (value? (char*)value : "")); } } /*! * \internal * \brief Add environment variables suitable for an action * * \param[in] op Action to use */ static void add_action_env_vars(const svc_action_t *op) { void (*env_setter)(gpointer, gpointer, gpointer) = NULL; if (op->agent == NULL) { env_setter = set_alert_env; /* we deal with alert handler */ } else if (pcmk__str_eq(op->standard, PCMK_RESOURCE_CLASS_OCF, pcmk__str_casei)) { env_setter = set_ocf_env_with_prefix; } if (env_setter != NULL && op->params != NULL) { g_hash_table_foreach(op->params, env_setter, NULL); } if (env_setter == NULL || env_setter == set_alert_env) { return; } set_ocf_env("OCF_RA_VERSION_MAJOR", PCMK_OCF_MAJOR_VERSION, NULL); set_ocf_env("OCF_RA_VERSION_MINOR", PCMK_OCF_MINOR_VERSION, NULL); set_ocf_env("OCF_ROOT", OCF_ROOT_DIR, NULL); set_ocf_env("OCF_EXIT_REASON_PREFIX", PCMK_OCF_REASON_PREFIX, NULL); if (op->rsc) { set_ocf_env("OCF_RESOURCE_INSTANCE", op->rsc, NULL); } if (op->agent != NULL) { set_ocf_env("OCF_RESOURCE_TYPE", op->agent, NULL); } /* Notes: this is not added to specification yet. Sept 10,2004 */ if (op->provider != NULL) { set_ocf_env("OCF_RESOURCE_PROVIDER", op->provider, NULL); } } static void pipe_in_single_parameter(gpointer key, gpointer value, gpointer user_data) { svc_action_t *op = user_data; char *buffer = crm_strdup_printf("%s=%s\n", (char *)key, (char *) value); int ret, total = 0, len = strlen(buffer); do { errno = 0; ret = write(op->opaque->stdin_fd, buffer + total, len - total); if (ret > 0) { total += ret; } } while ((errno == EINTR) && (total < len)); free(buffer); } /*! * \internal * \brief Pipe parameters in via stdin for action * * \param[in] op Action to use */ static void pipe_in_action_stdin_parameters(const svc_action_t *op) { crm_debug("sending args"); if (op->params) { g_hash_table_foreach(op->params, pipe_in_single_parameter, (gpointer) op); } } gboolean recurring_action_timer(gpointer data) { svc_action_t *op = data; crm_debug("Scheduling another invocation of %s", op->id); /* Clean out the old result */ free(op->stdout_data); op->stdout_data = NULL; free(op->stderr_data); op->stderr_data = NULL; op->opaque->repeat_timer = 0; services_action_async(op, NULL); return FALSE; } /*! * \internal * \brief Finalize handling of an asynchronous operation * * Given a completed asynchronous operation, cancel or reschedule it as * appropriate if recurring, call its callback if registered, stop tracking it, * and clean it up. * * \param[in,out] op Operation to finalize * * \return Standard Pacemaker return code * \retval EINVAL Caller supplied NULL or invalid \p op * \retval EBUSY Uncanceled recurring action has only been cleaned up * \retval pcmk_rc_ok Action has been freed * * \note If the return value is not pcmk_rc_ok, the caller is responsible for * freeing the action. */ int services__finalize_async_op(svc_action_t *op) { CRM_CHECK((op != NULL) && !(op->synchronous), return EINVAL); if (op->interval_ms != 0) { // Recurring operations must be either cancelled or rescheduled if (op->cancel) { services__set_cancelled(op); cancel_recurring_action(op); } else { op->opaque->repeat_timer = g_timeout_add(op->interval_ms, recurring_action_timer, (void *) op); } } if (op->opaque->callback != NULL) { op->opaque->callback(op); } // Stop tracking the operation (as in-flight or blocked) op->pid = 0; services_untrack_op(op); if ((op->interval_ms != 0) && !(op->cancel)) { // Do not free recurring actions (they will get freed when cancelled) services_action_cleanup(op); return EBUSY; } services_action_free(op); return pcmk_rc_ok; } static void close_op_input(svc_action_t *op) { if (op->opaque->stdin_fd >= 0) { close(op->opaque->stdin_fd); } } static void finish_op_output(svc_action_t *op, bool is_stderr) { mainloop_io_t **source; int fd; if (is_stderr) { source = &(op->opaque->stderr_gsource); fd = op->opaque->stderr_fd; } else { source = &(op->opaque->stdout_gsource); fd = op->opaque->stdout_fd; } if (op->synchronous || *source) { crm_trace("Finish reading %s[%d] %s", op->id, op->pid, (is_stderr? "stdout" : "stderr")); svc_read_output(fd, op, is_stderr); if (op->synchronous) { close(fd); } else { mainloop_del_fd(*source); *source = NULL; } } } // Log an operation's stdout and stderr static void log_op_output(svc_action_t *op) { char *prefix = crm_strdup_printf("%s[%d] error output", op->id, op->pid); - crm_log_output(LOG_NOTICE, prefix, op->stderr_data); + /* The library caller has better context to know how important the output + * is, so log it at info and debug severity here. They can log it again at + * higher severity if appropriate. + */ + crm_log_output(LOG_INFO, prefix, op->stderr_data); strcpy(prefix + strlen(prefix) - strlen("error output"), "output"); crm_log_output(LOG_DEBUG, prefix, op->stdout_data); free(prefix); } // Truncate exit reasons at this many characters #define EXIT_REASON_MAX_LEN 128 static void parse_exit_reason_from_stderr(svc_action_t *op) { const char *reason_start = NULL; const char *reason_end = NULL; const int prefix_len = strlen(PCMK_OCF_REASON_PREFIX); if ((op->stderr_data == NULL) || // Only OCF agents have exit reasons in stderr !pcmk__str_eq(op->standard, PCMK_RESOURCE_CLASS_OCF, pcmk__str_none)) { return; } // Find the last occurrence of the magic string indicating an exit reason for (const char *cur = strstr(op->stderr_data, PCMK_OCF_REASON_PREFIX); cur != NULL; cur = strstr(cur, PCMK_OCF_REASON_PREFIX)) { cur += prefix_len; // Skip over magic string reason_start = cur; } if ((reason_start == NULL) || (reason_start[0] == '\n') || (reason_start[0] == '\0')) { return; // No or empty exit reason } // Exit reason goes to end of line (or end of output) reason_end = strchr(reason_start, '\n'); if (reason_end == NULL) { reason_end = reason_start + strlen(reason_start); } // Limit size of exit reason to something reasonable if (reason_end > (reason_start + EXIT_REASON_MAX_LEN)) { reason_end = reason_start + EXIT_REASON_MAX_LEN; } free(op->opaque->exit_reason); op->opaque->exit_reason = strndup(reason_start, reason_end - reason_start); } /*! * \internal * \brief Process the completion of an asynchronous child process * * \param[in] p Child process that completed * \param[in] pid Process ID of child * \param[in] core (unused) * \param[in] signo Signal that interrupted child, if any * \param[in] exitcode Exit status of child process */ static void async_action_complete(mainloop_child_t *p, pid_t pid, int core, int signo, int exitcode) { svc_action_t *op = mainloop_child_userdata(p); mainloop_clear_child_userdata(p); CRM_CHECK(op->pid == pid, services__set_result(op, services__generic_error(op), PCMK_EXEC_ERROR, "Bug in mainloop handling"); return); /* Depending on the priority the mainloop gives the stdout and stderr * file descriptors, this function could be called before everything has * been read from them, so force a final read now. */ finish_op_output(op, true); finish_op_output(op, false); close_op_input(op); if (signo == 0) { crm_debug("%s[%d] exited with status %d", op->id, op->pid, exitcode); services__set_result(op, exitcode, PCMK_EXEC_DONE, NULL); log_op_output(op); parse_exit_reason_from_stderr(op); } else if (mainloop_child_timeout(p)) { - crm_warn("%s[%d] timed out after %dms", op->id, op->pid, op->timeout); + crm_info("%s[%d] timed out after %dms", op->id, op->pid, op->timeout); services__set_result(op, services__generic_error(op), PCMK_EXEC_TIMEOUT, "Process did not exit within specified timeout"); } else if (op->cancel) { /* If an in-flight recurring operation was killed because it was * cancelled, don't treat that as a failure. */ crm_info("%s[%d] terminated with signal %d (%s)", op->id, op->pid, signo, strsignal(signo)); services__set_result(op, PCMK_OCF_OK, PCMK_EXEC_CANCELLED, NULL); } else { - crm_warn("%s[%d] terminated with signal %d (%s)", + crm_info("%s[%d] terminated with signal %d (%s)", op->id, op->pid, signo, strsignal(signo)); services__set_result(op, PCMK_OCF_UNKNOWN_ERROR, PCMK_EXEC_ERROR, "Process interrupted by signal"); } services__finalize_async_op(op); } /*! * \internal * \brief Return agent standard's exit status for "generic error" * * When returning an internal error for an action, a value that is appropriate * to the action's agent standard must be used. This function returns a value * appropriate for errors in general. * * \param[in] op Action that error is for * * \return Exit status appropriate to agent standard * \note Actions without a standard will get PCMK_OCF_UNKNOWN_ERROR. */ int services__generic_error(svc_action_t *op) { if ((op == NULL) || (op->standard == NULL)) { return PCMK_OCF_UNKNOWN_ERROR; } if (pcmk__str_eq(op->standard, PCMK_RESOURCE_CLASS_LSB, pcmk__str_casei) && pcmk__str_eq(op->action, "status", pcmk__str_casei)) { return PCMK_LSB_STATUS_UNKNOWN; } #if SUPPORT_NAGIOS if (pcmk__str_eq(op->standard, PCMK_RESOURCE_CLASS_NAGIOS, pcmk__str_casei)) { return