diff --git a/cts/CM_LinuxHAv2.py.in b/cts/CM_LinuxHAv2.py.in index 070a5dff51..9539989386 100755 --- a/cts/CM_LinuxHAv2.py.in +++ b/cts/CM_LinuxHAv2.py.in @@ -1,612 +1,612 @@ #!@PYTHON@ '''CTS: Cluster Testing System: LinuxHA v2 dependent modules... ''' __copyright__=''' Author: Huang Zhen Copyright (C) 2004 International Business Machines Additional Audits, Revised Start action, Default Configuration: Copyright (C) 2004 Andrew Beekhof ''' # # 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 os,sys,CTS,CTSaudits,CTStests, warnings from CTS import * from CM_hb import HeartbeatCM from xml.dom.minidom import * from CTSaudits import ClusterAudit from CTStests import * from CIB import * ####################################################################### # # LinuxHA v2 dependent modules # ####################################################################### class LinuxHAv2(HeartbeatCM): ''' The linux-ha version 2 cluster manager class. It implements the things we need to talk to and manipulate linux-ha version 2 clusters ''' def __init__(self, Environment, randseed=None): HeartbeatCM.__init__(self, Environment, randseed=randseed) self.cib_installed = 0 self.config = None self.cluster_monitor = 0 self.use_short_names = 1 self.update({ "Name" : "linux-ha-v2", "DeadTime" : 300, "StartTime" : 300, # Max time to start up "StableTime" : 30, "StartCmd" : "@INITDIR@/heartbeat@INIT_EXT@ start > /dev/null 2>&1", "StopCmd" : "@INITDIR@/heartbeat@INIT_EXT@ stop > /dev/null 2>&1", "ElectionCmd" : "@sbindir@/crmadmin -E %s", "StatusCmd" : "@sbindir@/crmadmin -S %s 2>/dev/null", "EpocheCmd" : "@sbindir@/ccm_tool -e", "QuorumCmd" : "@sbindir@/ccm_tool -q", "CibQuery" : "@sbindir@/cibadmin -Ql", "ParitionCmd" : "@sbindir@/ccm_tool -p", "IsRscRunning" : "@libdir@/heartbeat/lrmadmin -E %s monitor 0 0 EVERYTIME 2>/dev/null|grep return", "ExecuteRscOp" : "@libdir@/heartbeat/lrmadmin -E %s %s 0 0 EVERYTIME 2>/dev/null", "CIBfile" : "%s:@HA_VARLIBDIR@/heartbeat/crm/cib.xml", "TmpDir" : "/tmp", "BreakCommCmd2" : "@libdir@/heartbeat/TestHeartbeatComm break-communication %s>/dev/null 2>&1", "IsIPAddrRscRunning" : "", "StandbyCmd" : "@sbindir@/crm_standby -U %s -v %s 2>/dev/null", "UUIDQueryCmd" : "@sbindir@/crmadmin -N", "StandbyQueryCmd" : "@sbindir@/crm_standby -GQ -U %s 2>/dev/null", # Patterns to look for in the log files for various occasions... "Pat:DC_IDLE" : "crmd:.*State transition.*-> S_IDLE", # This wont work if we have multiple partitions # Use: "Pat:They_started" : "%s crmd:.*State transition.*-> S_NOT_DC", "Pat:They_started" : "Updating node state to member for %s", "Pat:We_started" : "%s crmd:.*State transition.*-> S_IDLE", "Pat:We_stopped" : "%s heartbeat.*Heartbeat shutdown complete", "Pat:Logd_stopped" : "%s logd:.*Exiting write process", "Pat:They_stopped" : "%s crmd:.*LOST:.* %s ", "Pat:All_stopped" : "%s .*heartbeat.*Heartbeat shutdown complete", "Pat:They_dead" : "node %s.*: is dead", "Pat:TransitionComplete" : "Transition status: Complete: complete", # Bad news Regexes. Should never occur. "BadRegexes" : ( r"ERROR:", r"CRIT:", r"Shutting down\.", r"Forcing shutdown\.", r"Timer I_TERMINATE just popped", r"input=I_ERROR", r"input=I_FAIL", r"input=I_INTEGRATED cause=C_TIMER_POPPED", r"input=I_FINALIZED cause=C_TIMER_POPPED", r"input=I_ERROR", r", exiting\.", r"WARN.*Ignoring HA message.*vote.*not in our membership list", r"pengine:.*Attempting recovery of resource", r"pengine:.*Handling failed ", r"tengine:.*is taking more than 2x its timeout", r"Confirm not received from", r"Welcome reply not received from", r"Attempting to schedule .* after a stop", r"Resource .