diff --git a/cts/CTSaudits.py.in b/cts/CTSaudits.py.in index 20af9feaff..0f5aaf58aa 100755 --- a/cts/CTSaudits.py.in +++ b/cts/CTSaudits.py.in @@ -1,684 +1,697 @@ #!@PYTHON@ '''CTS: Cluster Testing System: Audit module ''' __copyright__=''' Copyright (C) 2000, 2001,2005 Alan Robertson Licensed under the GNU GPL. ''' # # This program is free software; you can redistribute it and/or # modify it under the terms of the GNU General Public License # as published by the Free Software Foundation; either version 2 # of the License, or (at your option) any later version. # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # You should have received a copy of the GNU General Public License # along with this program; if not, write to the Free Software # Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. import time, os, popen2, string, re import CTS import os import popen2 class ClusterAudit: def __init__(self, cm): self.CM = cm def __call__(self): raise ValueError("Abstract Class member (__call__)") def is_applicable(self): '''Return TRUE if we are applicable in the current test configuration''' raise ValueError("Abstract Class member (is_applicable)") return 1 def name(self): raise ValueError("Abstract Class member (name)") AllAuditClasses = [ ] class ResourceAudit(ClusterAudit): def name(self): return "ResourceAudit" def _doauditRsc(self, resource): ResourceNodes = [] for node in self.CM.Env["nodes"]: if self.CM.ShouldBeStatus[node] == self.CM["up"]: if resource.IsRunningOn(node): ResourceNodes.append(node) return ResourceNodes def _doaudit(self): '''Check to see if all resources are running in exactly one place in the cluster. We also verify that the members of a resource group are all running on the same node in the cluster, and we monitor that they are all running "properly". ''' Fatal = 0 result = [] # Thought: use self.CM.find_partitions() and make this audit # aware of partitions. Since in a split cluster one # partition may have quorum (and permission to run resources) # and the other not. Groups = self.CM.ResourceGroups() for group in Groups: GrpServedBy = None lastResource = None for resource in group: # # _doauditRsc returns the set of nodes serving # the given resource. This is normally a single node. # ResourceNodes = self._doauditRsc(resource) # Is the resource served without quorum present? if not self.CM.HasQuorum(None) and len(ResourceNodes) != 0 and resource.needs_quorum: result.append("Resource " + repr(resource) + " active without Quorum: " + repr(ResourceNodes)) # Is the resource served at all? elif len(ResourceNodes) == 0 and self.CM.HasQuorum(None): result.append("Resource " + repr(resource) + " not served anywhere.") # Is the resource served too many times? elif len(ResourceNodes) > 1: result.append("Resource " + repr(resource) + " served too many times: " + repr(ResourceNodes)) self.CM.log("Resource " + repr(resource) + " served too many times: " + repr(ResourceNodes)) Fatal = 1 elif GrpServedBy == None: GrpServedBy = ResourceNodes # Are all the members of the Rsc Grp served by the same node? elif GrpServedBy != ResourceNodes: result.append("Resource group resources" + repr(resource) + " running on different nodes: " + repr(ResourceNodes)+" vs "+repr(GrpServedBy) + "(otherRsc = " + repr(lastResource) + ")") self.CM.log("Resource group resources" + repr(resource) + " running on different nodes: " + repr(ResourceNodes)+" vs "+repr(GrpServedBy) + "(otherRsc = " + repr(lastResource) + ")") Fatal = 1 if self.CM.Env.has_key("SuppressMonitoring") and \ self.CM.Env["SuppressMonitoring"]: continue # Is the resource working correctly ? if not Fatal and len(ResourceNodes) == 1: beforearpchild = popen2.Popen3("date;/sbin/arp -n|cut -c1-15,26-50,75-" , None) beforearpchild.tochild.close() # /dev/null