NAGIOS_STATE_UNKNOWN; } #endif return PCMK_OCF_UNKNOWN_ERROR; } /*! * \internal * \brief Return agent standard's exit status for "not installed" * * When returning an internal error for an action, a value that is appropriate * to the action's agent standard must be used. This function returns a value * appropriate for "not installed" errors. * * \param[in] op Action that error is for * * \return Exit status appropriate to agent standard * \note Actions without a standard will get PCMK_OCF_UNKNOWN_ERROR. */ int services__not_installed_error(svc_action_t *op) { if ((op == NULL) || (op->standard == NULL)) { return PCMK_OCF_UNKNOWN_ERROR; } if (pcmk__str_eq(op->standard, PCMK_RESOURCE_CLASS_LSB, pcmk__str_casei) && pcmk__str_eq(op->action, "status", pcmk__str_casei)) { return PCMK_LSB_STATUS_NOT_INSTALLED; } #if SUPPORT_NAGIOS if (pcmk__str_eq(op->standard, PCMK_RESOURCE_CLASS_NAGIOS, pcmk__str_casei)) { return NAGIOS_STATE_UNKNOWN; } #endif return PCMK_OCF_NOT_INSTALLED; } /*! * \internal * \brief Return agent standard's exit status for "insufficient privileges" * * When returning an internal error for an action, a value that is appropriate * to the action's agent standard must be used. This function returns a value * appropriate for "insufficient privileges" errors. * * \param[in] op Action that error is for * * \return Exit status appropriate to agent standard * \note Actions without a standard will get PCMK_OCF_UNKNOWN_ERROR. */ int services__authorization_error(svc_action_t *op) { if ((op == NULL) || (op->standard == NULL)) { return PCMK_OCF_UNKNOWN_ERROR; } if (pcmk__str_eq(op->standard, PCMK_RESOURCE_CLASS_LSB, pcmk__str_casei) && pcmk__str_eq(op->action, "status", pcmk__str_casei)) { return PCMK_LSB_STATUS_INSUFFICIENT_PRIV; } #if SUPPORT_NAGIOS if (pcmk__str_eq(op->standard, PCMK_RESOURCE_CLASS_NAGIOS, pcmk__str_casei)) { return NAGIOS_INSUFFICIENT_PRIV; } #endif return PCMK_OCF_INSUFFICIENT_PRIV; } /*! * \internal * \brief Return agent standard's exit status for "not configured" * * When returning an internal error for an action, a value that is appropriate * to the action's agent standard must be used. This function returns a value * appropriate for "not configured" errors. * * \param[in] op Action that error is for * \param[in] is_fatal Whether problem is cluster-wide instead of only local * * \return Exit status appropriate to agent standard * \note Actions without a standard will get PCMK_OCF_UNKNOWN_ERROR. */ int services__configuration_error(svc_action_t *op, bool is_fatal) { if ((op == NULL) || (op->standard == NULL)) { return PCMK_OCF_UNKNOWN_ERROR; } if (pcmk__str_eq(op->standard, PCMK_RESOURCE_CLASS_LSB, pcmk__str_casei) && pcmk__str_eq(op->action, "status", pcmk__str_casei)) { return PCMK_LSB_NOT_CONFIGURED; } #if SUPPORT_NAGIOS if (pcmk__str_eq(op->standard, PCMK_RESOURCE_CLASS_NAGIOS, pcmk__str_casei)) { return NAGIOS_STATE_UNKNOWN; } #endif return is_fatal? PCMK_OCF_NOT_CONFIGURED : PCMK_OCF_INVALID_PARAM; } /*! * \internal * \brief Set operation rc and status per errno from stat(), fork() or execvp() * * \param[in,out] op Operation to set rc and status for * \param[in] error Value of errno after system call * * \return void */ void services__handle_exec_error(svc_action_t * op, int error) { switch (error) { /* see execve(2), stat(2) and fork(2) */ case ENOENT: /* No such file or directory */ case EISDIR: /* Is a directory */ case ENOTDIR: /* Path component is not a directory */ case EINVAL: /* Invalid executable format */ case ENOEXEC: /* Invalid executable format */ services__set_result(op, services__not_installed_error(op), PCMK_EXEC_NOT_INSTALLED, pcmk_rc_str(error)); break; case EACCES: /* permission denied (various errors) */ case EPERM: /* permission denied (various errors) */ services__set_result(op, services__authorization_error(op), PCMK_EXEC_ERROR, pcmk_rc_str(error)); break; default: services__set_result(op, services__generic_error(op), PCMK_EXEC_ERROR, pcmk_rc_str(error)); } } /*! * \internal * \brief Exit a child process that failed before executing agent * * \param[in] op Action that failed * \param[in] exit_status Exit status code to use * \param[in] exit_reason Exit reason to output if for OCF agent */ static void exit_child(svc_action_t *op, int exit_status, const char *exit_reason) { if ((op != NULL) && (exit_reason != NULL) && pcmk__str_eq(op->standard, PCMK_RESOURCE_CLASS_OCF, pcmk__str_none)) { fprintf(stderr, PCMK_OCF_REASON_PREFIX "%s\n", exit_reason); } _exit(exit_status); } static void action_launch_child(svc_action_t *op) { int rc; /* SIGPIPE is ignored (which is different from signal blocking) by the gnutls library. * Depending on the libqb version in use, libqb may set SIGPIPE to be ignored as well. * We do not want this to be inherited by the child process. By resetting this the signal * to the default behavior, we avoid some potential odd problems that occur during OCF * scripts when SIGPIPE is ignored by the environment. */ signal(SIGPIPE, SIG_DFL); #if defined(HAVE_SCHED_SETSCHEDULER) if (sched_getscheduler(0) != SCHED_OTHER) { struct sched_param sp; memset(&sp, 0, sizeof(sp)); sp.sched_priority = 0; if (sched_setscheduler(0, SCHED_OTHER, &sp) == -1) { - crm_warn("Could not reset scheduling policy for %s", op->id); + crm_info("Could not reset scheduling policy for %s", op->id); } } #endif if (setpriority(PRIO_PROCESS, 0, 0) == -1) { - crm_warn("Could not reset process priority for %s", op->id); + crm_info("Could not reset process priority for %s", op->id); } /* Man: The call setpgrp() is equivalent to setpgid(0,0) * _and_ compiles on BSD variants too * need to investigate if it works the same too. */ setpgid(0, 0); pcmk__close_fds_in_child(false); /* It would be nice if errors in this function could be reported as * execution status (for example, PCMK_EXEC_NO_SECRETS for the secrets error * below) instead of exit status. However, we've already forked, so * exit status is all we have. At least for OCF actions, we can output an * exit reason for the parent to parse. */ #if SUPPORT_CIBSECRETS rc = pcmk__substitute_secrets(op->rsc, op->params); if (rc != pcmk_rc_ok) { if (pcmk__str_eq(op->action, "stop", pcmk__str_casei)) { crm_info("Proceeding with stop operation for %s " "despite being unable to load CIB secrets (%s)", op->rsc, pcmk_rc_str(rc)); } else { crm_err("Considering %s unconfigured " "because unable to load CIB secrets: %s", - op->rsc, pcmk_rc_str(rc)); + op->rsc, pcmk_rc_str(rc)); exit_child(op, services__configuration_error(op, false), "Unable to load CIB secrets"); } } #endif add_action_env_vars(op); /* Become the desired user */ if (op->opaque->uid && (geteuid() == 0)) { // If requested, set effective group if (op->opaque->gid && (setgid(op->opaque->gid) < 0)) { crm_err("Considering %s unauthorized because could not set " "child group to %d: %s", op->id, op->opaque->gid, strerror(errno)); exit_child(op, services__authorization_error(op), "Could not set group for child process"); } // Erase supplementary group list // (We could do initgroups() if we kept a copy of the username) if (setgroups(0, NULL) < 0) { crm_err("Considering %s unauthorized because could not " "clear supplementary groups: %s", op->id, strerror(errno)); exit_child(op, services__authorization_error(op), "Could not clear supplementary groups for child process"); } // Set effective user if (setuid(op->opaque->uid) < 0) { crm_err("Considering %s unauthorized because could not set user " "to %d: %s", op->id, op->opaque->uid, strerror(errno)); exit_child(op, services__authorization_error(op), "Could not set user for child process"); } } // Execute the agent (doesn't return if successful) execvp(op->opaque->exec, op->opaque->args); // An earlier stat() should have avoided most possible errors rc = errno; services__handle_exec_error(op, rc); crm_err("Unable to execute %s: %s", op->id, strerror(rc)); exit_child(op, op->rc, "Child process was unable to execute file"); } /*! * \internal * \brief Wait for synchronous action to complete, and set its result * * \param[in] op Action to wait for * \param[in] data Child signal data */ static void wait_for_sync_result(svc_action_t *op, struct sigchld_data_s *data) { int status = 0; int timeout = op->timeout; time_t start = time(NULL); struct pollfd fds[3]; int wait_rc = 0; const char *wait_reason = NULL; fds[0].fd = op->opaque->stdout_fd; fds[0].events = POLLIN; fds[0].revents = 0; fds[1].fd = op->opaque->stderr_fd; fds[1].events = POLLIN; fds[1].revents = 0; fds[2].fd = sigchld_open(data); fds[2].events = POLLIN; fds[2].revents = 0; crm_trace("Waiting for %s[%d]", op->id, op->pid); do { int poll_rc = poll(fds, 3, timeout); wait_reason = NULL; if (poll_rc > 0) { if (fds[0].revents & POLLIN) { svc_read_output(op->opaque->stdout_fd, op, FALSE); } if (fds[1].revents & POLLIN) { svc_read_output(op->opaque->stderr_fd, op, TRUE); } if ((fds[2].revents & POLLIN) && sigchld_received(fds[2].fd)) { wait_rc = waitpid(op->pid, &status, WNOHANG); if ((wait_rc > 0) || ((wait_rc < 0) && (errno == ECHILD))) { // Child process exited or doesn't exist break; } else if (wait_rc < 0) { wait_reason = pcmk_rc_str(errno); - crm_warn("Wait for completion of %s[%d] failed: %s " + crm_info("Wait for completion of %s[%d] failed: %s " CRM_XS " source=waitpid", op->id, op->pid, wait_reason); wait_rc = 0; // Act as if process is still running } } } else if (poll_rc == 0) { // Poll timed out with no descriptors ready timeout = 0; break; } else if ((poll_rc < 0) && (errno != EINTR)) { wait_reason = pcmk_rc_str(errno); - crm_err("Wait for completion of %s[%d] failed: %s " - CRM_XS " source=poll", op->id, op->pid, wait_reason); + crm_info("Wait for completion of %s[%d] failed: %s " + CRM_XS " source=poll", op->id, op->pid, wait_reason); break; } timeout = op->timeout - (time(NULL) - start) * 1000; } while ((op->timeout < 0 || timeout > 0)); crm_trace("Stopped waiting for %s[%d]", op->id, op->pid); finish_op_output(op, true); finish_op_output(op, false); close_op_input(op); sigchld_close(fds[2].fd); if (wait_rc <= 0) { if ((op->timeout > 0) && (timeout <= 0)) { services__set_result(op, services__generic_error(op), PCMK_EXEC_TIMEOUT, "Process did not exit within specified timeout"); - crm_warn("%s[%d] timed out after %dms", + crm_info("%s[%d] timed out after %dms", op->id, op->pid, op->timeout); } else { services__set_result(op, services__generic_error(op), PCMK_EXEC_ERROR, wait_reason); } /* If only child hasn't been successfully waited for, yet. This is to limit killing wrong target a bit more. */ if ((wait_rc == 0) && (waitpid(op->pid, &status, WNOHANG) == 0)) { if (kill(op->pid, SIGKILL)) { crm_warn("Could not kill rogue child %s[%d]: %s", op->id, op->pid, pcmk_strerror(errno)); } /* Safe to skip WNOHANG here as we sent non-ignorable signal. */ while ((waitpid(op->pid, &status, 0) == (pid_t) -1) && (errno == EINTR)) { /* keep waiting */; } } } else if (WIFEXITED(status)) { services__set_result(op, WEXITSTATUS(status), PCMK_EXEC_DONE, NULL); parse_exit_reason_from_stderr(op); crm_info("%s[%d] exited with status %d", op->id, op->pid, op->rc); } else if (WIFSIGNALED(status)) { int signo = WTERMSIG(status); services__set_result(op, services__generic_error(op), PCMK_EXEC_ERROR, "Process interrupted by signal"); - crm_err("%s[%d] terminated with signal %d (%s)", - op->id, op->pid, signo, strsignal(signo)); + crm_info("%s[%d] terminated with signal %d (%s)", + op->id, op->pid, signo, strsignal(signo)); #ifdef WCOREDUMP if (WCOREDUMP(status)) { crm_warn("%s[%d] dumped core", op->id, op->pid); } #endif } else { // Shouldn't be possible to get here services__set_result(op, services__generic_error(op), PCMK_EXEC_ERROR, "Unable to wait for child to complete"); } } /*! * \internal * \brief Execute an action whose standard uses executable files * * \param[in] op Action to execute * * \return Standard Pacemaker return value * \retval EBUSY Recurring operation could not be initiated * \retval pcmk_rc_error Synchronous action failed * \retval pcmk_rc_ok Synchronous action succeeded, or asynchronous action * should not be freed (because it already was or is * pending) * * \note If the return value for an asynchronous action is not pcmk_rc_ok, the * caller is responsible for freeing the action. */ int services__execute_file(svc_action_t *op) { int stdout_fd[2]; int stderr_fd[2]; int stdin_fd[2] = {-1, -1}; int rc; struct stat st; struct sigchld_data_s data; // Catch common failure conditions early if (stat(op->opaque->exec, &st) != 0) { rc = errno; - crm_warn("Cannot execute '%s': %s " CRM_XS " stat rc=%d", + crm_info("Cannot execute '%s': %s " CRM_XS " stat rc=%d", op->opaque->exec, pcmk_strerror(rc), rc); services__handle_exec_error(op, rc); goto done; } if (pipe(stdout_fd) < 0) { rc = errno; - crm_err("Cannot execute '%s': %s " CRM_XS " pipe(stdout) rc=%d", - op->opaque->exec, pcmk_strerror(rc), rc); + crm_info("Cannot execute '%s': %s " CRM_XS " pipe(stdout) rc=%d", + op->opaque->exec, pcmk_strerror(rc), rc); services__handle_exec_error(op, rc); goto done; } if (pipe(stderr_fd) < 0) { rc = errno; close_pipe(stdout_fd); - crm_err("Cannot execute '%s': %s " CRM_XS " pipe(stderr) rc=%d", - op->opaque->exec, pcmk_strerror(rc), rc); + crm_info("Cannot execute '%s': %s " CRM_XS " pipe(stderr) rc=%d", + op->opaque->exec, pcmk_strerror(rc), rc); services__handle_exec_error(op, rc); goto done; } if (pcmk_is_set(pcmk_get_ra_caps(op->standard), pcmk_ra_cap_stdin)) { if (pipe(stdin_fd) < 0) { rc = errno; close_pipe(stdout_fd); close_pipe(stderr_fd); - crm_err("Cannot execute '%s': %s " CRM_XS " pipe(stdin) rc=%d", - op->opaque->exec, pcmk_strerror(rc), rc); + crm_info("Cannot execute '%s': %s " CRM_XS " pipe(stdin) rc=%d", + op->opaque->exec, pcmk_strerror(rc), rc); services__handle_exec_error(op, rc); goto done; } } if (op->synchronous && !sigchld_setup(&data)) { close_pipe(stdin_fd); close_pipe(stdout_fd); close_pipe(stderr_fd); sigchld_cleanup(&data); services__set_result(op, services__generic_error(op), PCMK_EXEC_ERROR, "Could not manage signals for child process"); goto done; } op->pid = fork(); switch (op->pid) { case -1: rc = errno; close_pipe(stdin_fd); close_pipe(stdout_fd); close_pipe(stderr_fd); - crm_err("Cannot execute '%s': %s " CRM_XS " fork rc=%d", - op->opaque->exec, pcmk_strerror(rc), rc); + crm_info("Cannot execute '%s': %s " CRM_XS " fork rc=%d", + op->opaque->exec, pcmk_strerror(rc), rc); services__handle_exec_error(op, rc); if (op->synchronous) { sigchld_cleanup(&data); } goto done; break; case 0: /* Child */ close(stdout_fd[0]); close(stderr_fd[0]); if (stdin_fd[1] >= 0) { close(stdin_fd[1]); } if (STDOUT_FILENO != stdout_fd[1]) { if (dup2(stdout_fd[1], STDOUT_FILENO) != STDOUT_FILENO) { crm_warn("Can't redirect output from '%s': %s " CRM_XS " errno=%d", op->opaque->exec, pcmk_strerror(errno), errno); } close(stdout_fd[1]); } if (STDERR_FILENO != stderr_fd[1]) { if (dup2(stderr_fd[1], STDERR_FILENO) != STDERR_FILENO) { crm_warn("Can't redirect error output from '%s': %s " CRM_XS " errno=%d", op->opaque->exec, pcmk_strerror(errno), errno); } close(stderr_fd[1]); } if ((stdin_fd[0] >= 0) && (STDIN_FILENO != stdin_fd[0])) { if (dup2(stdin_fd[0], STDIN_FILENO) != STDIN_FILENO) { crm_warn("Can't redirect input to '%s': %s " CRM_XS " errno=%d", op->opaque->exec, pcmk_strerror(errno), errno); } close(stdin_fd[0]); } if (op->synchronous) { sigchld_cleanup(&data); } action_launch_child(op); CRM_ASSERT(0); /* action_launch_child is effectively noreturn */ } /* Only the parent reaches here */ close(stdout_fd[1]); close(stderr_fd[1]); if (stdin_fd[0] >= 0) { close(stdin_fd[0]); } op->opaque->stdout_fd = stdout_fd[0]; rc = pcmk__set_nonblocking(op->opaque->stdout_fd); if (rc != pcmk_rc_ok) { - crm_warn("Could not set '%s' output non-blocking: %s " + crm_info("Could not set '%s' output non-blocking: %s " CRM_XS " rc=%d", op->opaque->exec, pcmk_rc_str(rc), rc); } op->opaque->stderr_fd = stderr_fd[0]; rc = pcmk__set_nonblocking(op->opaque->stderr_fd); if (rc != pcmk_rc_ok) { - crm_warn("Could not set '%s' error output non-blocking: %s " + crm_info("Could not set '%s' error output non-blocking: %s " CRM_XS " rc=%d", op->opaque->exec, pcmk_rc_str(rc), rc); } op->opaque->stdin_fd = stdin_fd[1]; if (op->opaque->stdin_fd >= 0) { // using buffer behind non-blocking-fd here - that could be improved // as long as no other standard uses stdin_fd assume stonith rc = pcmk__set_nonblocking(op->opaque->stdin_fd); if (rc != pcmk_rc_ok) { - crm_warn("Could not set '%s' input non-blocking: %s " + crm_info("Could not set '%s' input non-blocking: %s " CRM_XS " fd=%d,rc=%d", op->opaque->exec, pcmk_rc_str(rc), op->opaque->stdin_fd, rc); } pipe_in_action_stdin_parameters(op); // as long as we are handling parameters directly in here just close close(op->opaque->stdin_fd); op->opaque->stdin_fd = -1; } // after fds are setup properly and before we plug anything into mainloop if (op->opaque->fork_callback) { op->opaque->fork_callback(op); } if (op->synchronous) { wait_for_sync_result(op, &data); sigchld_cleanup(&data); goto done; } crm_trace("Waiting async for '%s'[%d]", op->opaque->exec, op->pid); mainloop_child_add_with_flags(op->pid, op->timeout, op->id, op, pcmk_is_set(op->flags, SVC_ACTION_LEAVE_GROUP)? mainloop_leave_pid_group : 0, async_action_complete); op->opaque->stdout_gsource = mainloop_add_fd(op->id, G_PRIORITY_LOW, op->opaque->stdout_fd, op, &stdout_callbacks); op->opaque->stderr_gsource = mainloop_add_fd(op->id, G_PRIORITY_LOW, op->opaque->stderr_fd, op, &stderr_callbacks); services_add_inflight_op(op); return pcmk_rc_ok; done: if (op->synchronous) { return (op->rc == PCMK_OCF_OK)? pcmk_rc_ok : pcmk_rc_error; } else { return services__finalize_async_op(op); } } GList * services_os_get_single_directory_list(const char *root, gboolean files, gboolean executable) { GList *list = NULL; struct dirent **namelist; int entries = 0, lpc = 0; char buffer[PATH_MAX]; entries = scandir(root, &namelist, NULL, alphasort); if (entries <= 0) { return list; } for (lpc = 0; lpc < entries; lpc++) { struct stat sb; if ('.' == namelist[lpc]->d_name[0]) { free(namelist[lpc]); continue; } snprintf(buffer, sizeof(buffer), "%s/%s", root, namelist[lpc]->d_name); if (stat(buffer, &sb)) { continue; } if (S_ISDIR(sb.st_mode)) { if (files) { free(namelist[lpc]); continue; } } else if (S_ISREG(sb.st_mode)) { if (files == FALSE) { free(namelist[lpc]); continue; } else if (executable && (sb.st_mode & S_IXUSR) == 0 && (sb.st_mode & S_IXGRP) == 0 && (sb.st_mode & S_IXOTH) == 0) { free(namelist[lpc]); continue; } } list = g_list_append(list, strdup(namelist[lpc]->d_name)); free(namelist[lpc]); } free(namelist); return list; } GList * services_os_get_directory_list(const char *root, gboolean files, gboolean executable) { GList *result = NULL; char *dirs = strdup(root); char *dir = NULL; if (pcmk__str_empty(dirs)) { free(dirs); return result; } for (dir = strtok(dirs, ":"); dir != NULL; dir = strtok(NULL, ":")) { GList *tmp = services_os_get_single_directory_list(dir, files, executable); if (tmp) { result = g_list_concat(result, tmp); } } free(dirs); return result; } diff --git a/lib/services/systemd.c b/lib/services/systemd.c index 6f5bef9606..8e9fff484a 100644 --- a/lib/services/systemd.c +++ b/lib/services/systemd.c @@ -1,1053 +1,1055 @@ /* * Copyright 2012-2021 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 static void invoke_unit_by_path(svc_action_t *op, const char *unit); #define BUS_NAME "org.freedesktop.systemd1" #define BUS_NAME_MANAGER BUS_NAME ".Manager" #define BUS_NAME_UNIT BUS_NAME ".Unit" #define BUS_PATH "/org/freedesktop/systemd1" /*! * \internal * \brief Prepare a systemd action * * \param[in] op Action to prepare * * \return Standard Pacemaker return code */ int services__systemd_prepare(svc_action_t *op) { op->opaque->exec = strdup("systemd-dbus"); if (op->opaque->exec == NULL) { return ENOMEM; } return pcmk_rc_ok; } /*! * \internal * \brief Map a systemd result to a standard OCF result * * \param[in] exit_status Systemd result * * \return Standard OCF result */ enum ocf_exitcode services__systemd2ocf(int exit_status) { // This library uses OCF codes for systemd actions return (enum ocf_exitcode) exit_status; } static inline DBusMessage * systemd_new_method(const char *method) { crm_trace("Calling: %s on " BUS_NAME_MANAGER, method); return dbus_message_new_method_call(BUS_NAME, BUS_PATH, BUS_NAME_MANAGER, method); } /* * Functions to manage a static DBus connection */ static DBusConnection* systemd_proxy = NULL; static inline DBusPendingCall * systemd_send(DBusMessage *msg, void(*done)(DBusPendingCall *pending, void *user_data), void *user_data, int timeout) { return pcmk_dbus_send(msg, systemd_proxy, done, user_data, timeout); } static inline DBusMessage * systemd_send_recv(DBusMessage *msg, DBusError *error, int timeout) { return pcmk_dbus_send_recv(msg, systemd_proxy, error, timeout); } /*! * \internal * \brief Send a method to systemd without arguments, and wait for reply * * \param[in] method Method to send * * \return Systemd reply on success, NULL (and error will be logged) otherwise * * \note The caller must call dbus_message_unref() on the reply after * handling it. */ static DBusMessage * systemd_call_simple_method(const char *method) { DBusMessage *msg = systemd_new_method(method); DBusMessage *reply = NULL; DBusError error; /* Don't call systemd_init() here, because that calls this */ CRM_CHECK(systemd_proxy, return NULL); if (msg == NULL) { crm_err("Could not create message to send %s to systemd", method); return NULL; } dbus_error_init(&error); reply = systemd_send_recv(msg, &error, DBUS_TIMEOUT_USE_DEFAULT); dbus_message_unref(msg); if (dbus_error_is_set(&error)) { crm_err("Could not send %s to systemd: %s (%s)", method, error.message, error.name); dbus_error_free(&error); return NULL; } else if (reply == NULL) { crm_err("Could not send %s to systemd: no reply received", method); return NULL; } return reply; } static gboolean systemd_init(void) { static int need_init = 1; // https://dbus.freedesktop.org/doc/api/html/group__DBusConnection.html if (systemd_proxy && dbus_connection_get_is_connected(systemd_proxy) == FALSE) { crm_warn("Connection to System DBus is closed. Reconnecting..."); pcmk_dbus_disconnect(systemd_proxy); systemd_proxy = NULL; need_init = 1; } if (need_init) { need_init = 0; systemd_proxy = pcmk_dbus_connect(); } if (systemd_proxy == NULL) { return FALSE; } return TRUE; } static inline char * systemd_get_property(const char *unit, const char *name, void (*callback)(const char *name, const char *value, void *userdata), void *userdata, DBusPendingCall **pending, int timeout) { return systemd_proxy? pcmk_dbus_get_property(systemd_proxy, BUS_NAME, unit, BUS_NAME_UNIT, name, callback, userdata, pending, timeout) : NULL; } void systemd_cleanup(void) { if (systemd_proxy) { pcmk_dbus_disconnect(systemd_proxy); systemd_proxy = NULL; } } /* * end of systemd_proxy functions */ /*! * \internal * \brief Check whether a file name represents a manageable systemd unit * * \param[in] name File name to check * * \return Pointer to "dot" before filename extension if so, NULL otherwise */ static const char * systemd_unit_extension(const char *name) { if (name) { const char *dot = strrchr(name, '.'); if (dot && (!strcmp(dot, ".service") || !strcmp(dot, ".socket") || !strcmp(dot, ".mount") || !strcmp(dot, ".timer") || !strcmp(dot, ".path"))) { return dot; } } return NULL; } static char * systemd_service_name(const char *name) { if (name == NULL) { return NULL; } if (systemd_unit_extension(name)) { return strdup(name); } return crm_strdup_printf("%s.service", name); } static void systemd_daemon_reload_complete(DBusPendingCall *pending, void *user_data) { DBusError error; DBusMessage *reply = NULL; unsigned int reload_count = GPOINTER_TO_UINT(user_data); dbus_error_init(&error); if(pending) { reply = dbus_pending_call_steal_reply(pending); } if (pcmk_dbus_find_error(pending, reply, &error)) { - crm_err("Could not issue systemd reload %d: %s", reload_count, error.message); + crm_warn("Could not issue systemd reload %d: %s", + reload_count, error.message); dbus_error_free(&error); } else { crm_trace("Reload %d complete", reload_count); } if(pending) { dbus_pending_call_unref(pending); } if(reply) { dbus_message_unref(reply); } } static bool systemd_daemon_reload(int timeout) { static unsigned int reload_count = 0; DBusMessage *msg = systemd_new_method("Reload"); reload_count++; CRM_ASSERT(msg != NULL); systemd_send(msg, systemd_daemon_reload_complete, GUINT_TO_POINTER(reload_count), timeout); dbus_message_unref(msg); return TRUE; } /*! * \internal * \brief Set an action result based on a method error * * \param[in] op Action to set result for * \param[in] error Method error */ static void set_result_from_method_error(svc_action_t *op, const DBusError *error) { services__set_result(op, PCMK_OCF_UNKNOWN_ERROR, PCMK_EXEC_ERROR, "Unable to invoke systemd DBus method"); if (strstr(error->name, "org.freedesktop.systemd1.InvalidName") || strstr(error->name, "org.freedesktop.systemd1.LoadFailed") || strstr(error->name, "org.freedesktop.systemd1.NoSuchUnit")) { if (pcmk__str_eq(op->action, "stop", pcmk__str_casei)) { crm_trace("Masking systemd stop failure (%s) for %s " "because unknown service can be considered stopped", error->name, crm_str(op->rsc)); services__set_result(op, PCMK_OCF_OK, PCMK_EXEC_DONE, NULL); return; } services__set_result(op, PCMK_OCF_NOT_INSTALLED, PCMK_EXEC_NOT_INSTALLED, "systemd unit not found"); } - crm_err("DBus request for %s of systemd unit %s for resource %s failed: %s", - op->action, op->agent, crm_str(op->rsc), error->message); + crm_info("DBus request for %s of systemd unit %s for resource %s failed: %s", + op->action, op->agent, crm_str(op->rsc), error->message); } /*! * \internal * \brief Extract unit path from LoadUnit reply, and execute action * * \param[in] reply LoadUnit reply * \param[in] op Action to execute (or NULL to just return path) * * \return DBus object path for specified unit if successful (only valid for * lifetime of \p reply), otherwise NULL */ static const char * execute_after_loadunit(DBusMessage *reply, svc_action_t *op) { const char *path = NULL; DBusError error; /* path here is not used other than as a non-NULL flag to indicate that a * request was indeed sent */ if (pcmk_dbus_find_error((void *) &path, reply, &error)) { if (op != NULL) { set_result_from_method_error(op, &error); } dbus_error_free(&error); } else if (!pcmk_dbus_type_check(reply, NULL, DBUS_TYPE_OBJECT_PATH, __func__, __LINE__)) { if (op != NULL) { services__set_result(op, PCMK_OCF_UNKNOWN_ERROR, PCMK_EXEC_ERROR, "systemd DBus method had unexpected reply"); - crm_err("Could not load systemd unit %s for %s: " - "DBus reply has unexpected type", op->agent, op->id); + crm_info("Could not load systemd unit %s for %s: " + "DBus reply has unexpected type", op->agent, op->id); } else { - crm_err("Could not load systemd unit: " - "DBus reply has unexpected type"); + crm_info("Could not load systemd unit: " + "DBus reply has unexpected type"); } } else { dbus_message_get_args (reply, NULL, DBUS_TYPE_OBJECT_PATH, &path, DBUS_TYPE_INVALID); } if (op != NULL) { if (path != NULL) { invoke_unit_by_path(op, path); } else if (!