* was active at shutdown", r"duplicate entries for call_id", ), }) del self["Standby"] self.check_transitions = 0 self.check_elections = 0 self.CIBsync = {} self.default_cts_cib=CIB(self).cib() self.debug(self.default_cts_cib) def errorstoignore(self): # At some point implement a more elegant solution that # also produces a report at the end '''Return list of errors which are known and very noisey should be ignored''' if 1: return [ "crmadmin:" ] return [] def install_config(self, node): if not self.CIBsync.has_key(node) and self.Env["ClobberCIB"] == 1: self.CIBsync[node] = 1 - self.rsh.remote_python_call(node, "os", "system", "rm -f @HA_VARLIBDIR@/heartbeat/crm/cib.xml") - self.rsh.remote_python_call(node, "os", "system", "rm -f @HA_VARLIBDIR@/heartbeat/crm/cib.xml.sig") + self.rsh.remote_py(node, "os", "system", "rm -f @HA_VARLIBDIR@/heartbeat/crm/cib.xml") + self.rsh.remote_py(node, "os", "system", "rm -f @HA_VARLIBDIR@/heartbeat/crm/cib.xml.sig") # Only install the CIB on the first node, all the other ones will pick it up from there if self.cib_installed == 1: return None self.cib_installed = 1 if self.Env["CIBfilename"] == None: self.debug("Installing Generated CIB on node %s" %(node)) warnings.filterwarnings("ignore") cib_file=os.tmpnam() warnings.resetwarnings() os.system("rm -f "+cib_file) self.debug("Creating new CIB for " + node + " in: " + cib_file) os.system("echo \'" + self.default_cts_cib + "\' > " + cib_file) if 0!=self.rsh.cp(cib_file,"root@" + (self["CIBfile"]%node)): raise ValueError("Can not scp file to %s "%node) os.system("rm -f "+cib_file) else: self.debug("Installing CIB (%s) on node %s" %(self.Env["CIBfilename"], node)) if 0!=self.rsh.cp(self.Env["CIBfilename"], "root@" + (self["CIBfile"]%node)): raise ValueError("Can not scp file to %s "%node) - self.rsh.remote_python_call(node, "os", "system", "chown @HA_CCMUSER@ @HA_VARLIBDIR@/heartbeat/crm/cib.xml") + self.rsh.remote_py(node, "os", "system", "chown @HA_CCMUSER@ @HA_VARLIBDIR@/heartbeat/crm/cib.xml") def prepare(self): '''Finish the Initialization process. Prepare to test...''' for node in self.Env["nodes"]: self.ShouldBeStatus[node] = "" self.StataCM(node) def test_node_CM(self, node): '''Report the status of the cluster manager on a given node''' if not self.ns.WaitForNodeToComeUp(node): self.log("WaitForNodeToComeUp failed, node %s set to down" % node) self.ShouldBeStatus[node] == self["down"] return 3 watchpats = [ ] watchpats.append("Current ping state: (S_IDLE|S_NOT_DC)") watchpats.append(self["Pat:They_started"]%node) idle_watch = CTS.LogWatcher(self["LogFileName"], watchpats) idle_watch.setwatch() out=self.rsh.readaline(node, self["StatusCmd"]%node) ret= (string.find(out, 'ok') != -1) self.debug("Node %s status: %s" %(node, out)) if not ret: if self.ShouldBeStatus[node] == self["up"]: self.log( "Node status for %s is %s but we think it should be %s" %(node, self["down"], self.ShouldBeStatus[node])) self.ShouldBeStatus[node]=self["down"] return 0 if self.ShouldBeStatus[node] == self["down"]: self.log( "Node status for %s is %s but we think it should be %s: %s" %(node, self["up"], self.ShouldBeStatus[node], out)) self.ShouldBeStatus[node]=self["up"] # check the output first - because syslog-ng looses messages if string.find(out, 'S_NOT_DC') != -1: # Up and stable return 2 if string.find(out, 'S_IDLE') != -1: # Up and stable return 2 # fall back to syslog-ng and wait if not idle_watch.look(): # just up self.debug("Warn: Node %s is unstable: %s" %(node, out)) return 1 # Up and stable return 2 # Is the node up or is the node down def StataCM(self, node): '''Report the status of the cluster manager on a given node''' if self.test_node_CM(node) > 0: return 1 return None # Being up and being stable is not the same question... def node_stable(self, node): '''Report the status of the cluster manager on a given