if not resource.IsWorkingCorrectly(ResourceNodes[0]): afterarpchild = popen2.Popen3("/sbin/arp -n|cut -c1-15,26-50,75-" , None) afterarpchild.tochild.close() # /dev/null result.append("Resource " + repr(resource) + " not operating properly." + " Resource is running on " + ResourceNodes[0]); Fatal = 1 self.CM.log("ARP table before failure ========"); for line in beforearpchild.fromchild.readlines(): self.CM.log(line) self.CM.log("ARP table after failure ========"); for line in afterarpchild.fromchild.readlines(): self.CM.log(line) self.CM.log("End of ARP tables ========"); try: beforearpchild.wait() afterarpchild.wait() except OSError: pass afterarpchild.fromchild.close() beforearpchild.fromchild.close() lastResource = resource if (Fatal): result.insert(0, "FATAL") # Kludgy. return result def __call__(self): # # Audit the resources. Since heartbeat doesn't really # know when resource acquisition is complete, we will # poll until things get stable. # # Having a resource duplicately implemented is a Fatal Error # with no tolerance granted. # audresult = self._doaudit() # # Probably the constant below should be a CM parameter. # Then it could be 0 for FailSafe. # Of course, it really depends on what resources # you have in the test suite, and how long it takes # for them to settle. # Recently, we've changed heartbeat so we know better when # resource acquisition is done. # audcount=5; while(audcount > 0): audresult = self._doaudit() if (len(audresult) <= 0 or audresult[0] == "FATAL"): audcount=0 else: audcount = audcount - 1 if (audcount > 0): time.sleep(1) if (len(audresult) > 0): self.CM.log("Fatal Audit error: " + repr(audresult)) return (len(audresult) == 0) def is_applicable(self): if self.CM["Name"] == "heartbeat": return 1 return 0 class HAResourceAudit(ClusterAudit): def __init__(self, cm): self.CM = cm def _RscRunningNodes(self, resource): ResourceNodes = [] for node in self.CM.Env["nodes"]: if self.CM.ShouldBeStatus[node] == self.CM["up"]: if resource.IsRunningOn(node): ResourceNodes.append(node) return ResourceNodes def __call__(self): passed = 1 NodeofRsc = {} NumofInc = {} MaxofInc = {} self.CM.debug("Do Audit HAResourceAudit") #Calculate the count of active nodes up_count = 0; for node in self.CM.Env["nodes"]: if self.CM.ShouldBeStatus[node] == self.CM["up"]: up_count += 1 #Make sure the resouces are running on one and only one node Resources = self.CM.Resources() for resource in Resources : RunningNodes = self._RscRunningNodes(resource) NodeofRsc[resource.rid]=RunningNodes if resource.inc_name == None: #Is the resource served without quorum present? if not self.CM.HasQuorum(None) and len(RunningNodes) != 0 and resource.needs_quorum: self.CM.log("Resource " + repr(resource) + " active without Quorum: " + repr(RunningNodes)) passed = 0 #Is the resource served at all? elif len(RunningNodes) == 0 and self.CM.HasQuorum(None): self.CM.log("Resource " + repr(resource) + " not served anywhere.") passed = 0 # Is the resource served too many times? elif len(RunningNodes) > 1: self.CM.log("Resource " + repr(resource) + " served too many times: " + repr(RunningNodes)) passed = 0 else: if not NumofInc.has_key(resource.inc_name): NumofInc[resource.inc_name]=0 MaxofInc[resource.inc_name]=resource.inc_max running = 1 #Is the resource served without quorum present? if not self.CM.HasQuorum(None) and len(RunningNodes) != 0 and resource.needs_quorum == 1: self.CM.log("Resource " + repr(resource) + " active without Quorum: " + repr(RunningNodes)) passed = 0 #Is the resource served at all? elif len(RunningNodes) == 0 : running = 0 # Is the resource served too many times? elif len(RunningNodes) > 1: self.CM.log("Resource " + repr(resource) + " served too many times: " + repr(RunningNodes)) passed = 0 if running: NumofInc[resource.inc_name] += 1 + if self.CM.HasQuorum(None): for inc_name in NumofInc.keys(): - if NumofInc[inc_name] != min(up_count,MaxofInc[inc_name]): + if NumofInc[inc_name] != min(up_count, MaxofInc[inc_name]): passed = 0 - self.CM.log("Incarnation %s has %d instances(max %d instances).