diff --git a/cts/lab/CIB.py b/cts/lab/CIB.py index defbc1850b..2cbed3580e 100644 --- a/cts/lab/CIB.py +++ b/cts/lab/CIB.py @@ -1,518 +1,518 @@ """ CIB generator for Pacemaker's Cluster Test Suite (CTS) """ __copyright__ = "Copyright 2008-2023 the Pacemaker project contributors" __license__ = "GNU General Public License version 2 or later (GPLv2+) WITHOUT ANY WARRANTY" import os import warnings import tempfile from pacemaker.buildoptions import BuildOptions class CibBase(object): def __init__(self, Factory, tag, _id, **kwargs): self.tag = tag self.name = _id self.kwargs = kwargs self.children = [] self.Factory = Factory def __repr__(self): return "%s-%s" % (self.tag, self.name) def add_child(self, child): self.children.append(child) def __setitem__(self, key, value): if value: self.kwargs[key] = value else: self.kwargs.pop(key, None) from cts.cib_xml import * class ConfigBase(object): cts_cib = None version = "unknown" Factory = None def __init__(self, CM, factory, tmpfile=None): self.CM = CM self.Factory = factory if not tmpfile: warnings.filterwarnings("ignore") f=tempfile.NamedTemporaryFile(delete=True) f.close() tmpfile = f.name warnings.resetwarnings() self.Factory.tmpfile = tmpfile def version(self): return self.version def NextIP(self): ip = self.CM.Env["IPBase"] if ":" in ip: (prefix, sep, suffix) = ip.rpartition(":") suffix = str(hex(int(suffix, 16)+1)).lstrip("0x") else: (prefix, sep, suffix) = ip.rpartition(".") suffix = str(int(suffix)+1) ip = prefix + sep + suffix self.CM.Env["IPBase"] = ip return ip.strip() class CIB12(ConfigBase): version = "pacemaker-1.2" counter = 1 def _show(self, command=""): output = "" (_, result) = self.Factory.rsh(self.Factory.target, "HOME=/root CIB_file="+self.Factory.tmpfile+" cibadmin -Ql "+command, verbose=1) for line in result: output += line self.Factory.debug("Generated Config: "+line) return output def NewIP(self, name=None, standard="ocf"): if self.CM.Env["IPagent"] == "IPaddr2": ip = self.NextIP() if not name: if ":" in ip: (prefix, sep, suffix) = ip.rpartition(":") name = "r"+suffix else: name = "r"+ip r = Resource(self.Factory, name, self.CM.Env["IPagent"], standard) r["ip"] = ip if ":" in ip: r["cidr_netmask"] = "64" r["nic"] = "eth0" else: r["cidr_netmask"] = "32" else: if not name: name = "r%s%d" % (self.CM.Env["IPagent"], self.counter) self.counter = self.counter + 1 r = Resource(self.Factory, name, self.CM.Env["IPagent"], standard) r.add_op("monitor", "5s") return r def get_node_id(self, node_name): """ Check the cluster configuration for a node ID. """ # We can't account for every possible configuration, # so we only return a node ID if: # * The node is specified in /etc/corosync/corosync.conf # with "ring0_addr:" equal to node_name and "nodeid:" # explicitly specified. # In all other cases, we return 0. node_id = 0 # awkward command: use } as record separator # so each corosync.conf "object" is one record; # match the "node {" record that has "ring0_addr: node_name"; # then print the substring of that record after "nodeid:" (rc, output) = self.Factory.rsh(self.Factory.target, r"""awk -v RS="}" """ r"""'/^(\s*nodelist\s*{)?\s*node\s*{.*(ring0_addr|name):\s*%s(\s+|$)/""" r"""{gsub(/.*nodeid:\s*/,"");gsub(/\s+.*$/,"");print}' %s""" % (node_name, BuildOptions.COROSYNC_CONFIG_FILE), verbose=1) if rc == 0 and len(output) == 1: try: node_id = int(output[0]) except ValueError: node_id = 0 return node_id def install(self, target): old = self.Factory.tmpfile # Force a rebuild self.cts_cib = None self.Factory.tmpfile = BuildOptions.CIB_DIR + "/cib.xml" self.contents(target) self.Factory.rsh(self.Factory.target, "chown " + BuildOptions.DAEMON_USER + " " + self.Factory.tmpfile) self.Factory.tmpfile = old def contents(self, target=None): # fencing resource if self.cts_cib: return self.cts_cib if target: self.Factory.target = target self.Factory.rsh(self.Factory.target, "HOME=/root cibadmin --empty %s > %s" % (self.version, self.Factory.tmpfile)) self.num_nodes = len(self.CM.Env["nodes"]) no_quorum = "stop" if self.num_nodes < 3: no_quorum = "ignore" self.Factory.log("Cluster only has %d nodes, configuring: no-quorum-policy=ignore" % self.num_nodes) # We don't need a nodes section unless we add attributes stn = None # Fencing resource # Define first so that the shell doesn't reject every update if self.CM.Env["DoFencing"]: # Define the "real" fencing device st = Resource(self.Factory, "Fencing", ""+self.CM.Env["stonith-type"], "stonith") # Set a threshold for unreliable stonith devices such as the vmware one st.add_meta("migration-threshold", "5") st.add_op("monitor", "120s", timeout="120s") st.add_op("stop", "0", timeout="60s") st.add_op("start", "0", timeout="60s") # For remote node tests, a cluster node is stopped and brought back up # as a remote node with the name "remote-OLDNAME". To allow fencing # devices to fence these nodes, create a list of all possible node names. all_node_names = [ prefix+n for n in self.CM.Env["nodes"] for prefix in ('', 'remote-') ] # Add all parameters specified by user entries = self.CM.Env["stonith-params"].split(',') for entry in entries: try: (name, value) = entry.split('=', 1) except ValueError: print("Warning: skipping invalid fencing parameter: %s" % entry) continue # Allow user to specify "all" as the node list, and expand it here if name in [ "hostlist", "pcmk_host_list" ] and value == "all": value = ' '.join(all_node_names) st[name] = value st.commit() # Test advanced fencing logic if True: stf_nodes = [] stt_nodes = [] attr_nodes = {} # Create the levels stl = FencingTopology(self.Factory) for node in self.CM.Env["nodes"]: # Remote node tests will rename the node remote_node = "remote-" + node # Randomly assign node to a fencing method - ftype = self.CM.Env.RandomGen.choice(["levels-and", "levels-or ", "broadcast "]) + ftype = self.CM.Env.random_gen.choice(["levels-and", "levels-or ", "broadcast "]) # For levels-and, randomly choose targeting by node name or attribute by = "" if ftype == "levels-and": node_id = self.get_node_id(node) - if node_id == 0 or self.CM.Env.RandomGen.choice([True, False]): + if node_id == 0 or self.CM.Env.random_gen.choice([True, False]): by = " (by name)" else: attr_nodes[node] = node_id by = " (by attribute)" self.CM.log(" - Using %s fencing for node: %s%s" % (ftype, node, by)) if ftype == "levels-and": # If targeting by name, add a topology level for this node if node not in attr_nodes: stl.level(1, node, "FencingPass,Fencing") # Always target remote nodes by name, otherwise we would need to add # an attribute to the remote node only during remote tests (we don't # want nonexistent remote nodes showing up in the non-remote tests). # That complexity is not worth the effort. stl.level(1, remote_node, "FencingPass,Fencing") # Add the node (and its remote equivalent) to the list of levels-and nodes. stt_nodes.extend([node, remote_node]) elif ftype == "levels-or ": for n in [ node, remote_node ]: stl.level(1, n, "FencingFail") stl.level(2, n, "Fencing") stf_nodes.extend([node, remote_node]) # If any levels-and nodes were targeted by attribute, # create the attributes and a level for the attribute. if attr_nodes: stn = Nodes(self.Factory) for (node_name, node_id) in list(attr_nodes.items()): stn.add_node(node_name, node_id, { "cts-fencing" : "levels-and" }) stl.level(1, None, "FencingPass,Fencing", "cts-fencing", "levels-and") # Create a Dummy agent that always passes for levels-and if len(stt_nodes): stt = Resource(self.Factory, "FencingPass", "fence_dummy", "stonith") stt["pcmk_host_list"] = " ".join(stt_nodes) # Wait this many seconds before doing anything, handy for letting disks get flushed too stt["random_sleep_range"] = "30" stt["mode"] = "pass" stt.commit() # Create a Dummy agent that always fails for levels-or if len(stf_nodes): stf = Resource(self.Factory, "FencingFail", "fence_dummy", "stonith") stf["pcmk_host_list"] = " ".join(stf_nodes) # Wait this many seconds before doing anything, handy for letting disks get flushed too stf["random_sleep_range"] = "30" stf["mode"] = "fail" stf.commit() # Now commit the levels themselves stl.commit() o = Option(self.Factory) o["stonith-enabled"] = self.CM.Env["DoFencing"] o["start-failure-is-fatal"] = "false" o["pe-input-series-max"] = "5000" o["shutdown-escalation"] = "5min" o["batch-limit"] = "10" o["dc-deadtime"] = "5s" o["no-quorum-policy"] = no_quorum if self.CM.Env["DoBSC"]: o["ident-string"] = "Linux-HA TEST configuration file - REMOVEME!!" o.commit() o = OpDefaults(self.Factory) o["timeout"] = "90s" o.commit() # Commit the nodes section if we defined one if stn is not None: stn.commit() # Add an alerts section if possible if self.Factory.rsh.exists_on_all(self.CM.Env["notification-agent"], self.CM.Env["nodes"]): alerts = Alerts(self.Factory) alerts.add_alert(self.CM.Env["notification-agent"], self.CM.Env["notification-recipient"]) alerts.commit() # Add resources? if self.CM.Env["CIBResource"]: self.add_resources() if self.CM.cluster_monitor == 1: mon = Resource(self.Factory, "cluster_mon", "ocf", "ClusterMon", "pacemaker") mon.add_op("start", "0", requires="nothing") mon.add_op("monitor", "5s", requires="nothing") mon["update"] = "10" mon["extra_options"] = "-r -n" mon["user"] = "abeekhof" mon["htmlfile"] = "/suse/abeekhof/Export/cluster.html" mon.commit() #self._create('''location prefer-dc cluster_mon rule -INFINITY: \#is_dc eq false''') # generate cib self.cts_cib = self._show() if self.Factory.tmpfile != BuildOptions.CIB_DIR + "/cib.xml": self.Factory.rsh(self.Factory.target, "rm -f "+self.Factory.tmpfile) return self.cts_cib def add_resources(self): # Per-node resources for node in self.CM.Env["nodes"]: name = "rsc_"+node r = self.NewIP(name) r.prefer(node, "100") r.commit() # Migrator # Make this slightly sticky (since we have no other location constraints) to avoid relocation during Reattach m = Resource(self.Factory, "migrator","Dummy", "ocf", "pacemaker") m["passwd"] = "whatever" m.add_meta("resource-stickiness","1") m.add_meta("allow-migrate", "1") m.add_op("monitor", "P10S") m.commit() # Ping the test exerciser p = Resource(self.Factory, "ping-1","ping", "ocf", "pacemaker") p.add_op("monitor", "60s") p["host_list"] = self.CM.Env["cts-exerciser"] p["name"] = "connected" p["debug"] = "true" c = Clone(self.Factory, "Connectivity", p) c["globally-unique"] = "false" c.commit() # promotable clone resource s = Resource(self.Factory, "stateful-1", "Stateful", "ocf", "pacemaker") s.add_op("monitor", "15s", timeout="60s") s.add_op("monitor", "16s", timeout="60s", role="Promoted") ms = Clone(self.Factory, "promotable-1", s) ms["promotable"] = "true" ms["clone-max"] = self.num_nodes ms["clone-node-max"] = 1 ms["promoted-max"] = 1 ms["promoted-node-max"] = 1 # Require connectivity to run the promotable clone r = Rule(self.Factory, "connected", "-INFINITY", op="or") r.add_child(Expression(self.Factory, "m1-connected-1", "connected", "lt", "1")) r.add_child(Expression(self.Factory, "m1-connected-2", "connected", "not_defined", None)) ms.prefer("connected", rule=r) ms.commit() # Group Resource g = Group(self.Factory, "group-1") g.add_child(self.NewIP()) if self.CM.Env["have_systemd"]: sysd = Resource(self.Factory, "petulant", "pacemaker-cts-dummyd@10", "service") sysd.add_op("monitor", "P10S") g.add_child(sysd) else: g.add_child(self.NewIP()) g.add_child(self.NewIP()) # Make group depend on the promotable clone g.after("promotable-1", first="promote", then="start") g.colocate("promotable-1", "INFINITY", withrole="Promoted") g.commit() # LSB resource lsb = Resource(self.Factory, "lsb-dummy", "LSBDummy", "lsb") lsb.add_op("monitor", "5s") # LSB with group lsb.after("group-1") lsb.colocate("group-1") lsb.commit() class CIB20(CIB12): version = "pacemaker-2.5" class CIB30(CIB12): version = "pacemaker-3.7" #class HASI(CIB10): # def add_resources(self): # # DLM resource # self._create('''primitive dlm ocf:pacemaker:controld op monitor interval=120s''') # self._create('''clone dlm-clone dlm meta globally-unique=false interleave=true''') # O2CB resource # self._create('''primitive o2cb ocf:ocfs2:o2cb op monitor interval=120s''') # self._create('''clone o2cb-clone o2cb meta globally-unique=false interleave=true''') # self._create('''colocation o2cb-with-dlm INFINITY: o2cb-clone dlm-clone''') # self._create('''order start-o2cb-after-dlm mandatory: dlm-clone o2cb-clone''') class ConfigFactory(object): def __init__(self, CM): self.CM = CM self.rsh = self.CM.rsh self.register("pacemaker12", CIB12, CM, self) self.register("pacemaker20", CIB20, CM, self) self.register("pacemaker30", CIB30, CM, self) # self.register("hae", HASI, CM, self) if not self.CM.Env["ListTests"]: self.target = self.CM.Env["nodes"][0] self.tmpfile = None def log(self, args): self.CM.log("cib: %s" % args) def debug(self, args): self.CM.debug("cib: %s" % args) def register(self, methodName, constructor, *args, **kargs): """register a constructor""" _args = [constructor] _args.extend(args) setattr(self, methodName, ConfigFactoryItem(*_args, **kargs)) def unregister(self, methodName): """unregister a constructor""" delattr(self, methodName) def createConfig(self, name="pacemaker-1.0"): if name == "pacemaker-1.0": name = "pacemaker10"; elif name == "pacemaker-1.2": name = "pacemaker12"; elif name == "pacemaker-2.0": name = "pacemaker20"; elif name.startswith("pacemaker-3."): name = "pacemaker30"; elif name == "hasi": name = "hae"; if hasattr(self, name): return getattr(self, name)() else: self.CM.log("Configuration variant '%s' is unknown. Defaulting to latest config" % name) return self.pacemaker30() class ConfigFactoryItem(object): def __init__(self, function, *args, **kargs): self._function = function self._args = args self._kargs = kargs def __call__(self, *args, **kargs): """call function""" _args = list(self._args) _args.extend(args) _kargs = self._kargs.copy() _kargs.update(kargs) return self._function(*_args,**_kargs) if __name__ == '__main__': """ Unit test (pass cluster node names as command line arguments) """ import cts.CTS import cts.CM_corosync import sys if len(sys.argv) < 2: print("Usage: %s ..." % sys.argv[0]) sys.exit(1) args = [ "--nodes", " ".join(sys.argv[1:]), "--clobber-cib", "--populate-resources", "--stack", "corosync", "--test-ip-base", "fe80::1234:56:7890:1000", "--stonith", "rhcs", ] env = CTS.CtsLab(args) cm = CM_corosync.crm_corosync() CibFactory = ConfigFactory(cm) cib = CibFactory.createConfig("pacemaker-3.0") print(cib.contents()) diff --git a/cts/lab/CTSscenarios.py b/cts/lab/CTSscenarios.py index 79d2aab648..7af5fc8626 100644 --- a/cts/lab/CTSscenarios.py +++ b/cts/lab/CTSscenarios.py @@ -1,601 +1,601 @@ """ Test scenario classes for Pacemaker's Cluster Test Suite (CTS) """ __copyright__ = "Copyright 2000-2021 the Pacemaker project contributors" __license__ = "GNU General Public License version 2 or later (GPLv2+) WITHOUT ANY WARRANTY" import os import re import sys import time from cts.CTStests import CTSTest from cts.CTSaudits import ClusterAudit from cts.watcher import LogWatcher class ScenarioComponent(object): def __init__(self, Env): self.Env = Env def IsApplicable(self): '''Return True if the current ScenarioComponent is applicable in the given LabEnvironment given to the constructor. ''' raise ValueError("Abstract Class member (IsApplicable)") def SetUp(self, CM): '''Set up the given ScenarioComponent''' raise ValueError("Abstract Class member (Setup)") def TearDown(self, CM): '''Tear down (undo) the given ScenarioComponent''' raise ValueError("Abstract Class member (Setup)") class Scenario(object): ( '''The basic idea of a scenario is that of an ordered list of ScenarioComponent objects. Each ScenarioComponent is SetUp() in turn, and then after the tests have been run, they are torn down using TearDown() (in reverse order). A Scenario is applicable to a particular cluster manager iff each ScenarioComponent is applicable. A partially set up scenario is torn down if it fails during setup. ''') def __init__(self, ClusterManager, Components, Audits, Tests): "Initialize the Scenario from the list of ScenarioComponents" self.ClusterManager = ClusterManager self.Components = Components self.Audits = Audits self.Tests = Tests self.BadNews = None self.TestSets = [] self.Stats = {"success":0, "failure":0, "BadNews":0, "skipped":0} self.Sets = [] #self.ns=CTS.NodeStatus(self.Env) for comp in Components: if not issubclass(comp.__class__, ScenarioComponent): raise ValueError("Init value must be subclass of ScenarioComponent") for audit in Audits: if not issubclass(audit.