(op->synchronous)) { services__set_result(op, PCMK_OCF_UNKNOWN_ERROR, PCMK_EXEC_ERROR, "No DBus object found for systemd unit"); services__finalize_async_op(op); } } return path; } /*! * \internal * \brief Execute a systemd action after its LoadUnit completes * * \param[in] pending If not NULL, DBus call associated with LoadUnit request * \param[in] user_data Action to execute */ static void loadunit_completed(DBusPendingCall *pending, void *user_data) { DBusMessage *reply = NULL; svc_action_t *op = user_data; crm_trace("LoadUnit result for %s arrived", op->id); // Grab the reply if (pending != NULL) { reply = dbus_pending_call_steal_reply(pending); } // The call is no longer pending CRM_LOG_ASSERT(pending == op->opaque->pending); services_set_op_pending(op, NULL); // Execute the desired action based on the reply execute_after_loadunit(reply, user_data); if (reply != NULL) { dbus_message_unref(reply); } } /*! * \internal * \brief Execute a systemd action, given the unit name * * \param[in] arg_name Unit name (possibly shortened, i.e. without ".service") * \param[in] op Action to execute (if NULL, just get the object path) * \param[out] path If non-NULL and \p op is NULL or synchronous, where to * store DBus object path for specified unit * * \return Standard Pacemaker return code (for NULL \p op, pcmk_rc_ok means unit * was found; for synchronous actions, pcmk_rc_ok means unit was * executed, with the actual result stored in \p op; for asynchronous * actions, pcmk_rc_ok means action was initiated) * \note It is the caller's responsibility to free the path. */ static int invoke_unit_by_name(const char *arg_name, svc_action_t *op, char **path) { DBusMessage *msg; DBusMessage *reply = NULL; DBusPendingCall *pending = NULL; char *name = NULL; if (!systemd_init()) { if (op != NULL) { services__set_result(op, PCMK_OCF_UNKNOWN_ERROR, PCMK_EXEC_ERROR, "No DBus connection"); } return ENOTCONN; } /* Create a LoadUnit DBus method (equivalent to GetUnit if already loaded), * which makes the unit usable via further DBus methods. * * * * * */ msg = systemd_new_method("LoadUnit"); CRM_ASSERT(msg != NULL); // Add the (expanded) unit name as the argument name = systemd_service_name(arg_name); CRM_LOG_ASSERT(dbus_message_append_args(msg, DBUS_TYPE_STRING, &name, DBUS_TYPE_INVALID)); free(name); if ((op == NULL) || op->synchronous) { // For synchronous ops, wait for a reply and extract the result const char *unit = NULL; int rc = pcmk_rc_ok; reply = systemd_send_recv(msg, NULL, (op? op->timeout : DBUS_TIMEOUT_USE_DEFAULT)); dbus_message_unref(msg); unit = execute_after_loadunit(reply, op); if (unit == NULL) { rc = ENOENT; if (path != NULL) { *path = NULL; } } else if (path != NULL) { *path = strdup(unit); if (*path == NULL) { rc = ENOMEM; } } if (reply != NULL) { dbus_message_unref(reply); } return rc; } // For asynchronous ops, initiate the LoadUnit call and return pending = systemd_send(msg, loadunit_completed, op, op->timeout); if (pending == NULL) { services__set_result(op, PCMK_OCF_UNKNOWN_ERROR, PCMK_EXEC_ERROR, "Unable to send DBus message"); dbus_message_unref(msg); return ECOMM; } // LoadUnit was successfully initiated services__set_result(op, PCMK_OCF_UNKNOWN, PCMK_EXEC_PENDING, NULL); services_set_op_pending(op, pending); dbus_message_unref(msg); return pcmk_rc_ok; } /*! * \internal * \brief Compare two strings alphabetically (case-insensitive) * * \param[in] a First string to compare * \param[in] b Second string to compare * * \return 0 if strings are equal, -1 if a < b, 1 if a > b * * \note Usable as a GCompareFunc with g_list_sort(). * NULL is considered less than non-NULL. */ static gint sort_str(gconstpointer a, gconstpointer b) { if (!a && !b) { return 0; } else if (!a) { return -1; } else if (!b) { return 1; } return strcasecmp(a, b); } GList * systemd_unit_listall(void) { int nfiles = 0; GList *units = NULL; DBusMessageIter args; DBusMessageIter unit; DBusMessageIter elem; DBusMessage *reply = NULL; if (systemd_init() == FALSE) { return NULL; } /* " \n" \ " \n" \ " \n" \ */ reply = systemd_call_simple_method("ListUnitFiles"); if (reply == NULL) { return NULL; } if (!dbus_message_iter_init(reply, &args)) { crm_err("Could not list systemd unit files: systemd reply has no arguments"); dbus_message_unref(reply); return NULL; } if (!pcmk_dbus_type_check(reply, &args, DBUS_TYPE_ARRAY, __func__, __LINE__)) { crm_err("Could not list systemd unit files: systemd reply has invalid arguments"); dbus_message_unref(reply); return NULL; } dbus_message_iter_recurse(&args, &unit); for (; dbus_message_iter_get_arg_type(&unit) != DBUS_TYPE_INVALID; dbus_message_iter_next(&unit)) { DBusBasicValue value; const char *match = NULL; char *unit_name = NULL; char *basename = NULL; if(!pcmk_dbus_type_check(reply, &unit, DBUS_TYPE_STRUCT, __func__, __LINE__)) { crm_warn("Skipping systemd reply argument with unexpected type"); continue; } dbus_message_iter_recurse(&unit, &elem); if(!pcmk_dbus_type_check(reply, &elem, DBUS_TYPE_STRING, __func__, __LINE__)) { crm_warn("Skipping systemd reply argument with no string"); continue; } dbus_message_iter_get_basic(&elem, &value); if (value.str == NULL) { crm_debug("ListUnitFiles reply did not provide a string"); continue; } crm_trace("DBus ListUnitFiles listed: %s", value.str); match = systemd_unit_extension(value.str); if (match == NULL) { // This is not a unit file type we know how to manage crm_debug("ListUnitFiles entry '%s' is not supported as resource", value.str); continue; } // ListUnitFiles returns full path names, we just want base name basename = strrchr(value.str, '/'); if (basename) { basename = basename + 1; } else { basename = value.str; } if (!strcmp(match, ".service")) { // Service is the "default" unit type, so strip it unit_name = strndup(basename, match - basename); } else { unit_name = strdup(basename); } nfiles++; units = g_list_prepend(units, unit_name); } dbus_message_unref(reply); crm_trace("Found %d manageable systemd unit files", nfiles); units = g_list_sort(units, sort_str); return units; } gboolean systemd_unit_exists(const char *name) { char *path = NULL; char *state = NULL; /* Note: Makes a blocking dbus calls * Used by resources_find_service_class() when resource class=service */ if ((invoke_unit_by_name(name, NULL, &path) != pcmk_rc_ok) || (path == NULL)) { return FALSE; } /* A successful LoadUnit is not sufficient to determine the unit's * existence; it merely means the LoadUnit request received a reply. * We must make another blocking call to check the LoadState property. */ state = systemd_get_property(path, "LoadState", NULL, NULL, NULL, DBUS_TIMEOUT_USE_DEFAULT); free(path); if (pcmk__str_any_of(state, "loaded", "masked", NULL)) { free(state); return TRUE; } free(state); return FALSE; } #define METADATA_FORMAT \ "\n" \ "\n" \ "\n" \ " 1.1\n" \ " \n" \ " %s\n" \ " \n" \ " systemd unit file for %s\n" \ " \n" \ " \n" \ " \n" \ " \n" \ " \n" \ " \n" \ " \n" \ " \n" \ " \n" \ "\n" static char * systemd_unit_metadata(const char *name, int timeout) { char *meta = NULL; char *desc = NULL; char *path = NULL; if (invoke_unit_by_name(name, NULL, &path) == pcmk_rc_ok) { /* TODO: Worth a making blocking call for? Probably not. Possibly if cached. */ desc = systemd_get_property(path, "Description", NULL, NULL, NULL, timeout); } else { desc = crm_strdup_printf("Systemd unit file for %s", name); } meta = crm_strdup_printf(METADATA_FORMAT, name, desc, name); free(desc); free(path); return meta; } /*! * \internal * \brief Determine result of method from reply * * \param[in] reply Reply to start, stop, or restart request * \param[in] op Action that was executed */ static void process_unit_method_reply(DBusMessage *reply, svc_action_t *op) { DBusError error; /* The first use of error here is not used other than as a non-NULL flag to * indicate that a request was indeed sent */ if (pcmk_dbus_find_error((void *) &error, reply, &error)) { set_result_from_method_error(op, &error); dbus_error_free(&error); } else if (!pcmk_dbus_type_check(reply, NULL, DBUS_TYPE_OBJECT_PATH, __func__, __LINE__)) { - crm_warn("DBus request for %s of %s succeeded but " + crm_info("DBus request for %s of %s succeeded but " "return type was