\ - Now %d nodes are up"%(str(inc_name),NumofInc[inc_name], \ - MaxofInc[inc_name],up_count)) - + self.CM.log("Cloned resource "+ inc_name + +" has "+ NumofInc[inc_name] + +" active instances (max: "+ MaxofInc[inc_name] + +", active nodes: "+ up_count + ")") + Groups = self.CM.ResourceGroups() for group in Groups : - RunningNodes = NodeofRsc[group[0].rid] + group_printed = 0 + first_rsc = group[0].rid + RunningNodes = NodeofRsc[first_rsc] for rsc in group : - if RunningNodes != NodeofRsc[rsc.rid] : + if RunningNodes != NodeofRsc[rsc.rid]: passed = 0 + + if group_printed == 0: + group_printed = 1 + self.CM.log("Group audit failed for: %s" % repr(group)) + if not NodeofRsc[first_rsc] or len(NodeofRsc[first_rsc]) == 0: + self.CM.log("* %s not running" % first_rsc) + else: + self.CM.log("* %s running on %s" + %(first_rsc, repr(NodeofRsc[first_rsc]))) + if not NodeofRsc[rsc.rid] or len(NodeofRsc[rsc.rid]) == 0: - self.CM.log("Not all resources in group %s running on same node" % repr(group)) self.CM.log("* %s not running" % rsc.rid) else: - self.CM.log("Not all resources in group %s running on same node"% repr(group)) - self.CM.log("* %s running on %s"%(group[0].rid, repr(NodeofRsc[group[0].rid]))) - self.CM.log("* %s running on %s"%(rsc.rid, repr(NodeofRsc[rsc.rid]))) + self.CM.log("* %s running on %s" + %(rsc.rid, repr(NodeofRsc[rsc.rid]))) - #Make sure the resouces with "must","placement" constraint are running on the same node + # Make sure the resouces with "must","placement" constraint + # are running on the same node Dependancies = self.CM.Dependencies() for dependency in Dependancies: if dependency["type"] == "placement" and dependency["strength"] == "must": if NodeofRsc[dependency["from"]] != NodeofRsc[dependency["to"]]: print dependency["from"] + " and " + dependency["to"] + " should be run on same node" passed = 0 return passed def is_applicable(self): if self.CM["Name"] == "linux-ha-v2" and self.CM.Env["ResCanStop"] == 0: return 1 return 0 def name(self): return "HAResourceAudit" class CrmdStateAudit(ClusterAudit): def __init__(self, cm): self.CM = cm self.Stats = {"calls":0 , "success":0 , "failure":0 , "skipped":0 , "auditfail":0} def has_key(self, key): return self.Stats.has_key(key) def __setitem__(self, key, value): self.Stats[key] = value def __getitem__(self, key): return self.Stats[key] def incr(self, name): '''Increment (or initialize) the value associated with the given name''' if not self.Stats.has_key(name): self.Stats[name]=0 self.Stats[name] = self.Stats[name]+1 def __call__(self): passed = 1 up_are_down = 0 down_are_up = 0 unstable_list = [] self.CM.debug("Do Audit %s"%self.name()) for node in self.CM.Env["nodes"]: should_be = self.CM.ShouldBeStatus[node] rc = self.CM.StataCM(node) if rc: if should_be == self.CM["down"]: down_are_up = down_are_up + 1 if not self.CM.node_stable(node): unstable_list.append(node) elif should_be == self.CM["up"]: up_are_down = up_are_down + 1 if len(unstable_list) > 0: passed = 0 self.CM.log("Cluster is not stable: %d (of %d): %s" %(len(unstable_list), self.CM.upcount(), repr(unstable_list))) if up_are_down > 0: passed = 0 self.CM.log("%d (of %d) nodes expected to be up were down." %(up_are_down, len(self.CM.Env["nodes"]))) if down_are_up > 0: passed = 0 self.CM.log("%d (of %d) nodes expected to be down were up." %(down_are_up, len(self.CM.Env["nodes"]))) return passed def name(self): return "CrmdStateAudit" def is_applicable(self): if