__class__, ClusterAudit): raise ValueError("Init value must be subclass of ClusterAudit") for test in Tests: if not issubclass(test.__class__, CTSTest): raise ValueError("Init value must be a subclass of CTSTest") def IsApplicable(self): ( '''A Scenario IsApplicable() iff each of its ScenarioComponents IsApplicable() ''' ) for comp in self.Components: if not comp.IsApplicable(): return None return True def SetUp(self): '''Set up the Scenario. Return TRUE on success.''' self.ClusterManager.prepare() self.audit() # Also detects remote/local log config self.ClusterManager.StatsMark(0) self.ClusterManager.ns.WaitForAllNodesToComeUp(self.ClusterManager.Env["nodes"]) self.audit() self.ClusterManager.install_support() self.BadNews = LogWatcher(self.ClusterManager.Env["LogFileName"], self.ClusterManager.templates.get_patterns("BadNews"), "BadNews", 0, kind=self.ClusterManager.Env["LogWatcher"], hosts=self.ClusterManager.Env["nodes"]) self.BadNews.setwatch() # Call after we've figured out what type of log watching to do in LogAudit j = 0 while j < len(self.Components): if not self.Components[j].SetUp(self.ClusterManager): # OOPS! We failed. Tear partial setups down. self.audit() self.ClusterManager.log("Tearing down partial setup") self.TearDown(j) return None j = j + 1 self.audit() return 1 def TearDown(self, max=None): '''Tear Down the Scenario - in reverse order.''' if max == None: max = len(self.Components)-1 j = max while j >= 0: self.Components[j].TearDown(self.ClusterManager) j = j - 1 self.audit() self.ClusterManager.StatsExtract() self.ClusterManager.install_support("uninstall") def incr(self, name): '''Increment (or initialize) the value associated with the given name''' if not name in self.Stats: self.Stats[name] = 0 self.Stats[name] = self.Stats[name]+1 def run(self, Iterations): self.ClusterManager.oprofileStart() try: self.run_loop(Iterations) self.ClusterManager.oprofileStop() except: self.ClusterManager.oprofileStop() raise def run_loop(self, Iterations): raise ValueError("Abstract Class member (run_loop)") def run_test(self, test, testcount): - nodechoice = self.ClusterManager.Env.RandomNode() + nodechoice = self.ClusterManager.Env.random_node() ret = 1 where = "" did_run = 0 self.ClusterManager.StatsMark(testcount) self.ClusterManager.instance_errorstoignore_clear() self.ClusterManager.log(("Running test %s" % test.name).ljust(35) + (" (%s) " % nodechoice).ljust(15) + "[" + ("%d" % testcount).rjust(3) + "]") starttime = test.set_timer() if not test.setup(nodechoice): self.ClusterManager.log("Setup failed") ret = 0 elif not test.canrunnow(nodechoice): self.ClusterManager.log("Skipped") test.skipped() else: did_run = 1 ret = test(nodechoice) if not test.teardown(nodechoice): self.ClusterManager.log("Teardown failed") if self.ClusterManager.Env["continue"]: answer = "Y" else: try: answer = input('Continue? [nY]') except EOFError as e: answer = "n" if answer and answer == "n": raise ValueError("Teardown of %s on %s failed" % (test.name, nodechoice)) ret = 0 stoptime = time.time() self.ClusterManager.oprofileSave(testcount) elapsed_time = stoptime - starttime test_time = stoptime - test.get_timer() if not test["min_time"]: test["elapsed_time"] = elapsed_time test["min_time"] = test_time test["max_time"] = test_time else: test["elapsed_time"] = test["elapsed_time"] + elapsed_time if test_time < test["min_time"]: test["min_time"] = test_time if test_time > test["max_time"]: test["max_time"] = test_time if ret: self.incr("success") test.log_timer() else: self.incr("failure") self.ClusterManager.statall() did_run = 1 # Force the test count to be incremented anyway so test extraction works self.audit(test.errorstoignore()) return did_run def summarize(self): self.ClusterManager.log("****************") self.ClusterManager.log("Overall Results:" + repr(self.Stats)) self.ClusterManager.log("****************") stat_filter = { "calls":0, "failure":0, "skipped":0, "auditfail":0, } self.ClusterManager.log("Test Summary") for test in self.Tests: for key in list(stat_filter.keys()): stat_filter[key] = test.Stats[key] self.ClusterManager.log(("Test %s: "%test.name).ljust(25) + " %s"%repr(stat_filter)) self.ClusterManager.debug("Detailed Results") for test in self.Tests: self.ClusterManager.debug(("Test %s: "%test.name).ljust(25) + " %s"%repr(test.Stats)) self.ClusterManager.log("<<<<<<<<<<<<<<<< TESTS COMPLETED") def audit(self, LocalIgnore=[]): errcount = 0 ignorelist = [] ignorelist.append("CTS:") ignorelist.extend(LocalIgnore) ignorelist.extend(self.ClusterManager.errorstoignore()) ignorelist.extend(self.ClusterManager.instance_errorstoignore()) # This makes sure everything is stabilized before starting... failed = 0 for audit in self.Audits: if not audit(): self.ClusterManager.log("Audit " + audit.name() + " FAILED.") failed += 1 else: self.ClusterManager.debug("Audit " + audit.name() + " passed.") while errcount < 1000: match = None if self.BadNews: match = self.BadNews.look(0) if match: add_err = 1 for ignore in ignorelist: if add_err == 1 and re.search(ignore, match): add_err = 0 if add_err == 1: self.ClusterManager.log("BadNews: " + match) self.incr("BadNews") errcount = errcount + 1 else: break else: if self.ClusterManager.Env["continue"]: answer = "Y" else: try: answer = input('Big problems. Continue? [nY]') except EOFError as e: answer = "n" if answer and answer == "n": self.ClusterManager.log("Shutting down.") self.summarize() self.TearDown() raise ValueError("Looks like we hit a BadNews jackpot!") if self.BadNews: self.BadNews.end() return failed class AllOnce(Scenario): '''Every Test Once''' # Accessable as __doc__ def run_loop(self, Iterations): testcount = 1 for test in self.Tests: self.run_test(test, testcount) testcount += 1 class RandomTests(Scenario): '''Random Test Execution''' def run_loop(self, Iterations): testcount = 1 while testcount <= Iterations: - test = self.ClusterManager.Env.RandomGen.choice(self.Tests) + test = self.ClusterManager.Env.random_gen.choice(self.Tests) self.run_test(test, testcount) testcount += 1 class BasicSanity(Scenario): '''Basic Cluster Sanity''' def run_loop(self, Iterations): testcount = 1 while testcount <= Iterations: - test = self.Environment.RandomGen.choice(self.Tests) + test = self.Environment.random_gen.choice(self.Tests) self.run_test(test, testcount) testcount += 1 class Sequence(Scenario): '''Named Tests in Sequence''' def run_loop(self, Iterations): testcount = 1 while testcount <= Iterations: for test in self.Tests: self.run_test(test, testcount) testcount += 1 class Boot(Scenario): '''Start the Cluster''' def run_loop(self, Iterations): testcount = 0 class BootCluster(ScenarioComponent): ( '''BootCluster is the most basic of ScenarioComponents. This ScenarioComponent simply starts the cluster manager on all the nodes. It is fairly robust as it waits for all nodes to come up before starting as they might have been rebooted or crashed for some reason beforehand. ''') def __init__(self, Env): pass def IsApplicable(self): '''BootCluster is so generic it is always Applicable''' return True def SetUp(self, CM): '''Basic Cluster Manager startup. Start everything''' CM.prepare() # Clear out the cobwebs ;-) CM.stopall(verbose=True, force=True) # Now start the Cluster Manager on all the nodes. CM.log("Starting Cluster Manager on all nodes.") return CM.startall(verbose=True, quick=True) def TearDown(self, CM, force=False): '''Set up the given ScenarioComponent''' # Stop the cluster manager everywhere CM.log("Stopping Cluster Manager on all nodes") return CM.stopall(verbose=True, force=force) class LeaveBooted(BootCluster): def TearDown(self, CM): '''Set up the given ScenarioComponent''' # Stop the cluster manager everywhere CM.log("Leaving Cluster running on all nodes") return 1 class PingFest(ScenarioComponent): ( '''PingFest does a flood ping to each node in the cluster from the test machine. If the LabEnvironment Parameter PingSize is set, it will be used as the size of ping packet requested (via the -s option). If it is not set, it defaults to 1024 bytes. According to the manual page for ping: Outputs packets as fast as they come back or one hundred times per second, whichever is more. For every ECHO_REQUEST sent a period ``.'' is printed, while for every ECHO_REPLY received a backspace is printed. This provides a rapid display of how many packets are being dropped. Only the super-user may use this option. This can be very hard on a net- work and should be used with caution. ''' ) def __init__(self, Env): self.Env = Env def IsApplicable(self): '''PingFests are always applicable ;-) ''' return True def SetUp(self, CM): '''Start the PingFest!''' self.PingSize = 1024 if "PingSize" in list(CM.Env.keys()): self.PingSize = CM.Env["PingSize"] CM.log("Starting %d byte flood pings" % self.PingSize) self.PingPids = [] for node in CM.Env["nodes"]: self.PingPids.append(self._pingchild(node)) CM.log("Ping PIDs: " + repr(self.PingPids)) return 1 def TearDown(self, CM): '''Stop it right now! My ears are pinging!!''' for pid in self.PingPids: if pid != None: CM.log("Stopping ping process %d" % pid) os.kill(pid, signal.SIGKILL) def _pingchild(self, node): Args = ["ping", "-qfn", "-s", str(self.PingSize), node] sys.stdin.flush() sys.stdout.flush() sys.stderr.flush() pid = os.fork() if pid < 0: self.Env.log("Cannot fork ping child") return None if pid > 0: return pid # Otherwise, we're the child process. os.execvp("ping", Args) self.Env.log("Cannot execvp ping: " + repr(Args)) sys.exit(1) class PacketLoss(ScenarioComponent): ( ''' It would be useful to do some testing of CTS with a modest amount of packet loss enabled - so we could see that everything runs like it should with a certain amount of packet loss present. ''') def IsApplicable(self): '''always Applicable''' return True def SetUp(self, CM): '''Reduce the reliability of communications''' if float(CM.Env["XmitLoss"]) == 0 and float(CM.Env["RecvLoss"]) == 0 : return 1 for node in CM.Env["nodes"]: CM.reducecomm_node(node) CM.log("Reduce the reliability of communications") return 1 def TearDown(self, CM): '''Fix the reliability of communications''' if float(CM.Env["XmitLoss"]) == 0 and float(CM.Env["RecvLoss"]) == 0 : return 1 for node in CM.Env["nodes"]: CM.unisolate_node(node) CM.log("Fix the reliability of communications") class BasicSanityCheck(ScenarioComponent): ( ''' ''') def IsApplicable(self): return self.Env["DoBSC"] def SetUp(self, CM): CM.prepare() # Clear out the cobwebs self.TearDown(CM) # Now start the Cluster Manager on all the nodes. CM.log("Starting Cluster Manager on BSC node(s).") return CM.startall() def TearDown(self, CM): CM.log("Stopping Cluster Manager on BSC node(s).") return CM.stopall() class Benchmark(ScenarioComponent): ( ''' ''') def IsApplicable(self): return self.Env["benchmark"] def SetUp(self, CM): CM.prepare() # Clear out the cobwebs self.TearDown(CM, force=True) # Now start the Cluster Manager on all the nodes. CM.log("Starting Cluster Manager on all node(s).") return CM.startall() def TearDown(self, CM): CM.log("Stopping Cluster Manager on all node(s).") return CM.stopall() class RollingUpgrade(ScenarioComponent): ( ''' Test a rolling upgrade between two versions of the stack ''') def __init__(self, Env): self.Env = Env def IsApplicable(self): if not self.Env["rpm-dir"]: return None if not self.Env["current-version"]: return None if not self.Env["previous-version"]: return None return True def install(self, node, version): target_dir = "/tmp/rpm-%s" % version src_dir = "%s/%s" % (self.CM.Env["rpm-dir"], version) self.CM.rsh(node, "mkdir -p %s" % target_dir) rc = self.CM.cp("%s/*.rpm %s:%s" % (src_dir, node, target_dir)) self.CM.rsh(node, "rpm -Uvh --force %s/*.rpm" % (target_dir)) return self.success() def upgrade(self, node): return self.install(node, self.CM.Env["current-version"]) def downgrade(self, node): return self.install(node, self.CM.Env["previous-version"]) def SetUp(self, CM): print(repr(self)+"prepare") CM.prepare() # Clear out the cobwebs CM.stopall(force=True) CM.log("Downgrading all nodes to %s." % self.Env["previous-version"]) for node in self.Env["nodes"]: if not self.downgrade(node): CM.log("Couldn't downgrade %s" % node) return None return 1 def TearDown(self, CM): # Stop everything CM.log("Stopping Cluster Manager on Upgrade nodes.") CM.stopall() CM.log("Upgrading all nodes to %s." % self.Env["current-version"]) for node in self.Env["nodes"]: if not self.upgrade(node): CM.log("Couldn't upgrade %s" % node) return None return 1 diff --git a/cts/lab/CTStests.py b/cts/lab/CTStests.py index 1fa16f2531..3eb4d900b6 100644 --- a/cts/lab/CTStests.py +++ b/cts/lab/CTStests.py @@ -1,3178 +1,3178 @@ """ Test-specific classes for Pacemaker's Cluster Test Suite (CTS) """ __copyright__ = "Copyright 2000-2023 the Pacemaker project contributors" __license__ = "GNU General Public License version 2 or later (GPLv2+) WITHOUT ANY WARRANTY" # # SPECIAL NOTE: # # Tests may NOT implement any cluster-manager-specific code in them. # EXTEND the ClusterManager object to provide the base capabilities # the test needs if you need to do something that the current CM classes # do not. Otherwise you screw up the whole point of the object structure # in CTS. # # Thank you. # import os import re import time import subprocess import tempfile from stat import * from cts import CTS from cts.CTSaudits import * from cts.watcher import LogWatcher from pacemaker import BuildOptions from pacemaker._cts.environment import EnvFactory from pacemaker._cts.logging import LogFactory from pacemaker._cts.patterns import PatternSelector from pacemaker._cts.remote import RemoteFactory AllTestClasses = [ ] class CTSTest(object): ''' A Cluster test. We implement the basic set of properties and behaviors for a generic cluster test. Cluster tests track their own statistics. We keep each of the kinds of counts we track as separate {name,value} pairs. ''' def __init__(self, cm): #self.name="the unnamed test" self.Stats = {"calls":0 , "success":0 , "failure":0 , "skipped":0 , "auditfail":0} # if not issubclass(cm.__class__, ClusterManager): # raise ValueError("Must be a ClusterManager object") self.CM = cm self.Env = EnvFactory().getInstance() self.rsh = RemoteFactory().getInstance() self.logger = LogFactory() self.templates = PatternSelector(cm["Name"]) self.Audits = [] self.timeout = 120 self.passed = 1 self.is_loop = 0 self.is_unsafe = 0 self.is_experimental = 0 self.is_container = 0 self.is_valgrind = 0 self.benchmark = 0 # which tests to benchmark self.timer = {} # timers def log(self, args): self.logger.log(args) def debug(self, args): self.logger.debug(args) def has_key(self, key): return key in self.Stats def __setitem__(self, key, value): self.Stats[key] = value def __getitem__(self, key): if str(key) == "0": raise ValueError("Bad call to 'foo in X', should reference 'foo in X.Stats' instead") if key in self.Stats: return self.Stats[key] return None def log_mark(self, msg): self.debug("MARK: test %s %s %d" % (self.name,msg,time.time())) return def get_timer(self,key = "test"): try: return self.timer[key] except: return 0 def set_timer(self,key = "test"): self.timer[key] = time.time() return self.timer[key] def log_timer(self,key = "test"): elapsed = 0 if key in self.timer: elapsed = time.time() - self.timer[key] s = key == "test" and self.name or "%s:%s" % (self.name,key) self.debug("%s runtime: %.2f" % (s, elapsed)) del self.timer[key] return elapsed def incr(self, name): '''Increment (or initialize) the value associated with the given name''' if not name in self.Stats: self.Stats[name] = 0 self.Stats[name] = self.Stats[name]+1 # Reset the test passed boolean if name == "calls": self.passed = 1 def failure(self, reason="none"): '''Increment the failure count''' self.passed = 0 self.incr("failure") self.logger.log(("Test %s" % self.name).ljust(35) + " FAILED: %s" % reason) return None def success(self): '''Increment the success count''' self.incr("success") return 1 def skipped(self): '''Increment the skipped count''' self.incr("skipped") return 1 def __call__(self, node): '''Perform the given test''' raise ValueError("Abstract Class member (__call__)") self.incr("calls") return self.failure() def audit(self): passed = 1 if len(self.Audits) > 0: for audit in self.Audits: if not audit(): self.logger.log("Internal %s Audit %s FAILED." % (self.name, audit.name())) self.incr("auditfail") passed = 0 return passed def setup(self, node): '''Setup the given test''' return self.success() def teardown(self, node): '''Tear down the given test''' return self.success() def create_watch(self, patterns, timeout, name=None): if not name: name = self.name return LogWatcher(self.Env["LogFileName"], patterns, name, timeout, kind=self.Env["LogWatcher"], hosts=self.Env["nodes"]) def local_badnews(self, prefix, watch, local_ignore=[]): errcount = 0 if not prefix: prefix = "LocalBadNews:" ignorelist = [] ignorelist.append(" CTS: ") ignorelist.append(prefix) ignorelist.extend(local_ignore) while errcount < 100: match = watch.look(0) if match: add_err = 1 for ignore in ignorelist: if add_err == 1 and re.search(ignore, match): add_err = 0 if add_err == 1: self.logger.log(prefix + " " + match) errcount = errcount + 1 else: break else: self.logger.log("Too many errors!") watch.end() return errcount def is_applicable(self): return self.is_applicable_common() def is_applicable_common(self): '''Return True if we are applicable in the current test configuration''' #raise ValueError("Abstract Class member (is_applicable)") if self.is_loop and not self.Env["loop-tests"]: return False elif self.is_unsafe and not self.Env["unsafe-tests"]: return False elif self.is_valgrind and not self.Env["valgrind-tests"]: return False elif self.is_experimental and not self.Env["experimental-tests"]: return False elif self.is_container and not self.Env["container-tests"]: return False elif self.Env["benchmark"] and self.benchmark == 0: return False return True def find_ocfs2_resources(self, node): self.r_o2cb = None self.r_ocfs2 = [] (_, lines) = self.rsh(node, "crm_resource -c", verbose=1) for line in lines: if re.search("^Resource", line): r = AuditResource(self.CM, line) if r.rtype == "o2cb" and r.parent != "NA": self.debug("Found o2cb: %s" % self.r_o2cb) self.r_o2cb = r.parent if re.search("^Constraint", line): c = AuditConstraint(self.CM, line) if c.type == "rsc_colocation" and c.target == self.r_o2cb: self.r_ocfs2.append(c.rsc) self.debug("Found ocfs2 filesystems: %s" % repr(self.r_ocfs2)) return len(self.r_ocfs2) def canrunnow(self, node): '''Return TRUE if we can meaningfully run right now''' return 1 def errorstoignore(self): '''Return list of errors which are 'normal' and should be ignored''' return [] class StopTest(CTSTest): '''Stop (deactivate) the cluster manager on a node''' def __init__(self, cm): CTSTest.