node''' if self.test_node_CM(node) == 2: return 1 self.log("Warn: Node %s not stable" %(node)) return None def cluster_stable(self, timeout=None): watchpats = [ ] watchpats.append("Current ping state: S_IDLE") watchpats.append(self["Pat:DC_IDLE"]) if timeout == None: timeout = self["DeadTime"] idle_watch = CTS.LogWatcher(self["LogFileName"], watchpats, timeout) idle_watch.setwatch() if not self.ns.WaitForAllNodesToComeUp(self.Env["nodes"]): self.log("WaitForAllNodesToComeUp failed.") return None any_up = 0 for node in self.Env["nodes"]: # have each node dump its current state if self.ShouldBeStatus[node] == self["up"]: self.rsh.readaline(node, (self["StatusCmd"] %node) ) any_up = 1 if any_up == 0 or idle_watch.look(): return 1 self.log("Warn: Cluster Master not IDLE") return None def is_node_dc(self, node, status_line=None): rc = 0 if not self.ns.WaitForNodeToComeUp(node): self.log("WaitForNodeToComeUp %s failed." % node) return rc if not status_line: status_line = self.rsh.readaline(node, self["StatusCmd"]%node) if not status_line: rc = 0 elif string.find(status_line, 'S_IDLE') != -1: rc = 1 elif string.find(status_line, 'S_INTEGRATION') != -1: rc = 1 elif string.find(status_line, 'S_FINALIZE_JOIN') != -1: rc = 1 elif string.find(status_line, 'S_POLICY_ENGINE') != -1: rc = 1 elif string.find(status_line, 'S_TRANSITION_ENGINE') != -1: rc = 1 if rc == 1: self.debug("%s _is_ the DC" % node) return rc def isolate_node(self, node, allowlist): '''isolate the communication between the nodes''' if not self.ns.WaitForNodeToComeUp(node): self.log("WaitForNodeToComeUp %s failed." % node) return None rc = self.rsh(node, self["BreakCommCmd2"]%allowlist) if rc == 0: return 1 else: self.log("Could not break the communication from node: %s",node) return None def Configuration(self): if self.config: return self.config.getElementsByTagName('configuration')[0] warnings.filterwarnings("ignore") cib_file=os.tmpnam() warnings.resetwarnings() os.system("rm -f "+cib_file) if self.Env["ClobberCIB"] == 1: if self.Env["CIBfilename"] == None: self.debug("Creating new CIB in: " + cib_file) os.system("echo \'"+ self.default_cts_cib +"\' > "+ cib_file) else: os.system("cp "+self.Env["CIBfilename"]+" "+cib_file) else: if 0!=self.rsh.cp("root@"+self["CIBfile"]%self.Env["nodes"][0],cib_file): raise ValueError("Can not copy file to %s, maybe permission denied"%self["TmpDir"]) self.config = parse(cib_file) os.remove(cib_file) return self.config.getElementsByTagName('configuration')[0] def Resources(self): ResourceList = [] #read resources in cib configuration = self.Configuration() resources = configuration.getElementsByTagName('resources')[0] rscs = configuration.getElementsByTagName('primitive') incs = configuration.getElementsByTagName('clone') groups = configuration.getElementsByTagName('group') for rsc in rscs: if rsc in resources.childNodes: ResourceList.append(HAResource(self,rsc)) for grp in groups: for rsc in rscs: if rsc in grp.childNodes: if self.use_short_names: resource = HAResource(self,rsc) else: resource = HAResource(self,rsc,grp.getAttribute('id')) ResourceList.append(resource) for inc in incs: max = 0 inc_name = inc.getAttribute("id") instance_attributes = inc.getElementsByTagName('instance_attributes')[0] attributes = instance_attributes.getElementsByTagName('attributes')[0] nvpairs = attributes.getElementsByTagName('nvpair') for nvpair in nvpairs: if nvpair.getAttribute("name") == "clone_max": max = int(nvpair.getAttribute("value")) inc_rsc = inc.getElementsByTagName('primitive')[0] for i in range(0,max): rsc = HAResource(self,inc_rsc) rsc.inc_no = i rsc.inc_name = inc_name rsc.inc_max = max if self.use_short_names: rsc.rid = rsc.rid + ":%d"%i else: rsc.rid = inc_name+":"+rsc.rid + ":%d"%i