unexpected", op->action, crm_str(op->rsc)); services__set_result(op, PCMK_OCF_OK, PCMK_EXEC_DONE, "systemd DBus method had unexpected reply"); } else { const char *path = NULL; dbus_message_get_args(reply, NULL, DBUS_TYPE_OBJECT_PATH, &path, DBUS_TYPE_INVALID); crm_debug("DBus request for %s of %s using %s succeeded", op->action, crm_str(op->rsc), path); services__set_result(op, PCMK_OCF_OK, PCMK_EXEC_DONE, NULL); } } /*! * \internal * \brief Process the completion of an asynchronous unit start, stop, or restart * * \param[in] pending If not NULL, DBus call associated with request * \param[in] user_data Action that was executed */ static void unit_method_complete(DBusPendingCall *pending, void *user_data) { DBusMessage *reply = NULL; svc_action_t *op = user_data; crm_trace("Result for %s arrived", op->id); // Grab the reply if (pending != NULL) { reply = dbus_pending_call_steal_reply(pending); } // The call is no longer pending CRM_LOG_ASSERT(pending == op->opaque->pending); services_set_op_pending(op, NULL); // Determine result and finalize action process_unit_method_reply(reply, op); services__finalize_async_op(op); if (reply != NULL) { dbus_message_unref(reply); } } #define SYSTEMD_OVERRIDE_ROOT "/run/systemd/system/" /* When the cluster manages a systemd resource, we create a unit file override * to order the service "before" pacemaker. The "before" relationship won't * actually be used, since systemd won't ever start the resource -- we're * interested in the reverse shutdown ordering it creates, to ensure that * systemd doesn't stop the resource at shutdown while pacemaker is still * running. * * @TODO Add start timeout */ #define SYSTEMD_OVERRIDE_TEMPLATE \ "[Unit]\n" \ "Description=Cluster Controlled %s\n" \ "Before=pacemaker.service pacemaker_remote.service\n" \ "\n" \ "[Service]\n" \ "Restart=no\n" // Temporarily use rwxr-xr-x umask when opening a file for writing static FILE * create_world_readable(const char *filename) { mode_t orig_umask = umask(S_IWGRP | S_IWOTH); FILE *fp = fopen(filename, "w"); umask(orig_umask); return fp; } static void create_override_dir(const char *agent) { char *override_dir = crm_strdup_printf(SYSTEMD_OVERRIDE_ROOT "/%s.service.d", agent); int rc = pcmk__build_path(override_dir, 0755); if (rc != pcmk_rc_ok) { crm_warn("Could not create systemd override directory %s: %s", override_dir, pcmk_rc_str(rc)); } free(override_dir); } static char * get_override_filename(const char *agent) { return crm_strdup_printf(SYSTEMD_OVERRIDE_ROOT "/%s.service.d/50-pacemaker.conf", agent); } static void systemd_create_override(const char *agent, int timeout) { FILE *file_strm = NULL; char *override_file = get_override_filename(agent); create_override_dir(agent); /* Ensure the override file is world-readable. This is not strictly * necessary, but it avoids a systemd warning in the logs. */ file_strm = create_world_readable(override_file); if (file_strm == NULL) { crm_err("Cannot open systemd override file %s for writing", override_file); } else { char *override = crm_strdup_printf(SYSTEMD_OVERRIDE_TEMPLATE, agent); int rc = fprintf(file_strm, "%s\n", override); free(override); if (rc < 0) { crm_perror(LOG_WARNING, "Cannot write to systemd override file %s", override_file); } fflush(file_strm); fclose(file_strm); systemd_daemon_reload(timeout); } free(override_file); } static void systemd_remove_override(const char *agent, int timeout) { char *override_file = get_override_filename(agent); int rc = unlink(override_file); if (rc < 0) { // Stop may be called when already stopped, which is fine crm_perror(LOG_DEBUG, "Cannot remove systemd override file %s", override_file); } else { systemd_daemon_reload(timeout); } free(override_file); } /*! * \internal * \brief Parse result of systemd status check * * Set a status action's exit status and execution status based on a DBus * property check result, and finalize the action if asynchronous. * * \param[in] name DBus interface name for property that was checked * \param[in] state Property value * \param[in] userdata Status action that check was done for */ static void parse_status_result(const char *name, const char *state, void *userdata) { svc_action_t *op = userdata; crm_trace("Resource %s has %s='%s'", crm_str(op->rsc), name, crm_str(state)); if (pcmk__str_eq(state, "active", pcmk__str_none)) { services__set_result(op, PCMK_OCF_OK, PCMK_EXEC_DONE, NULL); } else if (pcmk__str_eq(state, "reloading", pcmk__str_none)) { services__set_result(op, PCMK_OCF_OK, PCMK_EXEC_DONE, NULL); } else if (pcmk__str_eq(state, "activating", pcmk__str_none)) { services__set_result(op, PCMK_OCF_UNKNOWN, PCMK_EXEC_PENDING, NULL); } else if (pcmk__str_eq(state, "deactivating", pcmk__str_none)) { services__set_result(op, PCMK_OCF_UNKNOWN, PCMK_EXEC_PENDING, NULL); } else { services__set_result(op, PCMK_OCF_NOT_RUNNING, PCMK_EXEC_DONE, state); } if (!(op->synchronous)) { services_set_op_pending(op, NULL); services__finalize_async_op(op); } } /*! * \internal * \brief Invoke a systemd unit, given its DBus object path * * \param[in] op Action to execute * \param[in] unit DBus object path of systemd unit to invoke */ static void invoke_unit_by_path(svc_action_t *op, const char *unit) { const char *method = NULL; DBusMessage *msg = NULL; DBusMessage *reply = NULL; if (pcmk__str_any_of(op->action, "monitor", "status", NULL)) { DBusPendingCall *pending = NULL; char *state; state = systemd_get_property(unit, "ActiveState", (op->synchronous? NULL : parse_status_result), op, (op->synchronous? NULL : &pending), op->timeout); if (op->synchronous) { parse_status_result("ActiveState", state, op); free(state); } else if (pending == NULL) { // Could not get ActiveState property services__set_result(op, PCMK_OCF_UNKNOWN_ERROR, PCMK_EXEC_ERROR, "Could not get unit state from DBus"); services__finalize_async_op(op); } else { services_set_op_pending(op, pending); } return; } else if (pcmk__str_eq(op->action, "start", pcmk__str_none)) { method = "StartUnit"; systemd_create_override(op->agent, op->timeout); } else if (pcmk__str_eq(op->action, "stop", pcmk__str_none)) { method = "StopUnit"; systemd_remove_override(op->agent, op->timeout); } else if (pcmk__str_eq(op->action, "restart", pcmk__str_none)) { method = "RestartUnit"; } else { services__set_result(op, PCMK_OCF_UNIMPLEMENT_FEATURE, PCMK_EXEC_ERROR, "Action not implemented for systemd resources"); if (!(op->synchronous)) { services__finalize_async_op(op); } return; } crm_trace("Calling %s for unit path %s named %s", method, unit, crm_str(op->rsc)); msg = systemd_new_method(method); CRM_ASSERT(msg != NULL); /* (ss) */ { const char *replace_s = "replace"; char *name = systemd_service_name(op->agent); CRM_LOG_ASSERT(dbus_message_append_args(msg, DBUS_TYPE_STRING, &name, DBUS_TYPE_INVALID)); CRM_LOG_ASSERT(dbus_message_append_args(msg, DBUS_TYPE_STRING, &replace_s, DBUS_TYPE_INVALID)); free(name); } if (op->synchronous) { reply = systemd_send_recv(msg, NULL, op->timeout); dbus_message_unref(msg); process_unit_method_reply(reply, op); if (reply != NULL) { dbus_message_unref(reply); } } else { DBusPendingCall *pending = systemd_send(msg, unit_method_complete, op, op->timeout); dbus_message_unref(msg); if (pending == NULL) { services__set_result(op, PCMK_OCF_UNKNOWN_ERROR, PCMK_EXEC_ERROR, "Unable to send DBus message"); services__finalize_async_op(op); } else { services_set_op_pending(op, pending); } } } static gboolean systemd_timeout_callback(gpointer p) { svc_action_t * op = p; op->opaque->timerid = 0; - crm_warn("%s operation on systemd unit %s named '%s' timed out", op->action, op->agent, op->rsc); + crm_info("%s action for systemd unit %s named '%s' timed out", + op->action, op->agent, op->rsc); services__set_result(op, PCMK_OCF_UNKNOWN_ERROR, PCMK_EXEC_TIMEOUT, "Systemd action did not complete within specified timeout"); services__finalize_async_op(op); return FALSE; } /*! * \internal * \brief Execute a systemd action * * \param[in] op Action to execute * * \return Standard Pacemaker return code * \retval EBUSY Recurring operation could not be initiated * \retval pcmk_rc_error Synchronous action failed * \retval pcmk_rc_ok Synchronous action succeeded, or asynchronous action * should not be freed (because it already was or is * pending) * * \note If the return value for an asynchronous action is not pcmk_rc_ok, the * caller is responsible for freeing the action. */ int services__execute_systemd(svc_action_t *op) { CRM_ASSERT(op != NULL); if ((op->action == NULL) || (op->agent == NULL)) { services__set_result(op, PCMK_OCF_NOT_CONFIGURED, PCMK_EXEC_ERROR_FATAL, "Bug in action caller"); goto done; } if (!systemd_init()) { services__set_result(op, PCMK_OCF_UNKNOWN_ERROR, PCMK_EXEC_ERROR, "No DBus connection"); goto done; } crm_debug("Performing %ssynchronous %s op on systemd unit %s named '%s'", (op->synchronous? "" : "a"), op->action, op->agent, crm_str(op->rsc)); if (pcmk__str_eq(op->action, "meta-data", pcmk__str_casei)) { op->stdout_data = systemd_unit_metadata(op->agent, op->timeout); services__set_result(op, PCMK_OCF_OK, PCMK_EXEC_DONE, NULL); goto done; } /* invoke_unit_by_name() should always override these values, which are here * just as a fail-safe in case there are any code paths that neglect to */ services__set_result(op, PCMK_OCF_UNKNOWN_ERROR, PCMK_EXEC_ERROR, "Bug in service library"); if (invoke_unit_by_name(op->agent, op, NULL) == pcmk_rc_ok) { op->opaque->timerid = g_timeout_add(op->timeout + 5000, systemd_timeout_callback, op); services_add_inflight_op(op); return pcmk_rc_ok; } done: if (op->synchronous) { return (op->rc == PCMK_OCF_OK)? pcmk_rc_ok : pcmk_rc_error; } else { return services__finalize_async_op(op); } } diff --git a/lib/services/upstart.c b/lib/services/upstart.c index 2fdc229adc..2ece803e1b 100644 --- a/lib/services/upstart.c +++ b/lib/services/upstart.c @@ -1,696 +1,697 @@ /* * Original copyright 2010 Senko Rasic * and Ante Karamatic * Later changes copyright 2012-2021 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 #define BUS_NAME "com.ubuntu.Upstart" #define BUS_PATH "/com/ubuntu/Upstart" #define UPSTART_06_API BUS_NAME"0_6" #define UPSTART_JOB_IFACE UPSTART_06_API".Job" #define BUS_PROPERTY_IFACE "org.freedesktop.DBus.Properties" /* http://upstart.ubuntu.com/wiki/DBusInterface */ static DBusConnection *upstart_proxy = NULL; /*! * \internal * \brief Prepare an Upstart action * * \param[in] op Action to prepare * * \return Standard Pacemaker return code */ int services__upstart_prepare(svc_action_t *op) { op->opaque->exec = strdup("upstart-dbus"); if (op->opaque->exec == NULL) { return ENOMEM; } return pcmk_rc_ok; } /*! * \internal * \brief Map a Upstart result to a standard OCF result * * \param[in] exit_status Upstart result * * \return Standard OCF result */ enum ocf_exitcode services__upstart2ocf(int exit_status) { // This library uses OCF codes for Upstart actions return (enum ocf_exitcode) exit_status; } static gboolean upstart_init(void) { static int need_init = 1; if (need_init) { need_init = 0; upstart_proxy = pcmk_dbus_connect(); } if (upstart_proxy == NULL) { return FALSE; } return TRUE; } void upstart_cleanup(void) { if (upstart_proxy) { pcmk_dbus_disconnect(upstart_proxy); upstart_proxy = NULL; } } /*! * \internal * \brief Get the DBus object path corresponding to a job name * * \param[in] arg_name Name of job to get path for * \param[out] path If not NULL, where to store DBus object path * \param[in] timeout Give up after this many seconds * * \return true if object path was found, false otherwise * \note The caller is responsible for freeing *path if it is non-NULL. */ static bool object_path_for_job(const gchar *arg_name, char **path, int timeout) { /* com.ubuntu.Upstart0_6.GetJobByName (in String name, out ObjectPath job) */ DBusError error; DBusMessage *msg; DBusMessage *reply = NULL; bool rc = false; if (path != NULL) { *path = NULL; } if (!upstart_init()) { return false; } msg = dbus_message_new_method_call(BUS_NAME, // target for the method call BUS_PATH, // object to call on UPSTART_06_API, // interface to call on "GetJobByName"); // method name dbus_error_init(&error); CRM_LOG_ASSERT(dbus_message_append_args(msg, DBUS_TYPE_STRING, &arg_name, DBUS_TYPE_INVALID)); reply = pcmk_dbus_send_recv(msg, upstart_proxy, &error, timeout); dbus_message_unref(msg); if (dbus_error_is_set(&error)) { crm_err("Could not get DBus object path for %s: %s", arg_name, error.message); dbus_error_free(&error); } else if (!pcmk_dbus_type_check(reply, NULL, DBUS_TYPE_OBJECT_PATH, __func__, __LINE__)) { crm_err("Could not get DBus object path for %s: Invalid return type", arg_name); } else { if (path != NULL) { dbus_message_get_args(reply, NULL, DBUS_TYPE_OBJECT_PATH, path, DBUS_TYPE_INVALID); if (*path != NULL) { *path = strdup(*path); } } rc = true; } if (reply != NULL) { dbus_message_unref(reply); } return rc; } static void fix(char *input, const char *search, char replace) { char *match = NULL; int shuffle = strlen(search) - 1; while (TRUE) { int len, lpc; match = strstr(input, search); if (match == NULL) { break; } crm_trace("Found: %s", match); match[0] = replace; len = strlen(match) - shuffle; for (lpc = 1; lpc <= len; lpc++) { match[lpc] = match[lpc + shuffle]; } } } static char * fix_upstart_name(const char *input) { char *output = strdup(input); fix(output, "_2b", '+'); fix(output, "_2c", ','); fix(output, "_2d", '-'); fix(output, "_2e", '.'); fix(output, "_40", '@'); fix(output, "_5f", '_'); return output; } GList * upstart_job_listall(void) { GList *units = NULL; DBusMessageIter args; DBusMessageIter unit; DBusMessage *msg = NULL; DBusMessage *reply = NULL; const char *method = "GetAllJobs"; DBusError error; int lpc = 0; if (upstart_init() == FALSE) { return NULL; } /* com.ubuntu.Upstart0_6.GetAllJobs (out jobs) */ dbus_error_init(&error); msg = dbus_message_new_method_call(BUS_NAME, // target for the method call BUS_PATH, // object to call on UPSTART_06_API, // interface to call on method); // method name CRM_ASSERT(msg != NULL); reply = pcmk_dbus_send_recv(msg, upstart_proxy, &error, DBUS_TIMEOUT_USE_DEFAULT); dbus_message_unref(msg); if (dbus_error_is_set(&error)) { crm_err("Call to %s failed: %s", method, error.message); dbus_error_free(&error); return NULL; } else if (!dbus_message_iter_init(reply, &args)) { crm_err("Call to %s failed: Message has no arguments", method); dbus_message_unref(reply); return NULL; } if(!pcmk_dbus_type_check(reply, &args, DBUS_TYPE_ARRAY, __func__, __LINE__)) { crm_err("Call to %s failed: Message has invalid arguments", method); dbus_message_unref(reply); return NULL; } dbus_message_iter_recurse(&args, &unit); while (dbus_message_iter_get_arg_type (&unit) != DBUS_TYPE_INVALID) { DBusBasicValue value; const char *job = NULL; char *path = NULL; if(!pcmk_dbus_type_check(reply, &unit, DBUS_TYPE_OBJECT_PATH, __func__, __LINE__)) { crm_warn("Skipping Upstart reply argument with unexpected type"); continue; } dbus_message_iter_get_basic(&unit, &value); if(value.str) { int llpc = 0; path = value.str; job = value.str; while (path[llpc] != 0) { if (path[llpc] == '/') { job = path + llpc + 1; } llpc++; } lpc++; crm_trace("%s -> %s", path, job); units = g_list_append(units, fix_upstart_name(job)); } dbus_message_iter_next (&unit); } dbus_message_unref(reply); crm_trace("Found %d upstart jobs", lpc); return units; } gboolean upstart_job_exists(const char *name) { return object_path_for_job(name, NULL, DBUS_TIMEOUT_USE_DEFAULT); } static char * get_first_instance(const gchar * job, int timeout) { char *instance = NULL; const char *method = "GetAllInstances"; DBusError error; DBusMessage *msg; DBusMessage *reply; DBusMessageIter args; DBusMessageIter unit; dbus_error_init(&error); msg = dbus_message_new_method_call(BUS_NAME, // target for the method call job, // object to call on UPSTART_JOB_IFACE, // interface to call on method); // method name CRM_ASSERT(msg != NULL); dbus_message_append_args(msg, DBUS_TYPE_INVALID); reply = pcmk_dbus_send_recv(msg, upstart_proxy, &error, timeout); dbus_message_unref(msg); if (dbus_error_is_set(&error)) { - crm_err("Call to %s failed: %s", method, error.message); + crm_info("Call to %s failed: %s", method, error.message); dbus_error_free(&error); goto done; } else if(reply == NULL) { - crm_err("Call to %s failed: no reply", method); + crm_info("Call to %s failed: no reply", method); goto done; } else if (!dbus_message_iter_init(reply, &args)) { - crm_err("Call to %s failed: Message has no arguments", method); + crm_info("Call to %s failed: Message has no arguments", method); goto done; } if(!pcmk_dbus_type_check(reply, &args, DBUS_TYPE_ARRAY, __func__, __LINE__)) { - crm_err("Call to %s failed: Message has invalid arguments", method); + crm_info("Call to %s failed: Message has invalid arguments", method); goto done; } dbus_message_iter_recurse(&args, &unit); if(pcmk_dbus_type_check(reply, &unit, DBUS_TYPE_OBJECT_PATH, __func__, __LINE__)) { DBusBasicValue value; dbus_message_iter_get_basic(&unit, &value); if(value.str) { instance = strdup(value.str); crm_trace("Result: %s", instance); } } done: if(reply) { dbus_message_unref(reply); } return instance; } /*! * \internal * \brief Parse result of Upstart status check * * \param[in] name DBus interface name for property that was checked * \param[in] state Property value * \param[in] userdata Status action that check was done for */ static void parse_status_result(const char *name, const char *state, void *userdata) { svc_action_t *op = userdata; if (pcmk__str_eq(state, "running", pcmk__str_none)) { services__set_result(op, PCMK_OCF_OK, PCMK_EXEC_DONE, NULL); } else { services__set_result(op, PCMK_OCF_NOT_RUNNING, PCMK_EXEC_DONE, state); } if (!