__init__(self, cm) self.name = "Stop" def __call__(self, node): '''Perform the 'stop' test. ''' self.incr("calls") if self.CM.ShouldBeStatus[node] != "up": return self.skipped() patterns = [] # Technically we should always be able to notice ourselves stopping patterns.append(self.templates["Pat:We_stopped"] % node) # Any active node needs to notice this one left # (note that this won't work if we have multiple partitions) for other in self.Env["nodes"]: if self.CM.ShouldBeStatus[other] == "up" and other != node: patterns.append(self.templates["Pat:They_stopped"] %(other, self.CM.key_for_node(node))) #self.debug("Checking %s will notice %s left"%(other, node)) watch = self.create_watch(patterns, self.Env["DeadTime"]) watch.setwatch() if node == self.CM.OurNode: self.incr("us") else: if self.CM.upcount() <= 1: self.incr("all") else: self.incr("them") self.CM.StopaCM(node) watch_result = watch.lookforall() failreason = None UnmatchedList = "||" if watch.unmatched: (_, output) = self.rsh(node, "/bin/ps axf", verbose=1) for line in output: self.debug(line) (_, output) = self.rsh(node, "/usr/sbin/dlm_tool dump 2>/dev/null", verbose=1) for line in output: self.debug(line) for regex in watch.unmatched: self.logger.log ("ERROR: Shutdown pattern not found: %s" % (regex)) UnmatchedList += regex + "||"; failreason = "Missing shutdown pattern" self.CM.cluster_stable(self.Env["DeadTime"]) if not watch.unmatched or self.CM.upcount() == 0: return self.success() if len(watch.unmatched) >= self.CM.upcount(): return self.failure("no match against (%s)" % UnmatchedList) if failreason == None: return self.success() else: return self.failure(failreason) # # We don't register StopTest because it's better when called by # another test... # class StartTest(CTSTest): '''Start (activate) the cluster manager on a node''' def __init__(self, cm, debug=None): CTSTest.__init__(self,cm) self.name = "start" self.debug = debug def __call__(self, node): '''Perform the 'start' test. ''' self.incr("calls") if self.CM.upcount() == 0: self.incr("us") else: self.incr("them") if self.CM.ShouldBeStatus[node] != "down": return self.skipped() elif self.CM.StartaCM(node): return self.success() else: return self.failure("Startup %s on node %s failed" % (self.Env["Name"], node)) # # We don't register StartTest because it's better when called by # another test... # class FlipTest(CTSTest): '''If it's running, stop it. If it's stopped start it. Overthrow the status quo... ''' def __init__(self, cm): CTSTest.__init__(self,cm) self.name = "Flip" self.start = StartTest(cm) self.stop = StopTest(cm) def __call__(self, node): '''Perform the 'Flip' test. ''' self.incr("calls") if self.CM.ShouldBeStatus[node] == "up": self.incr("stopped") ret = self.stop(node) type = "up->down" # Give the cluster time to recognize it's gone... time.sleep(self.Env["StableTime"]) elif self.CM.ShouldBeStatus[node] == "down": self.incr("started") ret = self.start(node) type = "down->up" else: return self.skipped() self.incr(type) if ret: return self.success() else: return self.failure("%s failure" % type) # Register FlipTest as a good test to run AllTestClasses.append(FlipTest) class RestartTest(CTSTest): '''Stop and restart a node''' def __init__(self, cm): CTSTest.__init__(self,cm) self.name = "Restart" self.start = StartTest(cm) self.stop = StopTest(cm) self.benchmark = 1 def __call__(self, node): '''Perform the 'restart' test. ''' self.incr("calls") self.incr("node:" + node) ret1 = 1 if self.CM.StataCM(node): self.incr("WasStopped") if not self.start(node): return self.failure("start (setup) failure: "+node) self.set_timer() if not self.stop(node): return self.failure("stop failure: "+node) if not self.start(node): return self.failure("start failure: "+node) return self.success() # Register RestartTest as a good test to run AllTestClasses.append(RestartTest) class StonithdTest(CTSTest): def __init__(self, cm): CTSTest.__init__(self, cm) self.name = "Stonithd" self.startall = SimulStartLite(cm) self.benchmark = 1 def __call__(self, node): self.incr("calls") if len(self.Env["nodes"]) < 2: return self.skipped() ret = self.startall(None) if not ret: return self.failure("Setup failed") is_dc = self.CM.is_node_dc(node) watchpats = [] watchpats.append(self.templates["Pat:Fencing_ok"] % node) watchpats.append(self.templates["Pat:NodeFenced"] % node) if self.Env["at-boot"] == 0: self.debug("Expecting %s to stay down" % node) self.CM.ShouldBeStatus[node] = "down" else: self.debug("Expecting %s to come up again %d" % (node, self.Env["at-boot"])) watchpats.append("%s.* S_STARTING -> S_PENDING" % node) watchpats.append("%s.* S_PENDING -> S_NOT_DC" % node) watch = self.create_watch(watchpats, 30 + self.Env["DeadTime"] + self.Env["StableTime"] + self.Env["StartTime"]) watch.setwatch() - origin = self.Env.RandomGen.choice(self.Env["nodes"]) + origin = self.Env.random_gen.choice(self.Env["nodes"]) (rc, _) = self.rsh(origin, "stonith_admin --reboot %s -VVVVVV" % node) if rc == 124: # CRM_EX_TIMEOUT # Look for the patterns, usually this means the required # device was running on the node to be fenced - or that # the required devices were in the process of being loaded # and/or moved # # Effectively the node committed suicide so there will be # no confirmation, but pacemaker should be watching and # fence the node again self.logger.log("Fencing command on %s to fence %s timed out" % (origin, node)) elif origin != node and rc != 0: self.debug("Waiting for the cluster to recover") self.CM.cluster_stable() self.debug("Waiting for fenced node to come back up") self.CM.ns.WaitForAllNodesToComeUp(self.Env["nodes"], 600) self.logger.log("Fencing command on %s failed to fence %s (rc=%d)" % (origin, node, rc)) elif origin == node and rc != 255: # 255 == broken pipe, ie. the node was fenced as expected self.logger.log("Locally originated fencing returned %d" % rc) self.set_timer("fence") matched = watch.lookforall() self.log_timer("fence") self.set_timer("reform") if watch.unmatched: self.logger.log("Patterns not found: " + repr(watch.unmatched)) self.debug("Waiting for the cluster to recover") self.CM.cluster_stable() self.debug("Waiting for fenced node to come back up") self.CM.ns.WaitForAllNodesToComeUp(self.Env["nodes"], 600) self.debug("Waiting for the cluster to re-stabilize with all nodes") is_stable = self.CM.cluster_stable(self.Env["StartTime"]) if not matched: return self.failure("Didn't find all expected patterns") elif not is_stable: return self.failure("Cluster did not become stable") self.log_timer("reform") return self.success() def errorstoignore(self): return [ self.templates["Pat:Fencing_start"] % ".*", self.templates["Pat:Fencing_ok"] % ".*", self.templates["Pat:Fencing_active"], r"error.*: Operation 'reboot' targeting .* by .* for stonith_admin.*: Timer expired", ] def is_applicable(self): if not self.is_applicable_common(): return False if "DoFencing" in list(self.Env.keys()): return self.Env["DoFencing"] return True AllTestClasses.append(StonithdTest) class StartOnebyOne(CTSTest): '''Start all the nodes ~ one by one''' def __init__(self, cm): CTSTest.__init__(self,cm) self.name = "StartOnebyOne" self.stopall = SimulStopLite(cm) self.start = StartTest(cm) self.ns = CTS.NodeStatus(cm.Env) def __call__(self, dummy): '''Perform the 'StartOnebyOne' test. ''' self.incr("calls") # We ignore the "node" parameter... # Shut down all the nodes... ret = self.stopall(None) if not ret: return self.failure("Test setup failed") failed = [] self.set_timer() for node in self.Env["nodes"]: if not self.start(node): failed.append(node) if len(failed) > 0: return self.failure("Some node failed to start: " + repr(failed)) return self.success() # Register StartOnebyOne as a good test to run AllTestClasses.append(StartOnebyOne) class SimulStart(CTSTest): '''Start all the nodes ~ simultaneously''' def __init__(self, cm): CTSTest.__init__(self,cm) self.name = "SimulStart" self.stopall = SimulStopLite(cm) self.startall = SimulStartLite(cm) def __call__(self, dummy): '''Perform the 'SimulStart' test. ''' self.incr("calls") # We ignore the "node" parameter... # Shut down all the nodes... ret = self.stopall(None) if not ret: return self.failure("Setup failed") if not self.startall(None): return self.failure("Startall failed") return self.success() # Register SimulStart as a good test to run AllTestClasses.append(SimulStart) class SimulStop(CTSTest): '''Stop all the nodes ~ simultaneously''' def __init__(self, cm): CTSTest.__init__(self,cm) self.name = "SimulStop" self.startall = SimulStartLite(cm) self.stopall = SimulStopLite(cm) def __call__(self, dummy): '''Perform the 'SimulStop' test. ''' self.incr("calls") # We ignore the "node" parameter... # Start up all the nodes... ret = self.startall(None) if not ret: return self.failure("Setup failed") if not self.stopall(None): return self.failure("Stopall failed") return self.success() # Register SimulStop as a good test to run AllTestClasses.append(SimulStop) class StopOnebyOne(CTSTest): '''Stop all the nodes in order''' def __init__(self, cm): CTSTest.__init__(self,cm) self.name = "StopOnebyOne" self.startall = SimulStartLite(cm) self.stop = StopTest(cm) def __call__(self, dummy): '''Perform the 'StopOnebyOne' test. ''' self.incr("calls") # We ignore the "node" parameter... # Start up all the nodes... ret = self.startall(None) if not ret: return self.failure("Setup failed") failed = [] self.set_timer() for node in self.Env["nodes"]: if not self.stop(node): failed.append(node) if len(failed) > 0: return self.failure("Some node failed to stop: " + repr(failed)) return self.success() # Register StopOnebyOne as a good test to run AllTestClasses.append(StopOnebyOne) class RestartOnebyOne(CTSTest): '''Restart all the nodes in order''' def __init__(self, cm): CTSTest.__init__(self,cm) self.name = "RestartOnebyOne" self.startall = SimulStartLite(cm) def __call__(self, dummy): '''Perform the 'RestartOnebyOne' test. ''' self.incr("calls") # We ignore the "node" parameter... # Start up all the nodes... ret = self.startall(None) if not ret: return self.failure("Setup failed") did_fail = [] self.set_timer() self.restart = RestartTest(self.CM) for node in self.Env["nodes"]: if not self.restart(node): did_fail.append(node) if did_fail: return self.failure("Could not restart %d nodes: %s" % (len(did_fail), repr(did_fail))) return self.success() # Register StopOnebyOne as a good test to run AllTestClasses.append(RestartOnebyOne) class PartialStart(CTSTest): '''Start a node - but tell it to stop before it finishes starting up''' def __init__(self, cm): CTSTest.__init__(self,cm) self.name = "PartialStart" self.startall = SimulStartLite(cm) self.stopall = SimulStopLite(cm) self.stop = StopTest(cm) #self.is_unsafe = 1 def __call__(self, node): '''Perform the 'PartialStart' test. ''' self.incr("calls") ret = self.stopall(None) if not ret: return self.failure("Setup failed") watchpats = [] watchpats.append("pacemaker-controld.*Connecting to .* cluster infrastructure") watch = self.create_watch(watchpats, self.Env["DeadTime"]+10) watch.setwatch() self.CM.StartaCMnoBlock(node) ret = watch.lookforall() if not ret: self.logger.log("Patterns not found: " + repr(watch.unmatched)) return self.failure("Setup of %s failed" % node) ret = self.stop(node) if not ret: return self.failure("%s did not stop in time" % node) return self.success() def errorstoignore(self): '''Return list of errors which should be ignored''' # We might do some fencing in the 2-node case if we make it up far enough return [ r"Executing reboot fencing operation", r"Requesting fencing \([^)]+\) of node ", ] # Register StopOnebyOne as a good test to run AllTestClasses.append(PartialStart) class StandbyTest(CTSTest): def __init__(self, cm): CTSTest.__init__(self,cm) self.name = "Standby" self.benchmark = 1 self.start = StartTest(cm) self.startall = SimulStartLite(cm) # make sure the node is active # set the node to standby mode # check resources, none resource should be running on the node # set the node to active mode # check resouces, resources should have been migrated back (SHOULD THEY?) def __call__(self, node): self.incr("calls") ret = self.startall(None) if not ret: return self.failure("Start all nodes failed") self.debug("Make sure node %s is active" % node) if self.CM.StandbyStatus(node) != "off": if not self.CM.SetStandbyMode(node, "off"): return self.failure("can't set node %s to active mode" % node) self.CM.cluster_stable() status = self.CM.StandbyStatus(node) if status != "off": return self.failure("standby status of %s is [%s] but we expect [off]" % (node, status)) self.debug("Getting resources running on node %s" % node) rsc_on_node = self.CM.active_resources(node) watchpats = [] watchpats.append(r"State transition .* -> S_POLICY_ENGINE") watch = self.create_watch(watchpats, self.Env["DeadTime"]+10) watch.setwatch() self.debug("Setting node %s to standby mode" % node) if not self.CM.SetStandbyMode(node, "on"): return self.failure("can't set node %s to standby mode" % node) self.set_timer("on") ret = watch.lookforall() if not ret: self.logger.log("Patterns not found: " + repr(watch.unmatched)) self.CM.SetStandbyMode(node, "off") return self.failure("cluster didn't react to standby change on %s" % node) self.CM.cluster_stable() status = self.CM.StandbyStatus(node) if status != "on": return self.failure("standby status of %s is [%s] but we expect [on]" % (node, status)) self.log_timer("on") self.debug("Checking resources") bad_run = self.CM.active_resources(node) if len(bad_run) > 0: rc = self.failure("%s set to standby, %s is still running on it" % (node, repr(bad_run))) self.debug("Setting node %s to active mode" % node) self.CM.SetStandbyMode(node, "off") return rc self.debug("Setting node %s to active mode" % node) if not self.CM.SetStandbyMode(node, "off"): return self.failure("can't set node %s to active mode" % node) self.set_timer("off") self.CM.cluster_stable() status = self.CM.StandbyStatus(node) if status != "off": return self.failure("standby status of %s is [%s] but we expect [off]" % (node, status)) self.log_timer("off") return self.success() AllTestClasses.append(StandbyTest) class ValgrindTest(CTSTest): '''Check for memory leaks''' def __init__(self, cm): CTSTest.