rsc.Instance = rsc.rid ResourceList.append(rsc) return ResourceList def ResourceGroups(self): GroupList = [] #read resources in cib configuration = self.Configuration() groups = configuration.getElementsByTagName('group') rscs = configuration.getElementsByTagName('primitive') for grp in groups: group = [] GroupList.append(group) for rsc in rscs: if rsc in grp.childNodes: if self.use_short_names: resource = HAResource(self,rsc) else: resource = HAResource(self,rsc,grp.getAttribute('id')) group.append(resource) return GroupList def Dependencies(self): DependencyList = [] #read dependency in cib configuration=self.Configuration() constraints=configuration.getElementsByTagName('constraints')[0] rsc_to_rscs=configuration.getElementsByTagName('rsc_to_rsc') for node in rsc_to_rscs: dependency = {} dependency["id"]=node.getAttribute('id') dependency["from"]=node.getAttribute('from') dependency["to"]=node.getAttribute('to') dependency["type"]=node.getAttribute('type') dependency["strength"]=node.getAttribute('strength') DependencyList.append(dependency) return DependencyList def find_partitions(self): ccm_partitions = [] if not self.ns.WaitForAllNodesToComeUp(self.Env["nodes"]): self.log("WaitForAllNodesToComeUp failed.") return None for node in self.Env["nodes"]: self.debug("Retrieving partition details for %s" %node) if self.ShouldBeStatus[node] == self["up"]: partition = self.rsh.readaline(node, self["ParitionCmd"]) if not partition: self.log("no partition details for %s" %node) elif len(partition) > 2: partition = partition[:-1] self.debug("partition details for %s: %s" %(node, partition)) found=0 for a_partition in ccm_partitions: if partition == a_partition: found = 1 if found == 0: self.debug("Adding partition") ccm_partitions.append(partition) else: self.log("bad partition details for %s" %node) return ccm_partitions def HasQuorum(self, node_list): # If we are auditing a partition, then one side will # have quorum and the other not. # So the caller needs to tell us which we are checking # If no value for node_list is specified... assume all nodes if not node_list: node_list = self.Env["nodes"] if not self.ns.WaitForAllNodesToComeUp(node_list): self.log("WaitForAllNodesToComeUp failed in HasQuorum.") return 0 for node in node_list: if self.ShouldBeStatus[node] == self["up"]: quorum = self.rsh.readaline(node, self["QuorumCmd"]) if string.find(quorum, "1") != -1: return 1 elif string.find(quorum, "0") != -1: return 0 else: self.log("WARN: Unexpected quorum test result from "+ node +":"+ quorum) return 0 def Components(self): complist = [Process("lrmd",self),Process("crmd",self)] if self.Env["DoFencing"] == 1 : complist.append(Process("stonithd",self)) complist.append(Process("heartbeat",self)) return complist def NodeUUID(self, node): if not self.ns.WaitForNodeToComeUp(node): self.log("WaitForNodeToComeUp %s failed in NodeUUID." % node) return "" lines = self.rsh.readlines(node, self["UUIDQueryCmd"]) for line in lines: self.debug("UUIDLine:"+ line) m = re.search(r'%s.+\((.+)\)' % node, line) if m: return m.group(1) return "" def StandbyStatus(self, node): if not self.ns.WaitForNodeToComeUp(node): self.log("WaitForNodeToComeUp %s failed in StandbyStatus." % node) return "" out=self.rsh.readaline(node, self["StandbyQueryCmd"]%node) if not out: return "off" out = out[:-1] self.debug("Standby result: "+out) return out # status == "on" : Enter Standby mode # status == "off": Enter Active mode def SetStandbyMode(self, node, status): if not self.ns.WaitForNodeToComeUp(node): self.log("WaitForNodeToComeUp %s failed in SetStandbyMode." % node) return True current_status = self.StandbyStatus(node) cmd = self["StandbyCmd"] % (node, status) ret = self.rsh(node, cmd) return True class