(op->synchronous)) { services_set_op_pending(op, NULL); services__finalize_async_op(op); } } #define METADATA_FORMAT \ "\n" \ "\n" \ "\n" \ " 1.1\n" \ " \n" \ " Upstart agent for controlling the system %s service\n" \ " \n" \ " Upstart job for %s\n" \ " \n" \ " \n" \ " \n" \ " \n" \ " \n" \ " \n" \ " \n" \ " \n" \ " \n" \ " \n" \ "\n" static char * upstart_job_metadata(const char *name) { return crm_strdup_printf(METADATA_FORMAT, name, name, name); } /*! * \internal * \brief Set an action result based on a method error * * \param[in] op Action to set result for * \param[in] error Method error */ static void set_result_from_method_error(svc_action_t *op, const DBusError *error) { services__set_result(op, PCMK_OCF_UNKNOWN_ERROR, PCMK_EXEC_ERROR, "Unable to invoke Upstart DBus method"); if (strstr(error->name, UPSTART_06_API ".Error.UnknownInstance")) { if (pcmk__str_eq(op->action, "stop", pcmk__str_casei)) { crm_trace("Masking stop failure (%s) for %s " "because unknown service can be considered stopped", error->name, crm_str(op->rsc)); services__set_result(op, PCMK_OCF_OK, PCMK_EXEC_DONE, NULL); return; } services__set_result(op, PCMK_OCF_NOT_INSTALLED, PCMK_EXEC_NOT_INSTALLED, "Upstart job not found"); } else if (pcmk__str_eq(op->action, "start", pcmk__str_casei) && strstr(error->name, UPSTART_06_API ".Error.AlreadyStarted")) { crm_trace("Masking start failure (%s) for %s " "because already started resource is OK", error->name, crm_str(op->rsc)); services__set_result(op, PCMK_OCF_OK, PCMK_EXEC_DONE, NULL); return; } - crm_err("DBus request for %s of Upstart job %s for resource %s failed: %s", - op->action, op->agent, crm_str(op->rsc), error->message); + crm_info("DBus request for %s of Upstart job %s for resource %s failed: %s", + op->action, op->agent, crm_str(op->rsc), error->message); } /*! * \internal * \brief Process the completion of an asynchronous job start, stop, or restart * * \param[in] pending If not NULL, DBus call associated with request * \param[in] user_data Action that was executed */ static void job_method_complete(DBusPendingCall *pending, void *user_data) { DBusError error; DBusMessage *reply = NULL; svc_action_t *op = user_data; // Grab the reply if (pending != NULL) { reply = dbus_pending_call_steal_reply(pending); } // Determine result dbus_error_init(&error); if (pcmk_dbus_find_error(pending, reply, &error)) { set_result_from_method_error(op, &error); dbus_error_free(&error); } else if (pcmk__str_eq(op->action, "stop", pcmk__str_none)) { // Call has no return value crm_debug("DBus request for stop of %s succeeded", crm_str(op->rsc)); services__set_result(op, PCMK_OCF_OK, PCMK_EXEC_DONE, NULL); } else if (!pcmk_dbus_type_check(reply, NULL, DBUS_TYPE_OBJECT_PATH, __func__, __LINE__)) { - crm_warn("DBus request for %s of %s succeeded but " + crm_info("DBus request for %s of %s succeeded but " "return type was unexpected", op->action, crm_str(op->rsc)); services__set_result(op, PCMK_OCF_OK, PCMK_EXEC_DONE, NULL); } else { const char *path = NULL; dbus_message_get_args(reply, NULL, DBUS_TYPE_OBJECT_PATH, &path, DBUS_TYPE_INVALID); crm_debug("DBus request for %s of %s using %s succeeded", op->action, crm_str(op->rsc), path); services__set_result(op, PCMK_OCF_OK, PCMK_EXEC_DONE, NULL); } // The call is no longer pending CRM_LOG_ASSERT(pending == op->opaque->pending); services_set_op_pending(op, NULL); // Finalize action services__finalize_async_op(op); if (reply != NULL) { dbus_message_unref(reply); } } /*! * \internal * \brief Execute an Upstart action * * \param[in] op Action to execute * * \return Standard Pacemaker return code * \retval EBUSY Recurring operation could not be initiated * \retval pcmk_rc_error Synchronous action failed * \retval pcmk_rc_ok Synchronous action succeeded, or asynchronous action * should not be freed (because it already was or is * pending) * * \note If the return value for an asynchronous action is not pcmk_rc_ok, the * caller is responsible for freeing the action. */ int services__execute_upstart(svc_action_t *op) { char *job = NULL; int arg_wait = TRUE; const char *arg_env = "pacemaker=1"; const char *action = op->action; DBusError error; DBusMessage *msg = NULL; DBusMessage *reply = NULL; DBusMessageIter iter, array_iter; CRM_ASSERT(op != NULL); if ((op->action == NULL) || (op->agent == NULL)) { services__set_result(op, PCMK_OCF_NOT_CONFIGURED, PCMK_EXEC_ERROR_FATAL, "Bug in action caller"); goto cleanup; } if (!upstart_init()) { services__set_result(op, PCMK_OCF_UNKNOWN_ERROR, PCMK_EXEC_ERROR, "No DBus connection"); goto cleanup; } if (pcmk__str_eq(op->action, "meta-data", pcmk__str_casei)) { op->stdout_data = upstart_job_metadata(op->agent); services__set_result(op, PCMK_OCF_OK, PCMK_EXEC_DONE, NULL); goto cleanup; } if (!object_path_for_job(op->agent, &job, op->timeout)) { if (pcmk__str_eq(action, "stop", pcmk__str_none)) { services__set_result(op, PCMK_OCF_OK, PCMK_EXEC_DONE, NULL); } else { services__set_result(op, PCMK_OCF_NOT_INSTALLED, PCMK_EXEC_NOT_INSTALLED, "Upstart job not found"); } goto cleanup; } if (job == NULL) { // Shouldn't normally be possible -- maybe a memory error op->rc = PCMK_OCF_UNKNOWN_ERROR; op->status = PCMK_EXEC_ERROR; goto cleanup; } if (pcmk__strcase_any_of(op->action, "monitor", "status", NULL)) { DBusPendingCall *pending = NULL; char *state = NULL; char *path = get_first_instance(job, op->timeout); services__set_result(op, PCMK_OCF_NOT_RUNNING, PCMK_EXEC_DONE, "No Upstart job instances found"); if (path == NULL) { goto cleanup; } state = pcmk_dbus_get_property(upstart_proxy, BUS_NAME, path, UPSTART_06_API ".Instance", "state", op->synchronous? NULL : parse_status_result, op, op->synchronous? NULL : &pending, op->timeout); free(path); if (op->synchronous) { parse_status_result("state", state, op); free(state); } else if (pending == NULL) { services__set_result(op, PCMK_OCF_UNKNOWN_ERROR, PCMK_EXEC_ERROR, "Could not get job state from DBus"); } else { // Successfully initiated async op free(job); services_set_op_pending(op, pending); services_add_inflight_op(op); return pcmk_rc_ok; } goto cleanup; } else if (pcmk__str_eq(action, "start", pcmk__str_none)) { action = "Start"; } else if (pcmk__str_eq(action, "stop", pcmk__str_none)) { action = "Stop"; } else if (pcmk__str_eq(action, "restart", pcmk__str_none)) { action = "Restart"; } else { services__set_result(op, PCMK_OCF_UNIMPLEMENT_FEATURE, PCMK_EXEC_ERROR_HARD, "Action not implemented for Upstart resources"); goto cleanup; } // Initialize rc/status in case called functions don't set them services__set_result(op, PCMK_OCF_UNKNOWN_ERROR, PCMK_EXEC_DONE, "Bug in service library"); crm_debug("Calling %s for %s on %s", action, crm_str(op->rsc), job); msg = dbus_message_new_method_call(BUS_NAME, // target for the method call job, // object to call on UPSTART_JOB_IFACE, // interface to call on action); // method name CRM_ASSERT(msg != NULL); dbus_message_iter_init_append (msg, &iter); CRM_LOG_ASSERT(dbus_message_iter_open_container(&iter, DBUS_TYPE_ARRAY, DBUS_TYPE_STRING_AS_STRING, &array_iter)); CRM_LOG_ASSERT(dbus_message_iter_append_basic(&array_iter, DBUS_TYPE_STRING, &arg_env)); CRM_LOG_ASSERT(dbus_message_iter_close_container(&iter, &array_iter)); CRM_LOG_ASSERT(dbus_message_append_args(msg, DBUS_TYPE_BOOLEAN, &arg_wait, DBUS_TYPE_INVALID)); if (!(op->synchronous)) { DBusPendingCall *pending = pcmk_dbus_send(msg, upstart_proxy, job_method_complete, op, op->timeout); if (pending == NULL) { services__set_result(op, PCMK_OCF_UNKNOWN_ERROR, PCMK_EXEC_ERROR, "Unable to send DBus message"); goto cleanup; } else { // Successfully initiated async op free(job); services_set_op_pending(op, pending); services_add_inflight_op(op); return pcmk_rc_ok; } } // Synchronous call dbus_error_init(&error); reply = pcmk_dbus_send_recv(msg, upstart_proxy, &error, op->timeout); if (dbus_error_is_set(&error)) { set_result_from_method_error(op, &error); dbus_error_free(&error); } else if (pcmk__str_eq(op->action, "stop", pcmk__str_none)) { // DBus call does not return a value services__set_result(op, PCMK_OCF_OK, PCMK_EXEC_DONE, NULL); } else if (!pcmk_dbus_type_check(reply, NULL, DBUS_TYPE_OBJECT_PATH, __func__, __LINE__)) { - crm_warn("Call to %s passed but return type was unexpected", op->action); + crm_info("Call to %s passed but return type was unexpected", + op->action); services__set_result(op, PCMK_OCF_OK, PCMK_EXEC_DONE, NULL); } else { const char *path = NULL; dbus_message_get_args(reply, NULL, DBUS_TYPE_OBJECT_PATH, &path, DBUS_TYPE_INVALID); - crm_info("Call to %s passed: %s", op->action, path); + crm_debug("Call to %s passed: %s", op->action, path); services__set_result(op, PCMK_OCF_OK, PCMK_EXEC_DONE, NULL); } cleanup: free(job); if (msg != NULL) { dbus_message_unref(msg); } if (reply != NULL) { dbus_message_unref(reply); } if (op->synchronous) { return (op->rc == PCMK_OCF_OK)? pcmk_rc_ok : pcmk_rc_error; } else { return services__finalize_async_op(op); } }