__init__(self,cm) self.name = "Valgrind" self.stopall = SimulStopLite(cm) self.startall = SimulStartLite(cm) self.is_valgrind = 1 self.is_loop = 1 def setup(self, node): self.incr("calls") ret = self.stopall(None) if not ret: return self.failure("Stop all nodes failed") # @TODO Edit /etc/sysconfig/pacemaker on all nodes to enable valgrind, # and clear any valgrind logs from previous runs. For now, we rely on # the user to do this manually. ret = self.startall(None) if not ret: return self.failure("Start all nodes failed") return self.success() def teardown(self, node): # Return all nodes to normal # @TODO Edit /etc/sysconfig/pacemaker on all nodes to disable valgrind ret = self.stopall(None) if not ret: return self.failure("Stop all nodes failed") return self.success() def find_leaks(self): # Check for leaks # (no longer used but kept in case feature is restored) leaked = [] self.stop = StopTest(self.CM) for node in self.Env["nodes"]: rc = self.stop(node) if not rc: self.failure("Couldn't shut down %s" % node) (rc, _) = self.rsh(node, "grep -e indirectly.*lost:.*[1-9] -e definitely.*lost:.*[1-9] -e (ERROR|error).*SUMMARY:.*[1-9].*errors %s" % self.logger.logPat) if rc != 1: leaked.append(node) self.failure("Valgrind errors detected on %s" % node) (_, output) = self.rsh(node, "grep -e lost: -e SUMMARY: %s" % self.logger.logPat, verbose=1) for line in output: self.logger.log(line) (_, output) = self.rsh(node, "cat %s" % self.logger.logPat, verbose=1) for line in output: self.debug(line) self.rsh(node, "rm -f %s" % self.logger.logPat, verbose=1) return leaked def __call__(self, node): #leaked = self.find_leaks() #if len(leaked) > 0: # return self.failure("Nodes %s leaked" % repr(leaked)) return self.success() def errorstoignore(self): '''Return list of errors which should be ignored''' return [ r"pacemaker-based.*: \*\*\*\*\*\*\*\*\*\*\*\*\*", r"pacemaker-based.*: .* avoid confusing Valgrind", r"HA_VALGRIND_ENABLED", ] class StandbyLoopTest(ValgrindTest): '''Check for memory leaks by putting a node in and out of standby for an hour''' # @TODO This is not a useful test for memory leaks def __init__(self, cm): ValgrindTest.__init__(self,cm) self.name = "StandbyLoop" def __call__(self, node): lpc = 0 delay = 2 failed = 0 done = time.time() + self.Env["loop-minutes"] * 60 while time.time() <= done and not failed: lpc = lpc + 1 time.sleep(delay) if not self.CM.SetStandbyMode(node, "on"): self.failure("can't set node %s to standby mode" % node) failed = lpc time.sleep(delay) if not self.CM.SetStandbyMode(node, "off"): self.failure("can't set node %s to active mode" % node) failed = lpc leaked = self.find_leaks() if failed: return self.failure("Iteration %d failed" % failed) elif len(leaked) > 0: return self.failure("Nodes %s leaked" % repr(leaked)) return self.success() #AllTestClasses.append(StandbyLoopTest) class BandwidthTest(CTSTest): # Tests should not be cluster-manager-specific # If you need to find out cluster manager configuration to do this, then # it should be added to the generic cluster manager API. '''Test the bandwidth which the cluster uses''' def __init__(self, cm): CTSTest.__init__(self, cm) self.name = "Bandwidth" self.start = StartTest(cm) self.__setitem__("min",0) self.__setitem__("max",0) self.__setitem__("totalbandwidth",0) (handle, self.tempfile) = tempfile.mkstemp(".cts") os.close(handle) self.startall = SimulStartLite(cm) def __call__(self, node): '''Perform the Bandwidth test''' self.incr("calls") if self.CM.upcount() < 1: return self.skipped() Path = self.CM.InternalCommConfig() if "ip" not in Path["mediatype"]: return self.skipped() port = Path["port"][0] port = int(port) ret = self.startall(None) if not ret: return self.failure("Test setup failed") time.sleep(5) # We get extra messages right after startup. fstmpfile = "/var/run/band_estimate" dumpcmd = "tcpdump -p -n -c 102 -i any udp port %d > %s 2>&1" \ % (port, fstmpfile) (rc, _) = self.rsh(node, dumpcmd) if rc == 0: farfile = "root@%s:%s" % (node, fstmpfile) self.rsh.copy(farfile, self.tempfile) Bandwidth = self.countbandwidth(self.tempfile) if not Bandwidth: self.logger.log("Could not compute bandwidth.") return self.success() intband = int(Bandwidth + 0.5) self.logger.log("...bandwidth: %d bits/sec" % intband) self.Stats["totalbandwidth"] = self.Stats["totalbandwidth"] + Bandwidth if self.Stats["min"] == 0: self.Stats["min"] = Bandwidth if Bandwidth > self.Stats["max"]: self.Stats["max"] = Bandwidth if Bandwidth < self.Stats["min"]: self.Stats["min"] = Bandwidth self.rsh(node, "rm -f %s" % fstmpfile) os.unlink(self.tempfile) return self.success() else: return self.failure("no response from tcpdump command [%d]!" % rc) def countbandwidth(self, file): fp = open(file, "r") fp.seek(0) count = 0 sum = 0 while 1: line = fp.readline() if not line: return None if re.search("udp",line) or re.search("UDP,", line): count = count + 1 linesplit = line.split(" ") for j in range(len(linesplit)-1): if linesplit[j] == "udp": break if linesplit[j] == "length:": break try: sum = sum + int(linesplit[j+1]) except ValueError: self.logger.log("Invalid tcpdump line: %s" % line) return None T1 = linesplit[0] timesplit = T1.split(":") time2split = timesplit[2].split(".") time1 = (int(timesplit[0])*60+int(timesplit[1]))*60+int(time2split[0])+int(time2split[1])*0.000001 break while count < 100: line = fp.readline() if not line: return None if re.search("udp",line) or re.search("UDP,", line): count = count+1 linessplit = line.split(" ") for j in range(len(linessplit)-1): if linessplit[j] == "udp": break if linessplit[j] == "length:": break try: sum = int(linessplit[j+1]) + sum except ValueError: self.logger.log("Invalid tcpdump line: %s" % line) return None T2 = linessplit[0] timesplit = T2.split(":") time2split = timesplit[2].split(".") time2 = (int(timesplit[0])*60+int(timesplit[1]))*60+int(time2split[0])+int(time2split[1])*0.000001 time = time2-time1 if (time <= 0): return 0 return int((sum*8)/time) def is_applicable(self): '''BandwidthTest never applicable''' return False AllTestClasses.append(BandwidthTest) ################################################################### class MaintenanceMode(CTSTest): ################################################################### def __init__(self, cm): CTSTest.__init__(self,cm) self.name = "MaintenanceMode" self.start = StartTest(cm) self.startall = SimulStartLite(cm) self.max = 30 #self.is_unsafe = 1 self.benchmark = 1 self.action = "asyncmon" self.interval = 0 self.rid = "maintenanceDummy" def toggleMaintenanceMode(self, node, action): pats = [] pats.append(self.templates["Pat:DC_IDLE"]) # fail the resource right after turning Maintenance mode on # verify it is not recovered until maintenance mode is turned off if action == "On": pats.append(self.templates["Pat:RscOpFail"] % (self.action, self.rid)) else: pats.append(self.templates["Pat:RscOpOK"] % ("stop", self.rid)) pats.append(self.templates["Pat:RscOpOK"] % ("start", self.rid)) watch = self.create_watch(pats, 60) watch.setwatch() self.debug("Turning maintenance mode %s" % action) self.rsh(node, self.templates["MaintenanceMode%s" % (action)]) if (action == "On"): self.rsh(node, "crm_resource -V -F -r %s -H %s &>/dev/null" % (self.rid, node)) self.set_timer("recover%s" % (action)) watch.lookforall() self.log_timer("recover%s" % (action)) if watch.unmatched: self.debug("Failed to find patterns when turning maintenance mode %s" % action) return repr(watch.unmatched) return "" def insertMaintenanceDummy(self, node): pats = [] pats.append(("%s.*" % node) + (self.templates["Pat:RscOpOK"] % ("start", self.rid))) watch = self.create_watch(pats, 60) watch.setwatch() self.CM.AddDummyRsc(node, self.rid) self.set_timer("addDummy") watch.lookforall() self.log_timer("addDummy") if watch.unmatched: self.debug("Failed to find patterns when adding maintenance dummy resource") return repr(watch.unmatched) return "" def removeMaintenanceDummy(self, node): pats = [] pats.append(self.templates["Pat:RscOpOK"] % ("stop", self.rid)) watch = self.create_watch(pats, 60) watch.setwatch() self.CM.RemoveDummyRsc(node, self.rid) self.set_timer("removeDummy") watch.lookforall() self.log_timer("removeDummy") if watch.unmatched: self.debug("Failed to find patterns when removing maintenance dummy resource") return repr(watch.unmatched) return "" def managedRscList(self, node): rscList = [] (_, lines) = self.rsh(node, "crm_resource -c", verbose=1) for line in lines: if re.search("^Resource", line): tmp = AuditResource(self.CM, line) if tmp.managed(): rscList.append(tmp.id) return rscList def verifyResources(self, node, rscList, managed): managedList = list(rscList) managed_str = "managed" if not managed: managed_str = "unmanaged" (_, lines) = self.rsh(node, "crm_resource -c", verbose=1) for line in lines: if re.search("^Resource", line): tmp = AuditResource(self.CM, line) if managed and not tmp.managed(): continue elif not managed and tmp.managed(): continue elif managedList.count(tmp.id): managedList.remove(tmp.id) if len(managedList) == 0: self.debug("Found all %s resources on %s" % (managed_str, node)) return True self.logger.log("Could not find all %s resources on %s. %s" % (managed_str, node, managedList)) return False def __call__(self, node): '''Perform the 'MaintenanceMode' test. ''' self.incr("calls") verify_managed = False verify_unmanaged = False failPat = "" ret = self.startall(None) if not ret: return self.failure("Setup failed") # get a list of all the managed resources. We use this list # after enabling maintenance mode to verify all managed resources # become un-managed. After maintenance mode is turned off, we use # this list to verify all the resources become managed again. managedResources = self.managedRscList(node) if len(managedResources) == 0: self.logger.log("No managed resources on %s" % node) return self.skipped() # insert a fake resource we can fail during maintenance mode # so we can verify recovery does not take place until after maintenance # mode is disabled. failPat = failPat + self.insertMaintenanceDummy(node) # toggle maintenance mode ON, then fail dummy resource. failPat = failPat + self.toggleMaintenanceMode(node, "On") # verify all the resources are now unmanaged if self.verifyResources(node, managedResources, False): verify_unmanaged = True # Toggle maintenance mode OFF, verify dummy is recovered. failPat = failPat + self.toggleMaintenanceMode(node, "Off") # verify all the resources are now managed again if self.verifyResources(node, managedResources, True): verify_managed = True # Remove our maintenance dummy resource. failPat = failPat + self.removeMaintenanceDummy(node) self.CM.cluster_stable() if failPat != "": return self.failure("Unmatched patterns: %s" % (failPat)) elif verify_unmanaged is False: return self.failure("Failed to verify resources became unmanaged during maintenance mode") elif verify_managed is False: return self.failure("Failed to verify resources switched back to managed after disabling maintenance mode") return self.success() def errorstoignore(self): '''Return list of errors which should be ignored''' return [ r"Updating failcount for %s" % self.rid, r"schedulerd.*: Recover\s+%s\s+\(.*\)" % self.rid, r"Unknown operation: fail", self.templates["Pat:RscOpOK"] % (self.action, self.rid), r"(ERROR|error).*: Action %s_%s_%d .* initiated outside of a transition" % (self.rid, self.action, self.interval), ] AllTestClasses.append(MaintenanceMode) class ResourceRecover(CTSTest): def __init__(self, cm): CTSTest.__init__(self,cm) self.name = "ResourceRecover" self.start = StartTest(cm) self.startall = SimulStartLite(cm) self.max = 30 self.rid = None self.rid_alt = None #self.is_unsafe = 1 self.benchmark = 1 # these are the values used for the new LRM API call self.action = "asyncmon" self.interval = 0 def __call__(self, node): '''Perform the 'ResourceRecover' test. ''' self.incr("calls") ret = self.startall(None) if not ret: return self.failure("Setup failed") # List all resources active on the node (skip test if none) resourcelist = self.CM.active_resources(node) if len(resourcelist) == 0: self.logger.log("No active resources on %s" % node) return self.skipped() # Choose one resource at random rsc = self.choose_resource(node, resourcelist) if rsc is None: return self.failure("Could not get details of resource '%s'" % self.rid) if rsc.id == rsc.clone_id: self.debug("Failing " + rsc.id) else: self.debug("Failing " + rsc.id + " (also known as " + rsc.clone_id + ")") # Log patterns to watch for (failure, plus restart if managed) pats = [] pats.append(self.templates["Pat:CloneOpFail"] % (self.action, rsc.id, rsc.clone_id)) if rsc.managed(): pats.append(self.templates["Pat:RscOpOK"] % ("stop", self.rid)) if rsc.unique(): pats.append(self.templates["Pat:RscOpOK"] % ("start", self.rid)) else: # Anonymous clones may get restarted with a different clone number pats.append(self.templates["Pat:RscOpOK"] % ("start", ".*")) # Fail resource. (Ideally, we'd fail it twice, to ensure the fail count # is incrementing properly, but it might restart on a different node. # We'd have to temporarily ban it from all other nodes and ensure the # migration-threshold hasn't been reached.) if self.fail_resource(rsc, node, pats) is None: return None # self.failure() already called return self.success() def choose_resource(self, node, resourcelist): """ Choose a random resource to target """ - self.rid = self.Env.RandomGen.choice(resourcelist) + self.rid = self.Env.random_gen.choice(resourcelist) self.rid_alt = self.rid (_, lines) = self.rsh(node, "crm_resource -c", verbose=1) for line in lines: if line.startswith("Resource: "): rsc = AuditResource(self.CM, line) if rsc.id == self.rid: # Handle anonymous clones that get renamed self.rid = rsc.clone_id return rsc return None def get_failcount(self, node): """ Check the fail count of targeted resource on given node """ (rc, lines) = self.rsh(node, "crm_failcount --quiet --query --resource %s " "--operation %s --interval %d " "--node %s" % (self.rid, self.action, self.interval, node), verbose=1) if rc != 0 or len(lines) != 1: self.logger.log("crm_failcount on %s failed (%d): %s" % (node, rc, " // ".join(map(str.strip, lines)))) return -1 try: failcount = int(lines[0]) except (IndexError, ValueError): self.logger.log("crm_failcount output on %s unparseable: %s" % (node, ' '.join(lines))) return -1 return failcount def fail_resource(self, rsc, node, pats): """ Fail the targeted resource, and verify as expected """ orig_failcount = self.get_failcount(node) watch = self.create_watch(pats, 60) watch.setwatch() self.rsh(node, "crm_resource -V -F -r %s -H %s &>/dev/null" % (self.rid, node)) self.set_timer("recover") watch.lookforall() self.log_timer("recover") self.CM.cluster_stable() recovered = self.CM.ResourceLocation(self.rid) if watch.unmatched: return self.failure("Patterns not found: %s" % repr(watch.unmatched)) elif rsc.unique() and len(recovered) > 1: return self.failure("%s is now active on more than one node: %s"%(self.rid, repr(recovered))) elif len(recovered) > 0: self.debug("%s is running on: %s" % (self.rid, repr(recovered))) elif rsc.managed(): return self.failure("%s was not recovered and is inactive" % self.rid) new_failcount = self.get_failcount(node) if new_failcount != (orig_failcount + 1): return self.failure("%s fail count is %d not %d" % (self.rid, new_failcount, orig_failcount + 1)) return 0 # Anything but None is success def errorstoignore(self): '''Return list of errors which should be ignored''' return [ r"Updating failcount for %s" % self.rid, r"schedulerd.*: Recover\s+(%s|%s)\s+\(.*\)" % (self.rid, self.rid_alt), r"Unknown operation: fail", self.templates["Pat:RscOpOK"] % (self.action, self.rid), r"(ERROR|error).*: Action %s_%s_%d .* initiated outside of a transition" % (self.rid, self.action, self.interval), ] AllTestClasses.append(ResourceRecover) class ComponentFail(CTSTest): def __init__(self, cm): CTSTest.