self.CM["Name"] == "linux-ha-v2": return 1 return 0 class CIBAudit(ClusterAudit): def __init__(self, cm): self.CM = cm self.Stats = {"calls":0 , "success":0 , "failure":0 , "skipped":0 , "auditfail":0} def has_key(self, key): return self.Stats.has_key(key) def __setitem__(self, key, value): self.Stats[key] = value def __getitem__(self, key): return self.Stats[key] def incr(self, name): '''Increment (or initialize) the value associated with the given name''' if not self.Stats.has_key(name): self.Stats[name]=0 self.Stats[name] = self.Stats[name]+1 def __call__(self): self.CM.debug("Do Audit %s"%self.name()) passed = 1 ccm_partitions = self.CM.find_partitions() if len(ccm_partitions) == 0: self.CM.debug("\tNo partitions to audit") return 1 for partition in ccm_partitions: self.CM.debug("\tAuditing CIB consistency for: %s" %partition) partition_passed = 0 if self.audit_cib_contents(partition) == 0: passed = 0 return passed def audit_cib_contents(self, hostlist): passed = 1 first_host = None first_host_xml = "" partition_hosts = hostlist.split() for a_host in partition_hosts: if first_host == None: first_host = a_host first_host_xml = self.store_remote_cib(a_host) #self.CM.debug("Retrieved CIB: %s" % first_host_xml) else: a_host_xml = self.store_remote_cib(a_host) diff_cmd="@sbindir@/crm_diff -c -VV -f -N \'%s\' -O '%s'" % (a_host_xml, first_host_xml) infile, outfile, errfile = os.popen3(diff_cmd) diff_lines = outfile.readlines() for line in diff_lines: if not re.search("", line): passed = 0 self.CM.log("CibDiff[%s-%s]: %s" % (first_host, a_host, line)) else: self.CM.debug("CibDiff[%s-%s] Ignoring: %s" % (first_host, a_host, line)) diff_lines = errfile.readlines() for line in diff_lines: passed = 0 self.CM.log("CibDiff[%s-%s] ERROR: %s" % (first_host, a_host, line)) return passed def store_remote_cib(self, node): combined = "" first_line = 1 extra_debug = 0 #self.CM.debug("\tRetrieving CIB from: %s" % node) lines = self.CM.rsh.readlines(node, self.CM["CibQuery"]) if extra_debug: self.CM.debug("Start Cib[%s]" % node) for line in lines: combined = combined + line[:-1] if first_line: self.CM.debug("[Cib]" + line) first_line = 0 elif extra_debug: self.CM.debug("[Cib]" + line) if extra_debug: self.CM.debug("End Cib[%s]" % node) #self.CM.debug("Complete CIB: %s" % combined) return combined def name(self): return "CibAudit" def is_applicable(self): if self.CM["Name"] == "linux-ha-v2": return 1 return 0 class PartitionAudit(ClusterAudit): def __init__(self, cm): self.CM = cm self.Stats = {"calls":0 , "success":0 , "failure":0 , "skipped":0 , "auditfail":0} self.NodeEpoche={} self.NodeState={} self.NodeQuorum={} def has_key(self, key): return self.Stats.has_key(key) def __setitem__(self, key, value): self.Stats[key] = value def __getitem__(self, key): return self.Stats[key] def incr(self, name): '''Increment (or initialize) the value associated with the given name''' if not self.Stats.has_key(name): self.Stats[name]=0 self.Stats[name] = self.Stats[name]+1 def __call__(self): self.CM.debug("Do Audit %s"%self.name()) passed = 1 ccm_partitions = self.CM.find_partitions() if len(ccm_partitions) == 0: return 1 if len(ccm_partitions) > 1: self.CM.log("Warn: %d cluster partitions detected:" %len(ccm_partitions)) for partition in ccm_partitions: self.CM.log("\t %s" %partition) for partition in ccm_partitions: partition_passed = 0 if self.audit_partition(partition) == 0: passed = 0 return passed def trim_string(self, avalue): if not avalue: return None if len(avalue) > 1: return avalue[:-1] def trim2int(self, avalue): if not avalue: return None if len(avalue) > 1: return int(avalue[:-1]) def audit_partition(self, partition): passed = 1 dc_found = [] dc_allowed_list = [] lowest_epoche = None node_list = partition.split() self.CM.debug("Auditing