__init__(self,cm) self.name = "ComponentFail" self.startall = SimulStartLite(cm) self.complist = cm.Components() self.patterns = [] self.okerrpatterns = [] self.is_unsafe = 1 def __call__(self, node): '''Perform the 'ComponentFail' test. ''' self.incr("calls") self.patterns = [] self.okerrpatterns = [] # start all nodes ret = self.startall(None) if not ret: return self.failure("Setup failed") if not self.CM.cluster_stable(self.Env["StableTime"]): return self.failure("Setup failed - unstable") node_is_dc = self.CM.is_node_dc(node, None) # select a component to kill - chosen = self.Env.RandomGen.choice(self.complist) + chosen = self.Env.random_gen.choice(self.complist) while chosen.dc_only == 1 and node_is_dc == 0: - chosen = self.Env.RandomGen.choice(self.complist) + chosen = self.Env.random_gen.choice(self.complist) self.debug("...component %s (dc=%d,boot=%d)" % (chosen.name, node_is_dc,chosen.triggersreboot)) self.incr(chosen.name) if chosen.name != "corosync": self.patterns.append(self.templates["Pat:ChildKilled"] %(node, chosen.name)) self.patterns.append(self.templates["Pat:ChildRespawn"] %(node, chosen.name)) self.patterns.extend(chosen.pats) if node_is_dc: self.patterns.extend(chosen.dc_pats) # @TODO this should be a flag in the Component if chosen.name in [ "corosync", "pacemaker-based", "pacemaker-fenced" ]: # Ignore actions for fence devices if fencer will respawn # (their registration will be lost, and probes will fail) self.okerrpatterns = [ self.templates["Pat:Fencing_active"] ] (_, lines) = self.rsh(node, "crm_resource -c", verbose=1) for line in lines: if re.search("^Resource", line): r = AuditResource(self.CM, line) if r.rclass == "stonith": self.okerrpatterns.append(self.templates["Pat:Fencing_recover"] % r.id) self.okerrpatterns.append(self.templates["Pat:Fencing_probe"] % r.id) # supply a copy so self.patterns doesn't end up empty tmpPats = [] tmpPats.extend(self.patterns) self.patterns.extend(chosen.badnews_ignore) # Look for STONITH ops, depending on Env["at-boot"] we might need to change the nodes status stonithPats = [] stonithPats.append(self.templates["Pat:Fencing_ok"] % node) stonith = self.create_watch(stonithPats, 0) stonith.setwatch() # set the watch for stable watch = self.create_watch( tmpPats, self.Env["DeadTime"] + self.Env["StableTime"] + self.Env["StartTime"]) watch.setwatch() # kill the component chosen.kill(node) self.debug("Waiting for the cluster to recover") self.CM.cluster_stable() self.debug("Waiting for any fenced node to come back up") self.CM.ns.WaitForAllNodesToComeUp(self.Env["nodes"], 600) self.debug("Waiting for the cluster to re-stabilize with all nodes") self.CM.cluster_stable(self.Env["StartTime"]) self.debug("Checking if %s was shot" % node) shot = stonith.look(60) if shot: self.debug("Found: " + repr(shot)) self.okerrpatterns.append(self.templates["Pat:Fencing_start"] % node) if self.Env["at-boot"] == 0: self.CM.ShouldBeStatus[node] = "down" # If fencing occurred, chances are many (if not all) the expected logs # will not be sent - or will be lost when the node reboots return self.success() # check for logs indicating a graceful recovery matched = watch.lookforall(allow_multiple_matches=1) if watch.unmatched: self.logger.log("Patterns not found: " + repr(watch.unmatched)) self.debug("Waiting for the cluster to re-stabilize with all nodes") is_stable = self.CM.cluster_stable(self.Env["StartTime"]) if not matched: return self.failure("Didn't find all expected %s patterns" % chosen.name) elif not is_stable: return self.failure("Cluster did not become stable after killing %s" % chosen.name) return self.success() def errorstoignore(self): '''Return list of errors which should be ignored''' # Note that okerrpatterns refers to the last time we ran this test # The good news is that this works fine for us... self.okerrpatterns.extend(self.patterns) return self.okerrpatterns AllTestClasses.append(ComponentFail) class SplitBrainTest(CTSTest): '''It is used to test split-brain. when the path between the two nodes break check the two nodes both take over the resource''' def __init__(self,cm): CTSTest.__init__(self,cm) self.name = "SplitBrain" self.start = StartTest(cm) self.startall = SimulStartLite(cm) self.is_experimental = 1 def isolate_partition(self, partition): other_nodes = [] other_nodes.extend(self.Env["nodes"]) for node in partition: try: other_nodes.remove(node) except ValueError: self.logger.log("Node "+node+" not in " + repr(self.Env["nodes"]) + " from " +repr(partition)) if len(other_nodes) == 0: return 1 self.debug("Creating partition: " + repr(partition)) self.debug("Everyone else: " + repr(other_nodes)) for node in partition: if not self.CM.isolate_node(node, other_nodes): self.logger.log("Could not isolate %s" % node) return 0 return 1 def heal_partition(self, partition): other_nodes = [] other_nodes.extend(self.Env["nodes"]) for node in partition: try: other_nodes.remove(node) except ValueError: self.logger.log("Node "+node+" not in " + repr(self.Env["nodes"])) if len(other_nodes) == 0: return 1 self.debug("Healing partition: " + repr(partition)) self.debug("Everyone else: " + repr(other_nodes)) for node in partition: self.CM.unisolate_node(node, other_nodes) def __call__(self, node): '''Perform split-brain test''' self.incr("calls") self.passed = 1 partitions = {} ret = self.startall(None) if not ret: return self.failure("Setup failed") while 1: # Retry until we get multiple partitions partitions = {} p_max = len(self.Env["nodes"]) for node in self.Env["nodes"]: - p = self.Env.RandomGen.randint(1, p_max) + p = self.Env.random_gen.randint(1, p_max) if not p in partitions: partitions[p] = [] partitions[p].append(node) p_max = len(list(partitions.keys())) if p_max > 1: break # else, try again self.debug("Created %d partitions" % p_max) for key in list(partitions.keys()): self.debug("Partition["+str(key)+"]:\t"+repr(partitions[key])) # Disabling STONITH to reduce test complexity for now self.rsh(node, "crm_attribute -V -n stonith-enabled -v false") for key in list(partitions.keys()): self.isolate_partition(partitions[key]) count = 30 while count > 0: if len(self.CM.find_partitions()) != p_max: time.sleep(10) else: break else: self.failure("Expected partitions were not created") # Target number of partitions formed - wait for stability if not self.CM.cluster_stable(): self.failure("Partitioned cluster not stable") # Now audit the cluster state self.CM.partitions_expected = p_max if not self.audit(): self.failure("Audits failed") self.CM.partitions_expected = 1 # And heal them again for key in list(partitions.keys()): self.heal_partition(partitions[key]) # Wait for a single partition to form count = 30 while count > 0: if len(self.CM.find_partitions()) != 1: time.sleep(10) count -= 1 else: break else: self.failure("Cluster did not reform") # Wait for it to have the right number of members count = 30 while count > 0: members = [] partitions = self.CM.find_partitions() if len(partitions) > 0: members = partitions[0].split() if len(members) != len(self.Env["nodes"]): time.sleep(10) count -= 1 else: break else: self.failure("Cluster did not completely reform") # Wait up to 20 minutes - the delay is more preferable than # trying to continue with in a messed up state if not self.CM.cluster_stable(1200): self.failure("Reformed cluster not stable") if self.Env["continue"]: answer = "Y" else: try: answer = input('Continue? [nY]') except EOFError as e: answer = "n" if answer and answer == "n": raise ValueError("Reformed cluster not stable") # Turn fencing back on if self.Env["DoFencing"]: self.rsh(node, "crm_attribute -V -D -n stonith-enabled") self.CM.cluster_stable() if self.passed: return self.success() return self.failure("See previous errors") def errorstoignore(self): '''Return list of errors which are 'normal' and should be ignored''' return [ r"Another DC detected:", r"(ERROR|error).*: .*Application of an update diff failed", r"pacemaker-controld.*:.*not in our membership list", r"CRIT:.*node.*returning after partition", ] def is_applicable(self): if not self.is_applicable_common(): return False return len(self.Env["nodes"]) > 2 AllTestClasses.append(SplitBrainTest) class Reattach(CTSTest): def __init__(self, cm): CTSTest.__init__(self,cm) self.name = "Reattach" self.startall = SimulStartLite(cm) self.restart1 = RestartTest(cm) self.stopall = SimulStopLite(cm) self.is_unsafe = 0 # Handled by canrunnow() def _is_managed(self, node): (_, is_managed) = self.rsh(node, "crm_attribute -t rsc_defaults -n is-managed -q -G -d true", verbose=1) is_managed = is_managed[0].strip() return is_managed == "true" def _set_unmanaged(self, node): self.debug("Disable resource management") self.rsh(node, "crm_attribute -t rsc_defaults -n is-managed -v false") def _set_managed(self, node): self.debug("Re-enable resource management") self.rsh(node, "crm_attribute -t rsc_defaults -n is-managed -D") def setup(self, node): attempt = 0 if not self.startall(None): return None # Make sure we are really _really_ stable and that all # resources, including those that depend on transient node # attributes, are started while not self.CM.cluster_stable(double_check=True): if attempt < 5: attempt += 1 self.debug("Not stable yet, re-testing") else: self.logger.log("Cluster is not stable") return None return 1 def teardown(self, node): # Make sure 'node' is up start = StartTest(self.CM) start(node) if not self._is_managed(node): self.logger.log("Attempting to re-enable resource management on %s" % node) self._set_managed(node) self.CM.cluster_stable() if not self._is_managed(node): self.logger.log("Could not re-enable resource management") return 0 return 1 def canrunnow(self, node): '''Return TRUE if we can meaningfully run right now''' if self.find_ocfs2_resources(node): self.logger.log("Detach/Reattach scenarios are not possible with OCFS2 services present") return 0 return 1 def __call__(self, node): self.incr("calls") pats = [] # Conveniently, the scheduler will display this message when disabling # management, even if fencing is not enabled, so we can rely on it. managed = self.create_watch(["No fencing will be done"], 60) managed.setwatch() self._set_unmanaged(node) if not managed.lookforall(): self.logger.log("Patterns not found: " + repr(managed.unmatched)) return self.failure("Resource management not disabled") pats = [] pats.append(self.templates["Pat:RscOpOK"] % ("start", ".*")) pats.append(self.templates["Pat:RscOpOK"] % ("stop", ".*")) pats.append(self.templates["Pat:RscOpOK"] % ("promote", ".*")) pats.append(self.templates["Pat:RscOpOK"] % ("demote", ".*")) pats.append(self.templates["Pat:RscOpOK"] % ("migrate", ".*")) watch = self.create_watch(pats, 60, "ShutdownActivity") watch.setwatch() self.debug("Shutting down the cluster") ret = self.stopall(None) if not ret: self._set_managed(node) return self.failure("Couldn't shut down the cluster") self.debug("Bringing the cluster back up") ret = self.startall(None) time.sleep(5) # allow ping to update the CIB if not ret: self._set_managed(node) return self.failure("Couldn't restart the cluster") if self.local_badnews("ResourceActivity:", watch): self._set_managed(node) return self.failure("Resources stopped or started during cluster restart") watch = self.create_watch(pats, 60, "StartupActivity") watch.setwatch() # Re-enable resource management (and verify it happened). self._set_managed(node) self.CM.cluster_stable() if not self._is_managed(node): return self.failure("Could not re-enable resource management") # Ignore actions for STONITH resources ignore = [] (_, lines) = self.rsh(node, "crm_resource -c", verbose=1) for line in lines: if re.search("^Resource", line): r = AuditResource(self.CM, line) if r.rclass == "stonith": self.debug("Ignoring start actions for %s" % r.id) ignore.append(self.templates["Pat:RscOpOK"] % ("start", r.id)) if self.local_badnews("ResourceActivity:", watch, ignore): return self.failure("Resources stopped or started after resource management was re-enabled") return ret def errorstoignore(self): '''Return list of errors which should be ignored''' return [ r"resource( was|s were) active at shutdown", ] def is_applicable(self): return True AllTestClasses.append(Reattach) class SpecialTest1(CTSTest): '''Set up a custom test to cause quorum failure issues for Andrew''' def __init__(self, cm): CTSTest.__init__(self,cm) self.name = "SpecialTest1" self.startall = SimulStartLite(cm) self.restart1 = RestartTest(cm) self.stopall = SimulStopLite(cm) def __call__(self, node): '''Perform the 'SpecialTest1' test for Andrew. ''' self.incr("calls") # Shut down all the nodes... ret = self.stopall(None) if not ret: return self.failure("Could not stop all nodes") # Test config recovery when the other nodes come up self.rsh(node, "rm -f " + BuildOptions.CIB_DIR + "/cib*") # Start the selected node ret = self.restart1(node) if not ret: return self.failure("Could not start "+node) # Start all remaining nodes ret = self.startall(None) if not ret: return self.failure("Could not start the remaining nodes") return self.success() def errorstoignore(self): '''Return list of errors which should be ignored''' # Errors that occur as a result of the CIB being wiped return [ r"error.*: v1 patchset error, patch failed to apply: Application of an update diff failed", r"error.*: Resource start-up disabled since no STONITH resources have been defined", r"error.*: Either configure some or disable STONITH with the stonith-enabled option", r"error.*: NOTE: Clusters with shared data need STONITH to ensure data integrity", ] AllTestClasses.append(SpecialTest1) class HAETest(CTSTest): '''Set up a custom test to cause quorum failure issues for Andrew''' def __init__(self, cm): CTSTest.__init__(self,cm) self.name = "HAETest" self.stopall = SimulStopLite(cm) self.startall = SimulStartLite(cm) self.is_loop = 1 def setup(self, node): # Start all remaining nodes ret = self.startall(None) if not ret: return self.failure("Couldn't start all nodes") return self.success() def teardown(self, node): # Stop everything ret = self.stopall(None) if not ret: return self.failure("Couldn't stop all nodes") return self.success() def wait_on_state(self, node, resource, expected_clones, attempts=240): while attempts > 0: active = 0 (rc, lines) = self.rsh(node, "crm_resource -r %s -W -Q" % resource, verbose=1) # Hack until crm_resource does the right thing if rc == 0 and lines: active = len(lines) if len(lines) == expected_clones: return 1 elif rc == 1: self.debug("Resource %s is still inactive" % resource) elif rc == 234: self.logger.log("Unknown resource %s" % resource) return 0 elif rc == 246: self.logger.log("Cluster is inactive") return 0 elif rc != 0: self.logger.log("Call to crm_resource failed, rc=%d" % rc) return 0 else: self.debug("Resource %s is active on %d times instead of %d" % (resource, active, expected_clones)) attempts -= 1 time.sleep(1) return 0 def find_dlm(self, node): self.r_dlm = None (_, lines) = self.rsh(node, "crm_resource -c", verbose=1) for line in lines: if re.search("^Resource", line): r = AuditResource(self.CM, line) if r.rtype == "controld" and r.parent != "NA": self.debug("Found dlm: %s" % self.r_dlm) self.r_dlm = r.parent return 1 return 0 def find_hae_resources(self, node): self.r_dlm = None self.r_o2cb = None self.r_ocfs2 = [] if self.find_dlm(node): self.find_ocfs2_resources(node) def is_applicable(self): if not self.is_applicable_common(): return False if self.Env["Schema"] == "hae": return True return None class HAERoleTest(HAETest): def __init__(self, cm): '''Lars' mount/unmount test for the HA extension. ''' HAETest.__init__(self,cm) self.name = "HAERoleTest" def change_state(self, node, resource, target): (rc, _) = self.rsh(node, "crm_resource -V -r %s -p target-role -v %s --meta" % (resource, target)) return rc def __call__(self, node): self.incr("calls") lpc = 0 failed = 0 delay = 2 done = time.time() + self.Env["loop-minutes"]*60 self.find_hae_resources(node) clone_max = len(self.Env["nodes"]) while time.time() <= done and not failed: lpc = lpc + 1 self.change_state(node, self.r_dlm, "Stopped") if not self.wait_on_state(node, self.r_dlm, 0): self.failure("%s did not go down correctly" % self.r_dlm) failed = lpc self.change_state(node, self.r_dlm, "Started") if not self.wait_on_state(node, self.r_dlm, clone_max): self.failure("%s did not come up correctly" % self.r_dlm) failed = lpc if not self.wait_on_state(node, self.r_o2cb, clone_max): self.failure("%s did not come up correctly" % self.r_o2cb) failed = lpc for fs in self.r_ocfs2: if not self.wait_on_state(node, fs, clone_max): self.failure("%s did not come up correctly" % fs) failed = lpc if failed: return self.failure("iteration %d failed" % failed) return self.success() AllTestClasses.append(HAERoleTest) class HAEStandbyTest(HAETest): '''Set up a custom test to cause quorum failure issues for Andrew''' def __init__(self, cm): HAETest.__init__(self,cm) self.name = "HAEStandbyTest" def change_state(self, node, resource, target): (rc, _) = self.rsh(node, "crm_standby -V -l reboot -v %s" % (target)) return rc def __call__(self, node): self.incr("calls") lpc = 0 failed = 0 done = time.time() + self.Env["loop-minutes"]*60 self.find_hae_resources(node) clone_max = len(self.Env["nodes"]) while time.time() <= done and not failed: lpc = lpc + 1 self.change_state(node, self.r_dlm, "true") if not self.wait_on_state(node, self.r_dlm, clone_max-1): self.failure("%s did not go down correctly" % self.r_dlm) failed = lpc self.change_state(node, self.r_dlm, "false") if not self.wait_on_state(node, self.r_dlm, clone_max): self.failure("%s did not come up correctly" % self.r_dlm) failed = lpc if not self.wait_on_state(node, self.r_o2cb, clone_max): self.failure("%s did not come up correctly" % self.r_o2cb) failed = lpc for fs in self.r_ocfs2: if not self.wait_on_state(node, fs, clone_max): self.failure("%s did not come up correctly" % fs) failed = lpc if failed: return self.failure("iteration %d failed" % failed) return self.success() AllTestClasses.append(HAEStandbyTest) class NearQuorumPointTest(CTSTest): ''' This test brings larger clusters near the quorum point (50%). In addition, it will test doing starts and stops at the same time. Here is how I think it should work: - loop over the nodes and decide randomly which will be up and which will be down Use a 50% probability for each of up/down. - figure out what to do to get into that state from the current state - in parallel, bring up those going up and bring those going down. ''' def __init__(self, cm): CTSTest.__init__(self,cm) self.name = "NearQuorumPoint" def __call__(self, dummy): '''Perform the 'NearQuorumPoint' test. ''' self.incr("calls") startset = [] stopset = [] stonith = self.CM.prepare_fencing_watcher("NearQuorumPoint") #decide what to do with each node for node in self.Env["nodes"]: - action = self.Env.RandomGen.choice(["start","stop"]) - #action = self.Env.RandomGen.choice(["start","stop","no change"]) + action = self.Env.random_gen.choice(["start","stop"]) + #action = self.Env.random_gen.choice(["start","stop","no change"]) if action == "start" : startset.append(node) elif action == "stop" : stopset.append(node) self.debug("start nodes:" + repr(startset)) self.debug("stop nodes:" + repr(stopset)) #add search patterns watchpats = [ ] for node in stopset: if self.CM.ShouldBeStatus[node] == "up": watchpats.append(self.templates["Pat:We_stopped"] % node) for node in startset: if self.CM.ShouldBeStatus[node] == "down": #watchpats.append(self.templates["Pat:NonDC_started"] % node) watchpats.append(self.templates["Pat:Local_started"] % node) else: for stopping in stopset: if self.CM.ShouldBeStatus[stopping] == "up": watchpats.append(self.templates["Pat:They_stopped"] % (node, self.CM.key_for_node(stopping))) if len(watchpats) == 0: return self.skipped() if len(startset) != 0: watchpats.append(self.templates["Pat:DC_IDLE"]) watch = self.create_watch(watchpats, self.Env["DeadTime"]+10) watch.setwatch() #begin actions for node in stopset: if self.CM.ShouldBeStatus[node] == "up": self.CM.StopaCMnoBlock(node) for node in startset: if self.CM.ShouldBeStatus[node] == "down": self.CM.StartaCMnoBlock(node) #get the result if watch.lookforall(): self.CM.cluster_stable() self.CM.fencing_cleanup("NearQuorumPoint", stonith) return self.success() self.logger.log("Warn: Patterns not found: " + repr(watch.unmatched)) #get the "bad" nodes upnodes = [] for node in stopset: if self.CM.StataCM(node) == 1: upnodes.append(node) downnodes = [] for node in startset: if self.CM.StataCM(node) == 0: downnodes.append(node) self.CM.fencing_cleanup("NearQuorumPoint", stonith) if upnodes == [] and downnodes == []: self.CM.cluster_stable() # Make sure they're completely down with no residule for node in stopset: self.rsh(node, self.templates["StopCmd"]) return self.success() if len(upnodes) > 0: self.logger.log("Warn: Unstoppable nodes: " + repr(upnodes)) if len(downnodes) > 0: self.logger.log("Warn: Unstartable nodes: " + repr(downnodes)) return self.failure() def is_applicable(self): return True AllTestClasses.append(NearQuorumPointTest) class RollingUpgradeTest(CTSTest): '''Perform a rolling upgrade of the cluster''' def __init__(self, cm): CTSTest.__init__(self,cm) self.name = "RollingUpgrade" self.start = StartTest(cm) self.stop = StopTest(cm) self.stopall = SimulStopLite(cm) self.startall = SimulStartLite(cm) def setup(self, node): # Start all remaining nodes ret = self.stopall(None) if not ret: return self.failure("Couldn't stop all nodes") for node in self.Env["nodes"]: if not self.downgrade(node, None): return self.failure("Couldn't downgrade %s" % node) ret = self.startall(None) if not ret: return self.failure("Couldn't start all nodes") return self.success() def teardown(self, node): # Stop everything ret = self.stopall(None) if not ret: return self.failure("Couldn't stop all nodes") for node in self.Env["nodes"]: if not self.upgrade(node, None): return self.failure("Couldn't upgrade %s" % node) return self.success() def install(self, node, version, start=1, flags="--force"): target_dir = "/tmp/rpm-%s" % version src_dir = "%s/%s" % (self.Env["rpm-dir"], version) self.logger.log("Installing %s on %s with %s" % (version, node, flags)) if not self.stop(node): return self.failure("stop failure: "+node) self.rsh(node, "mkdir -p %s" % target_dir) self.rsh(node, "rm -f %s/*.rpm" % target_dir) (_, lines) = self.rsh(node, "ls -1 %s/*.rpm" % src_dir, verbose=1) for line in lines: line = line[:-1] rc = self.rsh.copy("%s" % (line), "%s:%s/" % (node, target_dir)) self.rsh(node, "rpm -Uvh %s %s/*.rpm" % (flags, target_dir)) if start and not self.start(node): return self.failure("start failure: "+node) return self.success() def upgrade(self, node, start=1): return self.install(node, self.Env["current-version"], start) def downgrade(self, node, start=1): return self.install(node, self.Env["previous-version"], start, "--force --nodeps") def __call__(self, node): '''Perform the 'Rolling Upgrade' test. ''' self.incr("calls") for node in self.Env["nodes"]: if self.upgrade(node): return self.failure("Couldn't upgrade %s" % node) self.CM.cluster_stable() return self.success() def is_applicable(self): if not self.is_applicable_common(): return None if not "rpm-dir" in list(self.Env.keys()): return None if not "current-version" in list(self.Env.keys()): return None if not "previous-version" in list(self.Env.keys()): return None return 1 # Register RestartTest as a good test to run AllTestClasses.append(RollingUpgradeTest) class BSC_AddResource(CTSTest): '''Add a resource to the cluster''' def __init__(self, cm): CTSTest.__init__(self, cm) self.name = "AddResource" self.resource_offset = 0 self.cib_cmd = """cibadmin -C -o %s -X '%s' """ def __call__(self, node): self.incr("calls") self.resource_offset = self.resource_offset + 1 r_id = "bsc-rsc-%s-%d" % (node, self.resource_offset) start_pat = "pacemaker-controld.*%s_start_0.*confirmed.*ok" patterns = [] patterns.append(start_pat % r_id) watch = self.create_watch(patterns, self.Env["DeadTime"]) watch.setwatch() ip = self.NextIP() if not self.make_ip_resource(node, r_id, "ocf", "IPaddr", ip): return self.failure("Make resource %s failed" % r_id) failed = 0 watch_result = watch.lookforall() if watch.unmatched: for regex in watch.unmatched: self.logger.log ("Warn: Pattern not found: %s" % (regex)) failed = 1 if failed: return self.failure("Resource pattern(s) not found") if not self.CM.cluster_stable(self.Env["DeadTime"]): return self.failure("Unstable cluster") return self.success() def NextIP(self): ip = self.Env["IPBase"] if ":" in ip: fields = ip.rpartition(":") fields[2] = str(hex(int(fields[2], 16)+1)) print(str(hex(int(f[2], 16)+1))) else: fields = ip.rpartition('.') fields[2] = str(int(fields[2])+1) ip = fields[0] + fields[1] + fields[3]; self.Env["IPBase"] = ip return ip.strip() def make_ip_resource(self, node, id, rclass, type, ip): self.logger.log("Creating %s:%s:%s (%s) on %s" % (rclass,type,id,ip,node)) rsc_xml=""" """ % (id, rclass, type, id, id, ip) node_constraint = """ """ % (id, id, id, id, node) rc = 0 (rc, _) = self.rsh(node, self.cib_cmd % ("constraints", node_constraint), verbose=1) if rc != 0: self.logger.log("Constraint creation failed: %d" % rc) return None (rc, _) = self.rsh(node, self.cib_cmd % ("resources", rsc_xml), verbose=1) if rc != 0: self.logger.log("Resource creation failed: %d" % rc) return None return 1 def is_applicable(self): if self.Env["DoBSC"]: return True return None AllTestClasses.append(BSC_AddResource) class SimulStopLite(CTSTest): '''Stop any active nodes ~ simultaneously''' def __init__(self, cm): CTSTest.__init__(self,cm) self.name = "SimulStopLite" def __call__(self, dummy): '''Perform the 'SimulStopLite' setup work. ''' self.incr("calls") self.debug("Setup: " + self.name) # We ignore the "node" parameter... watchpats = [ ] for node in self.Env["nodes"]: if self.CM.ShouldBeStatus[node] == "up": self.incr("WasStarted") watchpats.append(self.templates["Pat:We_stopped"] % node) if len(watchpats) == 0: return self.success() # Stop all the nodes - at about the same time... watch = self.create_watch(watchpats, self.Env["DeadTime"]+10) watch.setwatch() self.set_timer() for node in self.Env["nodes"]: if self.CM.ShouldBeStatus[node] == "up": self.CM.StopaCMnoBlock(node) if watch.lookforall(): # Make sure they're completely down with no residule for node in self.Env["nodes"]: self.rsh(node, self.templates["StopCmd"]) return self.success() did_fail = 0 up_nodes = [] for node in self.Env["nodes"]: if self.CM.StataCM(node) == 1: did_fail = 1 up_nodes.append(node) if did_fail: return self.failure("Active nodes exist: " + repr(up_nodes)) self.logger.log("Warn: All nodes stopped but CTS didn't detect: " + repr(watch.unmatched)) return self.failure("Missing log message: "+repr(watch.unmatched)) def is_applicable(self): '''SimulStopLite is a setup test and never applicable''' return False class SimulStartLite(CTSTest): '''Start any stopped nodes ~ simultaneously''' def __init__(self, cm): CTSTest.__init__(self,cm) self.name = "SimulStartLite" def __call__(self, dummy): '''Perform the 'SimulStartList' setup work. ''' self.incr("calls") self.debug("Setup: " + self.name) # We ignore the "node" parameter... node_list = [] for node in self.Env["nodes"]: if self.CM.ShouldBeStatus[node] == "down": self.incr("WasStopped") node_list.append(node) self.set_timer() while len(node_list) > 0: # Repeat until all nodes come up watchpats = [ ] uppat = self.templates["Pat:NonDC_started"] if self.CM.upcount() == 0: uppat = self.templates["Pat:Local_started"] watchpats.append(self.templates["Pat:DC_IDLE"]) for node in node_list: watchpats.append(uppat % node) watchpats.append(self.templates["Pat:InfraUp"] % node) watchpats.append(self.templates["Pat:PacemakerUp"] % node) # Start all the nodes - at about the same time... watch = self.create_watch(watchpats, self.Env["DeadTime"]+10) watch.setwatch() stonith = self.CM.prepare_fencing_watcher(self.name) for node in node_list: self.CM.StartaCMnoBlock(node) watch.lookforall() node_list = self.CM.fencing_cleanup(self.name, stonith) if node_list == None: return self.failure("Cluster did not stabilize") # Remove node_list messages from watch.unmatched for node in node_list: self.logger.debug("Dealing with stonith operations for %s" % repr(node_list)) if watch.unmatched: try: watch.unmatched.remove(uppat % node) except: self.debug("Already matched: %s" % (uppat % node)) try: watch.unmatched.remove(self.templates["Pat:InfraUp"] % node) except: self.debug("Already matched: %s" % (self.templates["Pat:InfraUp"] % node)) try: watch.unmatched.remove(self.templates["Pat:PacemakerUp"] % node) except: self.debug("Already matched: %s" % (self.templates["Pat:PacemakerUp"] % node)) if watch.unmatched: for regex in watch.unmatched: self.logger.log ("Warn: Startup pattern not found: %s" %(regex)) if not self.CM.cluster_stable(): return self.failure("Cluster did not stabilize") did_fail = 0 unstable = [] for node in self.Env["nodes"]: if self.CM.StataCM(node) == 0: did_fail = 1 unstable.append(node) if did_fail: return self.failure("Unstarted nodes exist: " + repr(unstable)) unstable = [] for node in self.Env["nodes"]: if not self.CM.node_stable(node): did_fail = 1 unstable.append(node) if did_fail: return self.failure("Unstable cluster nodes exist: " + repr(unstable)) return self.success() def is_applicable(self): '''SimulStartLite is a setup test and never applicable''' return False def TestList(cm, audits): result = [] for testclass in AllTestClasses: bound_test = testclass(cm) if bound_test.is_applicable(): bound_test.Audits = audits result.append(bound_test) return result class RemoteLXC(CTSTest): def __init__(self, cm): CTSTest.__init__(self,cm) self.name = "RemoteLXC" self.start = StartTest(cm) self.startall = SimulStartLite(cm) self.num_containers = 2 self.is_container = 1 self.failed = 0 self.fail_string = "" def start_lxc_simple(self, node): # restore any artifacts laying around from a previous test. self.rsh(node, "/usr/share/pacemaker/tests/cts/lxc_autogen.sh -s -R &>/dev/null") # generate the containers, put them in the config, add some resources to them pats = [ ] watch = self.create_watch(pats, 120) watch.setwatch() pats.append(self.templates["Pat:RscOpOK"] % ("start", "lxc1")) pats.append(self.templates["Pat:RscOpOK"] % ("start", "lxc2")) pats.append(self.templates["Pat:RscOpOK"] % ("start", "lxc-ms")) pats.append(self.templates["Pat:RscOpOK"] % ("promote", "lxc-ms")) self.rsh(node, "/usr/share/pacemaker/tests/cts/lxc_autogen.sh -g -a -m -s -c %d &>/dev/null" % self.num_containers) self.set_timer("remoteSimpleInit") watch.lookforall() self.log_timer("remoteSimpleInit") if watch.unmatched: self.fail_string = "Unmatched patterns: %s" % (repr(watch.unmatched)) self.failed = 1 def cleanup_lxc_simple(self, node): pats = [ ] # if the test failed, attempt to clean up the cib and libvirt environment # as best as possible if self.failed == 1: # restore libvirt and cib self.rsh(node, "/usr/share/pacemaker/tests/cts/lxc_autogen.sh -s -R &>/dev/null") return watch = self.create_watch(pats, 120) watch.setwatch() pats.append(self.templates["Pat:RscOpOK"] % ("stop", "container1")) pats.append(self.templates["Pat:RscOpOK"] % ("stop", "container2")) self.rsh(node, "/usr/share/pacemaker/tests/cts/lxc_autogen.sh -p &>/dev/null") self.set_timer("remoteSimpleCleanup") watch.lookforall() self.log_timer("remoteSimpleCleanup") if watch.unmatched: self.fail_string = "Unmatched patterns: %s" % (repr(watch.unmatched)) self.failed = 1 # cleanup libvirt self.rsh(node, "/usr/share/pacemaker/tests/cts/lxc_autogen.sh -s -R &>/dev/null") def __call__(self, node): '''Perform the 'RemoteLXC' test. ''' self.incr("calls") ret = self.startall(None) if not ret: return self.failure("Setup failed, start all nodes failed.") (rc, _) = self.rsh(node, "/usr/share/pacemaker/tests/cts/lxc_autogen.sh -v &>/dev/null") if rc == 1: self.log("Environment test for lxc support failed.") return self.skipped() self.start_lxc_simple(node) self.cleanup_lxc_simple(node) self.debug("Waiting for the cluster to recover") self.CM.cluster_stable() if self.failed == 1: return self.failure(self.fail_string) return self.success() def errorstoignore(self): '''Return list of errors which should be ignored''' return [ r"Updating failcount for ping", r"schedulerd.*: Recover\s+(ping|lxc-ms|container)\s+\(.*\)", # The orphaned lxc-ms resource causes an expected transition error # that is a result of the scheduler not having knowledge that the # promotable resource used to be a clone. As a result, it looks like that # resource is running in multiple locations when it shouldn't... But in # this instance we know why this error is occurring and that it is expected. r"Calculated [Tt]ransition .*pe-error", r"Resource lxc-ms .* is active on 2 nodes attempting recovery", r"Unknown operation: fail", r"VirtualDomain.*ERROR: Unable to determine emulator", ] AllTestClasses.append(RemoteLXC) class RemoteDriver(CTSTest): def __init__(self, cm): CTSTest.__init__(self,cm) self.name = self.__class__.