partition: %s" %(partition)) for node in node_list: if self.CM.ShouldBeStatus[node] != self.CM["up"]: self.CM.log("Warn: Node %s appeared out of nowhere" %(node)) self.CM.ShouldBeStatus[node] = self.CM["up"] # not in itself a reason to fail the audit (not what we're # checking for in this audit) self.NodeState[node] = self.CM.rsh.readaline( node, self.CM["StatusCmd"]%node) self.NodeEpoche[node] = self.CM.rsh.readaline( node, self.CM["EpocheCmd"]) self.NodeQuorum[node] = self.CM.rsh.readaline( node, self.CM["QuorumCmd"]) self.NodeState[node] = self.trim_string(self.NodeState[node]) self.NodeEpoche[node] = self.trim2int(self.NodeEpoche[node]) self.NodeQuorum[node] = self.trim_string(self.NodeQuorum[node]) if not self.NodeEpoche[node]: self.CM.log("Warn: Node %s dissappeared: cant determin epoche" %(node)) self.CM.ShouldBeStatus[node] = self.CM["down"] # not in itself a reason to fail the audit (not what we're # checking for in this audit) elif lowest_epoche == None or self.NodeEpoche[node] < lowest_epoche: lowest_epoche = self.NodeEpoche[node] if not lowest_epoche: self.CM.log("Lowest epoche not determined in %s" % (partition)) passed = 0 for node in node_list: if self.CM.ShouldBeStatus[node] == self.CM["up"]: if self.CM.is_node_dc(node, self.NodeState[node]): dc_found.append(node) if self.NodeEpoche[node] == lowest_epoche: self.CM.debug("%s: OK" % node) elif not self.NodeEpoche[node]: self.CM.debug("Check on %s ignored: no node epoche" % node) elif not lowest_epoche: self.CM.debug("Check on %s ignored: no lowest epoche" % node) else: self.CM.log("DC %s is not the oldest node (%d vs. %d)" %(node, self.NodeEpoche[node], lowest_epoche)) passed = 0 if len(dc_found) == 0: self.CM.log("DC not found on any of the %d allowed nodes: %s (of %s)" %(len(dc_allowed_list), str(dc_allowed_list), str(node_list))) elif len(dc_found) > 1: self.CM.log("%d DCs (%s) found in cluster partition: %s" %(len(dc_found), str(dc_found), str(node_list))) passed = 0 elif self.CM.Env["CIBResource"] == 1 and self.NodeQuorum[dc_found[0]]: if self.audit_dc_resources(node_list, dc_found) == 0: passed = 0 Resources = self.CM.Resources() for node in node_list: if self.CM.ShouldBeStatus[node] == self.CM["up"]: for resource in Resources: if resource.rid == "rsc_"+node: if resource.IsRunningOn(node) == 0: self.CM.log("Node %s is not running its own resource" %(node)) passed = 0 elif self.CM.Env["CIBResource"] == 1: # no quorum means no resource management self.CM.debug("Not auditing resources - no quorum") if passed == 0: for node in node_list: if self.CM.ShouldBeStatus[node] == self.CM["up"]: self.CM.log("epoche %s : %s" %(self.NodeEpoche[node], self.NodeState[node])) return passed def audit_dc_resources(self, node_list, dc_list): passed = 1 Resources = self.CM.Resources() for resource in Resources: if resource.rid == "DcIPaddr": self.CM.debug("Auditing resource: %s" %(resource)) # All DCs are running the resource for dc in dc_list: if self.NodeQuorum[dc]: if resource.IsRunningOn(dc) == 0: self.CM.log("Resource %s not running on DC: %s" %(resource, dc)) passed = 0 # All nodes running the resource are DCs for node in node_list: if resource.IsRunningOn(node): if self.CM.is_node_dc(node, self.NodeState[node]) == 0: self.CM.log("Resource %s is running on non-DC node %s" %("DcIPaddr", node)) passed = 0 return passed def name(self): return "PartitionAudit" def is_applicable(self): if self.CM["Name"] == "linux-ha-v2": return 1 return 0 AllAuditClasses.append(CrmdStateAudit) AllAuditClasses.append(PartitionAudit) AllAuditClasses.append(ResourceAudit) AllAuditClasses.append(HAResourceAudit) AllAuditClasses.append(CIBAudit) def AuditList(cm): result = [] for auditclass in AllAuditClasses: result.append(auditclass(cm)) return result