__name__ self.start = StartTest(cm) self.startall = SimulStartLite(cm) self.stop = StopTest(cm) self.remote_rsc = "remote-rsc" self.cib_cmd = """cibadmin -C -o %s -X '%s' """ self.reset() def reset(self): self.pcmk_started = 0 self.failed = False self.fail_string = "" self.remote_node_added = 0 self.remote_rsc_added = 0 - self.remote_use_reconnect_interval = self.Env.RandomGen.choice([True,False]) + self.remote_use_reconnect_interval = self.Env.random_gen.choice([True,False]) def fail(self, msg): """ Mark test as failed. """ self.failed = True # Always log the failure. self.logger.log(msg) # Use first failure as test status, as it's likely to be most useful. if not self.fail_string: self.fail_string = msg def get_othernode(self, node): for othernode in self.Env["nodes"]: if othernode == node: # we don't want to try and use the cib that we just shutdown. # find a cluster node that is not our soon to be remote-node. continue else: return othernode def del_rsc(self, node, rsc): othernode = self.get_othernode(node) (rc, _) = self.rsh(othernode, "crm_resource -D -r %s -t primitive" % (rsc)) if rc != 0: self.fail("Removal of resource '%s' failed" % rsc) def add_rsc(self, node, rsc_xml): othernode = self.get_othernode(node) (rc, _) = self.rsh(othernode, self.cib_cmd % ("resources", rsc_xml)) if rc != 0: self.fail("resource creation failed") def add_primitive_rsc(self, node): rsc_xml = """ """ % { "node": self.remote_rsc } self.add_rsc(node, rsc_xml) if not self.failed: self.remote_rsc_added = 1 def add_connection_rsc(self, node): rsc_xml = """ """ % { "node": self.remote_node, "server": node } if self.remote_use_reconnect_interval: # Set reconnect interval on resource rsc_xml = rsc_xml + """ """ % (self.remote_node) rsc_xml = rsc_xml + """ """ % { "node": self.remote_node } self.add_rsc(node, rsc_xml) if not self.failed: self.remote_node_added = 1 def disable_services(self, node): self.corosync_enabled = self.Env.service_is_enabled(node, "corosync") if self.corosync_enabled: self.Env.disable_service(node, "corosync") self.pacemaker_enabled = self.Env.service_is_enabled(node, "pacemaker") if self.pacemaker_enabled: self.Env.disable_service(node, "pacemaker") def restore_services(self, node): if self.corosync_enabled: self.Env.enable_service(node, "corosync") if self.pacemaker_enabled: self.Env.enable_service(node, "pacemaker") def stop_pcmk_remote(self, node): # disable pcmk remote for i in range(10): (rc, _) = self.rsh(node, "service pacemaker_remote stop") if rc != 0: time.sleep(6) else: break def start_pcmk_remote(self, node): for i in range(10): (rc, _) = self.rsh(node, "service pacemaker_remote start") if rc != 0: time.sleep(6) else: self.pcmk_started = 1 break def freeze_pcmk_remote(self, node): """ Simulate a Pacemaker Remote daemon failure. """ # We freeze the process. self.rsh(node, "killall -STOP pacemaker-remoted") def resume_pcmk_remote(self, node): # We resume the process. self.rsh(node, "killall -CONT pacemaker-remoted") def start_metal(self, node): # Cluster nodes are reused as remote nodes in remote tests. If cluster # services were enabled at boot, in case the remote node got fenced, the # cluster node would join instead of the expected remote one. Meanwhile # pacemaker_remote would not be able to start. Depending on the chances, # the situations might not be able to be orchestrated gracefully any more. # # Temporarily disable any enabled cluster serivces. self.disable_services(node) pcmk_started = 0 # make sure the resource doesn't already exist for some reason self.rsh(node, "crm_resource -D -r %s -t primitive" % (self.remote_rsc)) self.rsh(node, "crm_resource -D -r %s -t primitive" % (self.remote_node)) if not self.stop(node): self.fail("Failed to shutdown cluster node %s" % node) return self.start_pcmk_remote(node) if self.pcmk_started == 0: self.fail("Failed to start pacemaker_remote on node %s" % node) return # Convert node to baremetal now that it has shutdown the cluster stack pats = [ ] watch = self.create_watch(pats, 120) watch.setwatch() pats.append(self.templates["Pat:RscOpOK"] % ("start", self.remote_node)) pats.append(self.templates["Pat:DC_IDLE"]) self.add_connection_rsc(node) self.set_timer("remoteMetalInit") watch.lookforall() self.log_timer("remoteMetalInit") if watch.unmatched: self.fail("Unmatched patterns: %s" % watch.unmatched) def migrate_connection(self, node): if self.failed: return pats = [ ] pats.append(self.templates["Pat:RscOpOK"] % ("migrate_to", self.remote_node)) pats.append(self.templates["Pat:RscOpOK"] % ("migrate_from", self.remote_node)) pats.append(self.templates["Pat:DC_IDLE"]) watch = self.create_watch(pats, 120) watch.setwatch() (rc, _) = self.rsh(node, "crm_resource -M -r %s" % (self.remote_node), verbose=1) if rc != 0: self.fail("failed to move remote node connection resource") return self.set_timer("remoteMetalMigrate") watch.lookforall() self.log_timer("remoteMetalMigrate") if watch.unmatched: self.fail("Unmatched patterns: %s" % watch.unmatched) return def fail_rsc(self, node): if self.failed: return watchpats = [ ] watchpats.append(self.templates["Pat:RscRemoteOpOK"] % ("stop", self.remote_rsc, self.remote_node)) watchpats.append(self.templates["Pat:RscRemoteOpOK"] % ("start", self.remote_rsc, self.remote_node)) watchpats.append(self.templates["Pat:DC_IDLE"]) watch = self.create_watch(watchpats, 120) watch.setwatch() self.debug("causing dummy rsc to fail.") self.rsh(node, "rm -f /var/run/resource-agents/Dummy*") self.set_timer("remoteRscFail") watch.lookforall() self.log_timer("remoteRscFail") if watch.unmatched: self.fail("Unmatched patterns during rsc fail: %s" % watch.unmatched) def fail_connection(self, node): if self.failed: return watchpats = [ ] watchpats.append(self.templates["Pat:Fencing_ok"] % self.remote_node) watchpats.append(self.templates["Pat:NodeFenced"] % self.remote_node) watch = self.create_watch(watchpats, 120) watch.setwatch() # freeze the pcmk remote daemon. this will result in fencing self.debug("Force stopped active remote node") self.freeze_pcmk_remote(node) self.debug("Waiting for remote node to be fenced.") self.set_timer("remoteMetalFence") watch.lookforall() self.log_timer("remoteMetalFence") if watch.unmatched: self.fail("Unmatched patterns: %s" % watch.unmatched) return self.debug("Waiting for the remote node to come back up") self.CM.ns.WaitForNodeToComeUp(node, 120); pats = [ ] watch = self.create_watch(pats, 240) watch.setwatch() pats.append(self.templates["Pat:RscOpOK"] % ("start", self.remote_node)) if self.remote_rsc_added == 1: pats.append(self.templates["Pat:RscRemoteOpOK"] % ("start", self.remote_rsc, self.remote_node)) # start the remote node again watch it integrate back into cluster. self.start_pcmk_remote(node) if self.pcmk_started == 0: self.fail("Failed to start pacemaker_remote on node %s" % node) return self.debug("Waiting for remote node to rejoin cluster after being fenced.") self.set_timer("remoteMetalRestart") watch.lookforall() self.log_timer("remoteMetalRestart") if watch.unmatched: self.fail("Unmatched patterns: %s" % watch.unmatched) return def add_dummy_rsc(self, node): if self.failed: return # verify we can put a resource on the remote node pats = [ ] watch = self.create_watch(pats, 120) watch.setwatch() pats.append(self.templates["Pat:RscRemoteOpOK"] % ("start", self.remote_rsc, self.remote_node)) pats.append(self.templates["Pat:DC_IDLE"]) # Add a resource that must live on remote-node self.add_primitive_rsc(node) # force that rsc to prefer the remote node. (rc, _) = self.CM.rsh(node, "crm_resource -M -r %s -N %s -f" % (self.remote_rsc, self.remote_node), verbose=1) if rc != 0: self.fail("Failed to place remote resource on remote node.") return self.set_timer("remoteMetalRsc") watch.lookforall() self.log_timer("remoteMetalRsc") if watch.unmatched: self.fail("Unmatched patterns: %s" % watch.unmatched) def test_attributes(self, node): if self.failed: return # This verifies permanent attributes can be set on a remote-node. It also # verifies the remote-node can edit its own cib node section remotely. (rc, line) = self.CM.rsh(node, "crm_attribute -l forever -n testattr -v testval -N %s" % (self.remote_node), verbose=1) if rc != 0: self.fail("Failed to set remote-node attribute. rc:%s output:%s" % (rc, line)) return (rc, _) = self.CM.rsh(node, "crm_attribute -l forever -n testattr -q -N %s" % (self.remote_node), verbose=1) if rc != 0: self.fail("Failed to get remote-node attribute") return (rc, _) = self.CM.rsh(node, "crm_attribute -l forever -n testattr -D -N %s" % (self.remote_node), verbose=1) if rc != 0: self.fail("Failed to delete remote-node attribute") return def cleanup_metal(self, node): self.restore_services(node) if self.pcmk_started == 0: return pats = [ ] watch = self.create_watch(pats, 120) watch.setwatch() if self.remote_rsc_added == 1: pats.append(self.templates["Pat:RscOpOK"] % ("stop", self.remote_rsc)) if self.remote_node_added == 1: pats.append(self.templates["Pat:RscOpOK"] % ("stop", self.remote_node)) self.set_timer("remoteMetalCleanup") self.resume_pcmk_remote(node) if self.remote_rsc_added == 1: # Remove dummy resource added for remote node tests self.debug("Cleaning up dummy rsc put on remote node") self.rsh(self.get_othernode(node), "crm_resource -U -r %s" % self.remote_rsc) self.del_rsc(node, self.remote_rsc) if self.remote_node_added == 1: # Remove remote node's connection resource self.debug("Cleaning up remote node connection resource") self.rsh(self.get_othernode(node), "crm_resource -U -r %s" % (self.remote_node)) self.del_rsc(node, self.remote_node) watch.lookforall() self.log_timer("remoteMetalCleanup") if watch.unmatched: self.fail("Unmatched patterns: %s" % watch.unmatched) self.stop_pcmk_remote(node) self.debug("Waiting for the cluster to recover") self.CM.cluster_stable() if self.remote_node_added == 1: # Remove remote node itself self.debug("Cleaning up node entry for remote node") self.rsh(self.get_othernode(node), "crm_node --force --remove %s" % self.remote_node) def setup_env(self, node): self.remote_node = "remote-%s" % (node) # we are assuming if all nodes have a key, that it is # the right key... If any node doesn't have a remote # key, we regenerate it everywhere. if self.rsh.exists_on_all("/etc/pacemaker/authkey", self.Env["nodes"]): return # create key locally (handle, keyfile) = tempfile.mkstemp(".cts") os.close(handle) subprocess.check_call(["dd", "if=/dev/urandom", "of=%s" % keyfile, "bs=4096", "count=1"], stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL) # sync key throughout the cluster for node in self.Env["nodes"]: self.rsh(node, "mkdir -p --mode=0750 /etc/pacemaker") self.rsh.copy(keyfile, "root@%s:/etc/pacemaker/authkey" % node) self.rsh(node, "chgrp haclient /etc/pacemaker /etc/pacemaker/authkey") self.rsh(node, "chmod 0640 /etc/pacemaker/authkey") os.unlink(keyfile) def is_applicable(self): if not self.is_applicable_common(): return False for node in self.Env["nodes"]: (rc, _) = self.rsh(node, "which pacemaker-remoted >/dev/null 2>&1") if rc != 0: return False return True def start_new_test(self, node): self.incr("calls") self.reset() ret = self.startall(None) if not ret: return self.failure("setup failed: could not start all nodes") self.setup_env(node) self.start_metal(node) self.add_dummy_rsc(node) return True def __call__(self, node): return self.failure("This base class is not meant to be called directly.") def errorstoignore(self): '''Return list of errors which should be ignored''' return [ r"""is running on remote.*which isn't allowed""", r"""Connection terminated""", r"""Could not send remote""", ] # RemoteDriver is just a base class for other tests, so it is not added to AllTestClasses class RemoteBasic(RemoteDriver): def __call__(self, node): '''Perform the 'RemoteBaremetal' test. ''' if not self.start_new_test(node): return self.failure(self.fail_string) self.test_attributes(node) self.cleanup_metal(node) self.debug("Waiting for the cluster to recover") self.CM.cluster_stable() if self.failed: return self.failure(self.fail_string) return self.success() AllTestClasses.append(RemoteBasic) class RemoteStonithd(RemoteDriver): def __call__(self, node): '''Perform the 'RemoteStonithd' test. ''' if not self.start_new_test(node): return self.failure(self.fail_string) self.fail_connection(node) self.cleanup_metal(node) self.debug("Waiting for the cluster to recover") self.CM.cluster_stable() if self.failed: return self.failure(self.fail_string) return self.success() def is_applicable(self): if not RemoteDriver.is_applicable(self): return False if "DoFencing" in list(self.Env.keys()): return self.Env["DoFencing"] return True def errorstoignore(self): ignore_pats = [ r"Lost connection to Pacemaker Remote node", r"Software caused connection abort", r"pacemaker-controld.*:\s+error.*: Operation remote-.*_monitor", r"pacemaker-controld.*:\s+error.*: Result of monitor operation for remote-.*", r"schedulerd.*:\s+Recover\s+remote-.*\s+\(.*\)", r"error: Result of monitor operation for .* on remote-.*: Internal communication failure", ] ignore_pats.extend(RemoteDriver.errorstoignore(self)) return ignore_pats AllTestClasses.append(RemoteStonithd) class RemoteMigrate(RemoteDriver): def __call__(self, node): '''Perform the 'RemoteMigrate' test. ''' if not self.start_new_test(node): return self.failure(self.fail_string) self.migrate_connection(node) self.cleanup_metal(node) self.debug("Waiting for the cluster to recover") self.CM.cluster_stable() if self.failed: return self.failure(self.fail_string) return self.success() def is_applicable(self): if not RemoteDriver.is_applicable(self): return 0 # This test requires at least three nodes: one to convert to a # remote node, one to host the connection originally, and one # to migrate the connection to. if len(self.Env["nodes"]) < 3: return 0 return 1 AllTestClasses.append(RemoteMigrate) class RemoteRscFailure(RemoteDriver): def __call__(self, node): '''Perform the 'RemoteRscFailure' test. ''' if not self.start_new_test(node): return self.failure(self.fail_string) # This is an important step. We are migrating the connection # before failing the resource. This verifies that the migration # has properly maintained control over the remote-node. self.migrate_connection(node) self.fail_rsc(node) self.cleanup_metal(node) self.debug("Waiting for the cluster to recover") self.CM.cluster_stable() if self.failed: return self.failure(self.fail_string) return self.success() def errorstoignore(self): ignore_pats = [ r"schedulerd.*: Recover\s+remote-rsc\s+\(.*\)", r"Dummy.*: No process state file found", ] ignore_pats.extend(RemoteDriver.errorstoignore(self)) return ignore_pats def is_applicable(self): if not RemoteDriver.is_applicable(self): return 0 # This test requires at least three nodes: one to convert to a # remote node, one to host the connection originally, and one # to migrate the connection to. if len(self.Env["nodes"]) < 3: return 0 return 1 AllTestClasses.append(RemoteRscFailure) # vim:ts=4:sw=4:et: diff --git a/python/pacemaker/_cts/environment.py b/python/pacemaker/_cts/environment.py index f301273c71..4cae274971 100644 --- a/python/pacemaker/_cts/environment.py +++ b/python/pacemaker/_cts/environment.py @@ -1,657 +1,657 @@ """ Test environment classes for Pacemaker's Cluster Test Suite (CTS) """ __all__ = ["EnvFactory"] __copyright__ = "Copyright 2014-2023 the Pacemaker project contributors" __license__ = "GNU General Public License version 2 or later (GPLv2+) WITHOUT ANY WARRANTY" import os import random import socket import sys import time from pacemaker._cts.logging import LogFactory from pacemaker._cts.remote import RemoteFactory class Environment: def __init__(self, args): self.data = {} - self.Nodes = [] + self._nodes = [] self["DeadTime"] = 300 self["StartTime"] = 300 self["StableTime"] = 30 self["tests"] = [] self["IPagent"] = "IPaddr2" self["DoFencing"] = True self["XmitLoss"] = "0.0" self["RecvLoss"] = "0.0" self["ClobberCIB"] = False self["CIBfilename"] = None self["CIBResource"] = False self["DoBSC"] = False self["oprofile"] = [] self["warn-inactive"] = False self["ListTests"] = False self["benchmark"] = False self["LogWatcher"] = "any" self["SyslogFacility"] = "daemon" self["LogFileName"] = "/var/log/messages" self["Schema"] = "pacemaker-3.0" self["Stack"] = "corosync" self["stonith-type"] = "external/ssh" self["stonith-params"] = "hostlist=all,livedangerously=yes" self["notification-agent"] = "/var/lib/pacemaker/notify.sh" self["notification-recipient"] = "/var/lib/pacemaker/notify.log" self["loop-minutes"] = 60 self["valgrind-procs"] = "pacemaker-attrd pacemaker-based pacemaker-controld pacemaker-execd pacemaker-fenced pacemaker-schedulerd" self["experimental-tests"] = False self["container-tests"] = False self["valgrind-tests"] = False self["unsafe-tests"] = True self["loop-tests"] = True self["scenario"] = "random" self["stats"] = False self["continue"] = False self["TruncateLog"] = False - self.RandomGen = random.Random() + self.random_gen = random.Random() self.logger = LogFactory() - self.SeedRandom() + self._seed_random() self.rsh = RemoteFactory().getInstance() self.target = "localhost" self.parse_args(args) if not self["ListTests"]: self.validate() self.discover() - def SeedRandom(self, seed=None): + def _seed_random(self, seed=None): if not seed: seed = int(time.time()) self["RandSeed"] = seed - self.RandomGen.seed(str(seed)) + self.random_gen.seed(str(seed)) def dump(self): keys = [] for key in list(self.data.keys()): keys.append(key) keys.sort() for key in keys: self.logger.debug("Environment["+key+"]:\t"+str(self[key])) def keys(self): return list(self.data.keys()) def has_key(self, key): if key == "nodes": return True return key in self.data def __getitem__(self, key): if str(key) == "0": raise ValueError("Bad call to 'foo in X', should reference 'foo in X.keys()' instead") if key == "nodes": - return self.Nodes + return self._nodes if key == "Name": return self.get_stack_short() if key in self.data: return self.data[key] return None def __setitem__(self, key, value): if key == "Stack": self.set_stack(value) elif key == "node-limit": self.data[key] = value self.filter_nodes() elif key == "nodes": - self.Nodes = [] + self._nodes = [] for node in value: # I don't think I need the IP address, etc. but this validates # the node name against /etc/hosts and/or DNS, so it's a # GoodThing(tm). try: n = node.strip() socket.gethostbyname_ex(n) - self.Nodes.append(n) + self._nodes.append(n) except: self.logger.log(node+" not found in DNS... aborting") raise self.filter_nodes() else: self.data[key] = value - def RandomNode(self): - '''Choose a random node from the cluster''' - return self.RandomGen.choice(self["nodes"]) + def random_node(self): + """ Choose a random node from the cluster """ + return self.random_gen.choice(self["nodes"]) def set_stack(self, name): # Normalize stack names if name in ["corosync", "cs", "mcp"]: self.data["Stack"] = "corosync 2+" else: raise ValueError("Unknown stack: "+name) def get_stack_short(self): # Create the Cluster Manager object if "Stack" not in self.data: return "unknown" if self.data["Stack"] == "corosync 2+": return "crm-corosync" LogFactory().log("Unknown stack: "+self["stack"]) raise ValueError("Unknown stack: "+self["stack"]) def detect_syslog(self): # Detect syslog variant if "syslogd" not in self.data: if self["have_systemd"]: # Systemd (_, lines) = self.rsh(self.target, "systemctl list-units | grep syslog.*\.service.*active.*running | sed 's:.service.*::'", verbose=1) self["syslogd"] = lines[0].strip() else: # SYS-V (_, lines) = self.rsh(self.target, "chkconfig --list | grep syslog.*on | awk '{print $1}' | head -n 1", verbose=1) self["syslogd"] = lines[0].strip() if "syslogd" not in self.data or not self["syslogd"]: # default self["syslogd"] = "rsyslog" def disable_service(self, node, service): if self["have_systemd"]: # Systemd (rc, _) = self.rsh(node, "systemctl disable %s" % service) return rc # SYS-V (rc, _) = self.rsh(node, "chkconfig %s off" % service) return rc def enable_service(self, node, service): if self["have_systemd"]: # Systemd (rc, _) = self.rsh(node, "systemctl enable %s" % service) return rc # SYS-V (rc, _) = self.rsh(node, "chkconfig %s on" % service) return rc def service_is_enabled(self, node, service): if self["have_systemd"]: # Systemd # With "systemctl is-enabled", we should check if the service is # explicitly "enabled" instead of the return code. For example it returns # 0 if the service is "static" or "indirect", but they don't really count # as "enabled". (rc, _) = self.rsh(node, "systemctl is-enabled %s | grep enabled" % service) return rc == 0 # SYS-V (rc, _) = self.rsh(node, "chkconfig --list | grep -e %s.*on" % service) return rc == 0 def detect_at_boot(self): # Detect if the cluster starts at boot if "at-boot" not in self.data: self["at-boot"] = self.service_is_enabled(self.target, "corosync") \ or self.service_is_enabled(self.target, "pacemaker") def detect_ip_offset(self): # Try to determine an offset for IPaddr resources if self["CIBResource"] and "IPBase" not in self.data: (_, lines) = self.rsh(self.target, "ip addr | grep inet | grep -v -e link -e inet6 -e '/32' -e ' lo' | awk '{print $2}'", verbose=0) network = lines[0].strip() (_, lines) = self.rsh(self.target, "nmap -sn -n %s | grep 'scan report' | awk '{print $NF}' | sed 's:(::' | sed 's:)::' | sort -V | tail -n 1" % network, verbose=0) self["IPBase"] = lines[0].strip() if not self["IPBase"]: self["IPBase"] = " fe80::1234:56:7890:1000" self.logger.log("Could not determine an offset for IPaddr resources. Perhaps nmap is not installed on the nodes.") self.logger.log("Defaulting to '%s', use --test-ip-base to override" % self["IPBase"]) elif int(self["IPBase"].split('.')[3]) >= 240: self.logger.log("Could not determine an offset for IPaddr resources. Upper bound is too high: %s %s" % (self["IPBase"], self["IPBase"].split('.')[3])) self["IPBase"] = " fe80::1234:56:7890:1000" self.logger.log("Defaulting to '%s', use --test-ip-base to override" % self["IPBase"]) def filter_nodes(self): if self['node-limit'] is not None and self["node-limit"] > 0: if len(self["nodes"]) > self["node-limit"]: self.logger.log("Limiting the number of nodes configured=%d (max=%d)" %(len(self["nodes"]), self["node-limit"])) while len(self["nodes"]) > self["node-limit"]: self["nodes"].pop(len(self["nodes"])-1) def validate(self): if not self["nodes"]: print("No nodes specified!") sys.exit(1) def discover(self): self.target = random.Random().choice(self["nodes"]) exerciser = socket.gethostname() # Use the IP where possible to avoid name lookup failures for ip in socket.gethostbyname_ex(exerciser)[2]: if ip != "127.0.0.1": exerciser = ip break self["cts-exerciser"] = exerciser if "have_systemd" not in self.data: (rc, _) = self.rsh(self.target, "systemctl list-units", verbose=0) self["have_systemd"] = rc == 0 self.detect_syslog() self.detect_at_boot() self.detect_ip_offset() def parse_args(self, args): skipthis=None if not args: args=sys.argv[1:] for i in range(0, len(args)): if skipthis: skipthis=None continue elif args[i] == "-l" or args[i] == "--limit-nodes": skipthis=1 self["node-limit"] = int(args[i+1]) elif args[i] == "-r" or args[i] == "--populate-resources": self["CIBResource"] = True self["ClobberCIB"] = True elif args[i] == "--outputfile": skipthis=1 self["OutputFile"] = args[i+1] LogFactory().add_file(self["OutputFile"]) elif args[i] == "-L" or args[i] == "--logfile": skipthis=1 self["LogWatcher"] = "remote" self["LogAuditDisabled"] = 1 self["LogFileName"] = args[i+1] elif args[i] == "--ip" or args[i] == "--test-ip-base": skipthis=1 self["IPBase"] = args[i+1] self["CIBResource"] = True self["ClobberCIB"] = True elif args[i] == "--oprofile": skipthis=1 self["oprofile"] = args[i+1].split(' ') elif args[i] == "--trunc": self["TruncateLog"] = True elif args[i] == "--list-tests" or args[i] == "--list" : self["ListTests"] = True elif args[i] == "--benchmark": self["benchmark"] = True elif args[i] == "--bsc": self["DoBSC"] = True self["scenario"] = "basic-sanity" elif args[i] == "--qarsh": RemoteFactory().enable_qarsh() elif args[i] == "--yes" or args[i] == "-y": self["continue"] = True elif args[i] == "--stonith" or args[i] == "--fencing": skipthis=1 if args[i+1] == "1" or args[i+1] == "yes": self["DoFencing"] = True elif args[i+1] == "0" or args[i+1] == "no": self["DoFencing"] = False elif args[i+1] == "rhcs" or args[i+1] == "xvm" or args[i+1] == "virt": self["DoStonith"]=1 self["stonith-type"] = "fence_xvm" elif args[i+1] == "scsi": self["DoStonith"]=1 self["stonith-type"] = "fence_scsi" elif args[i+1] == "ssh" or args[i+1] == "lha": self["DoStonith"]=1 self["stonith-type"] = "external/ssh" self["stonith-params"] = "hostlist=all,livedangerously=yes" elif args[i+1] == "north": self["DoStonith"]=1 self["stonith-type"] = "fence_apc" self["stonith-params"] = "ipaddr=north-apc,login=apc,passwd=apc,pcmk_host_map=north-01:2;north-02:3;north-03:4;north-04:5;north-05:6;north-06:7;north-07:9;north-08:10;north-09:11;north-10:12;north-11:13;north-12:14;north-13:15;north-14:18;north-15:17;north-16:19;" elif args[i+1] == "south": self["DoStonith"]=1 self["stonith-type"] = "fence_apc" self["stonith-params"] = "ipaddr=south-apc,login=apc,passwd=apc,pcmk_host_map=south-01:2;south-02:3;south-03:4;south-04:5;south-05:6;south-06:7;south-07:9;south-08:10;south-09:11;south-10:12;south-11:13;south-12:14;south-13:15;south-14:18;south-15:17;south-16:19;" elif args[i+1] == "east": self["DoStonith"]=1 self["stonith-type"] = "fence_apc" self["stonith-params"] = "ipaddr=east-apc,login=apc,passwd=apc,pcmk_host_map=east-01:2;east-02:3;east-03:4;east-04:5;east-05:6;east-06:7;east-07:9;east-08:10;east-09:11;east-10:12;east-11:13;east-12:14;east-13:15;east-14:18;east-15:17;east-16:19;" elif args[i+1] == "west": self["DoStonith"]=1 self["stonith-type"] = "fence_apc" self["stonith-params"] = "ipaddr=west-apc,login=apc,passwd=apc,pcmk_host_map=west-01:2;west-02:3;west-03:4;west-04:5;west-05:6;west-06:7;west-07:9;west-08:10;west-09:11;west-10:12;west-11:13;west-12:14;west-13:15;west-14:18;west-15:17;west-16:19;" elif args[i+1] == "openstack": self["DoStonith"]=1 self["stonith-type"] = "fence_openstack" print("Obtaining OpenStack credentials from the current environment") self["stonith-params"] = "region=%s,tenant=%s,auth=%s,user=%s,password=%s" % ( os.environ['OS_REGION_NAME'], os.environ['OS_TENANT_NAME'], os.environ['OS_AUTH_URL'], os.environ['OS_USERNAME'], os.environ['OS_PASSWORD'] ) elif args[i+1] == "rhevm": self["DoStonith"]=1 self["stonith-type"] = "fence_rhevm" print("Obtaining RHEV-M credentials from the current environment") self["stonith-params"] = "login=%s,passwd=%s,ipaddr=%s,ipport=%s,ssl=1,shell_timeout=10" % ( os.environ['RHEVM_USERNAME'], os.environ['RHEVM_PASSWORD'], os.environ['RHEVM_SERVER'], os.environ['RHEVM_PORT'], ) else: self.usage(args[i+1]) elif args[i] == "--stonith-type": self["stonith-type"] = args[i+1] skipthis=1 elif args[i] == "--stonith-args": self["stonith-params"] = args[i+1] skipthis=1 elif args[i] == "--clobber-cib" or args[i] == "-c": self["ClobberCIB"] = True elif args[i] == "--cib-filename": skipthis=1 self["CIBfilename"] = args[i+1] elif args[i] == "--xmit-loss": try: float(args[i+1]) except ValueError: print("--xmit-loss parameter should be float") self.usage(args[i+1]) skipthis=1 self["XmitLoss"] = args[i+1] elif args[i] == "--recv-loss": try: float(args[i+1]) except ValueError: print("--recv-loss parameter should be float") self.usage(args[i+1]) skipthis=1 self["RecvLoss"] = args[i+1] elif args[i] == "--choose": skipthis=1 self["tests"].append(args[i+1]) self["scenario"] = "sequence" elif args[i] == "--nodes": skipthis=1 self["nodes"] = args[i+1].split(' ') elif args[i] == "-g" or args[i] == "--group" or args[i] == "--dsh-group": skipthis=1 self["OutputFile"] = "%s/cluster-%s.log" % (os.environ['HOME'], args[i+1]) LogFactory().add_file(self["OutputFile"], "CTS") dsh_file = "%s/.dsh/group/%s" % (os.environ['HOME'], args[i+1]) # Hacks to make my life easier if args[i+1] == "virt1": self["Stack"] = "corosync" self["DoStonith"]=1 self["stonith-type"] = "fence_xvm" self["stonith-params"] = "delay=0" self["IPBase"] = " fe80::1234:56:7890:1000" elif args[i+1] == "east16" or args[i+1] == "nsew": self["Stack"] = "corosync" self["DoStonith"]=1 self["stonith-type"] = "fence_apc" self["stonith-params"] = "ipaddr=east-apc,login=apc,passwd=apc,pcmk_host_map=east-01:2;east-02:3;east-03:4;east-04:5;east-05:6;east-06:7;east-07:9;east-08:10;east-09:11;east-10:12;east-11:13;east-12:14;east-13:15;east-14:18;east-15:17;east-16:19;" self["IPBase"] = " fe80::1234:56:7890:2000" if args[i+1] == "east16": # Requires newer python than available via nsew self["IPagent"] = "Dummy" elif args[i+1] == "corosync8": self["Stack"] = "corosync" self["DoStonith"]=1 self["stonith-type"] = "fence_rhevm" print("Obtaining RHEV-M credentials from the current environment") self["stonith-params"] = "login=%s,passwd=%s,ipaddr=%s,ipport=%s,ssl=1,shell_timeout=10" % ( os.environ['RHEVM_USERNAME'], os.environ['RHEVM_PASSWORD'], os.environ['RHEVM_SERVER'], os.environ['RHEVM_PORT']) self["IPBase"] = " fe80::1234:56:7890:3000" if os.path.isfile(dsh_file): self["nodes"] = [] f = open(dsh_file, 'r') for line in f: l = line.strip().rstrip() if not l.startswith('#'): self["nodes"].append(l) f.close() else: print("Unknown DSH group: %s" % args[i+1]) elif args[i] == "--syslog-facility" or args[i] == "--facility": skipthis=1 self["SyslogFacility"] = args[i+1] elif args[i] == "--seed": skipthis=1 - self.SeedRandom(args[i+1]) + self._seed_random(args[i+1]) elif args[i] == "--warn-inactive": self["warn-inactive"] = True elif args[i] == "--schema": skipthis=1 self["Schema"] = args[i+1] elif args[i] == "--at-boot" or args[i] == "--cluster-starts-at-boot": skipthis=1 if args[i+1] == "1" or args[i+1] == "yes": self["at-boot"] = 1 elif args[i+1] == "0" or args[i+1] == "no": self["at-boot"] = 0 else: self.usage(args[i+1]) elif args[i] == "--stack": if args[i+1] == "fedora" or args[i+1] == "fedora-17" or args[i+1] == "fedora-18": self["Stack"] = "corosync" elif args[i+1] == "rhel-7": self["Stack"] = "corosync" else: self["Stack"] = args[i+1] skipthis=1 elif args[i] == "--once": self["scenario"] = "all-once" elif args[i] == "--boot": self["scenario"] = "boot" elif args[i] == "--notification-agent": self["notification-agent"] = args[i+1] skipthis = 1 elif args[i] == "--notification-recipient": self["notification-recipient"] = args[i+1] skipthis = 1 elif args[i] == "--valgrind-tests": self["valgrind-tests"] = True elif args[i] == "--valgrind-procs": self["valgrind-procs"] = args[i+1] skipthis = 1 elif args[i] == "--no-loop-tests": self["loop-tests"] = False elif args[i] == "--loop-minutes": skipthis=1 try: self["loop-minutes"]=int(args[i+1]) except ValueError: self.usage(args[i]) elif args[i] == "--no-unsafe-tests": self["unsafe-tests"] = False elif args[i] == "--experimental-tests": self["experimental-tests"] = True elif args[i] == "--container-tests": self["container-tests"] = True elif args[i] == "--set": skipthis=1 (name, value) = args[i+1].split('=') self[name] = value print("Setting %s = %s" % (name, value)) elif args[i] == "--help": self.usage(args[i], 0) elif args[i] == "--": break else: try: NumIter=int(args[i]) self["iterations"] = NumIter except ValueError: self.usage(args[i]) def usage(self, arg, status=1): if status: print("Illegal argument %s" % arg) print("""usage: %s [options] number-of-iterations Common options: \t [--nodes 'node list'] list of cluster nodes separated by whitespace \t [--group | -g 'name'] use the nodes listed in the named DSH group (~/.dsh/groups/$name) \t [--limit-nodes max] only use the first 'max' cluster nodes supplied with --nodes \t [--stack corosync] which cluster stack is installed \t [--list-tests] list the valid tests \t [--benchmark] add the timing information Options that CTS will usually auto-detect correctly: \t [--logfile path] where should the test software look for logs from cluster nodes \t [--syslog-facility name] which syslog facility should the test software log to \t [--at-boot (1|0)] does the cluster software start at boot time \t [--test-ip-base ip] offset for generated IP address resources Options for release testing: \t [--populate-resources | -r] generate a sample configuration \t [--choose name] run only the named test \t [--stonith (1 | 0 | yes | no | rhcs | ssh)] \t [--once] run all valid tests once Additional (less common) options: \t [--clobber-cib | -c ] erase any existing configuration \t [--outputfile path] optional location for the test software to write logs to \t [--trunc] truncate logfile before starting \t [--xmit-loss lost-rate(0.0-1.0)] \t [--recv-loss lost-rate(0.0-1.0)] \t [--fencing (1 | 0 | yes | no | rhcs | lha | openstack )] \t [--stonith-type type] \t [--stonith-args name=value] \t [--bsc] \t [--notification-agent path] script to configure for Pacemaker alerts \t [--notification-recipient r] recipient to pass to alert script \t [--no-loop-tests] don't run looping/time-based tests \t [--no-unsafe-tests] don't run tests that are unsafe for use with ocfs2/drbd \t [--valgrind-tests] include tests using valgrind \t [--experimental-tests] include experimental tests \t [--container-tests] include pacemaker_remote tests that run in lxc container resources \t [--oprofile 'node list'] list of cluster nodes to run oprofile on] \t [--qarsh] use the QARSH backdoor to access nodes instead of SSH \t [--seed random_seed] \t [--set option=value] \t [--yes | -y] continue to run cts when there is an interaction whether to continue running pacemaker-cts Example: \t python %s -g virt1 -r --stonith ssh --schema pacemaker-2.0 500""" % (sys.argv[0], sys.argv[0])) sys.exit(status) class EnvFactory: instance = None def getInstance(self, args=None): if not EnvFactory.instance: EnvFactory.instance = Environment(args) return EnvFactory.instance