HAResource(Resource): def __init__(self, cm, node, group=None): ''' Get information from xml node ''' if group == None : self.rid = str(node.getAttribute('id')) else : self.rid = group + ":" + str(node.getAttribute('id')) self.rclass = str(node.getAttribute('class')) self.rtype = str(node.getAttribute('type')) self.inc_name = None self.inc_no = -1 self.inc_max = -1 self.rparameters = {} nvpairs = [] list = node.getElementsByTagName('instance_attributes') if len(list) > 0: attributes = list[0] list = attributes.getElementsByTagName('attributes') if len(list) > 0: parameters = list[0] nvpairs = parameters.getElementsByTagName('nvpair') for nvpair in nvpairs: name=nvpair.getAttribute('name') value=nvpair.getAttribute('value') self.rparameters[name]=value # This should normally be called first... FIXME! Resource.__init__(self, cm, self.rtype, self.rid) # resources that dont need quorum will have: # ops = node.getElementsByTagName('op') for op in ops: if op.getAttribute('name') == "start" and op.getAttribute('prereq') == "nothing": self.needs_quorum = 0 if self.needs_quorum == 0: self.CM.debug("Resource %s does not need quorum" % self.rid) # else: # self.CM.debug("Resource %s DOES need quorum" % self.rid) def IsRunningOn(self, nodename): ''' This member function returns true if our resource is running on the given node in the cluster. We call the status operation for the resource script. ''' if not self.CM.ns.WaitForNodeToComeUp(nodename): self.CM.log("WaitForNodeToComeUp %s failed in IsRunningOn." % node) return 0 out=self.CM.rsh.readaline(nodename, self.CM["IsRscRunning"]%self.rid) return re.search("0",out) def RunningNodes(self): ResourceNodes = [] for node in self.CM.Env["nodes"]: if self.CM.ShouldBeStatus[node] == self.CM["up"]: if self.IsRunningOn(node): ResourceNodes.append(node) return ResourceNodes def _ResourceOperation(self, operation, nodename): ''' Execute an operation on the resource ''' if not self.CM.ns.WaitForNodeToComeUp(nodename): self.CM.log("WaitForNodeToComeUp %s failed in _ResourceOperation." % node) return 0 self.CM.rsh.readaline(nodename, self.CM["ExecuteRscOp"]%(self.rid,operation)) return self.CM.rsh.lastrc == 0 def Start(self, nodename): ''' This member function starts or activates the resource. ''' return self._ResourceOperation("start", nodename) def Stop(self, nodename): ''' This member function stops or deactivates the resource. ''' return self._ResourceOperation("stop", nodename) def IsWorkingCorrectly(self, nodename): return self._ResourceOperation("monitor", nodename) ####################################################################### # # A little test code... # # Which you are advised to completely ignore... # ####################################################################### if __name__ == '__main__': pass diff --git a/cts/CTS.py.in b/cts/CTS.py.in index 55231fa360..fc6c32ea7b 100755 --- a/cts/CTS.py.in +++ b/cts/CTS.py.in @@ -1,1157 +1,1173 @@ #!@PYTHON@ '''CTS: Cluster Testing System: Main module Classes related to testing high-availability clusters... Lots of things are implemented. Lots of things are not implemented. We have many more ideas of what to do than we've implemented. ''' __copyright__=''' Copyright (C) 2000, 2001 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 types, string, select, sys, time, re, os, struct, os, signal import base64, pickle, binascii from UserDict import UserDict from syslog import * from popen2 import Popen3 class RemoteExec: '''This is an abstract remote execution class. It runs a command on another machine - somehow. The somehow is up to us. This particular class uses ssh. Most of the work is done by fork/exec of ssh or scp. ''' def __init__(self): # -n: no stdin, -x: no X11 self.Command = "@SSH@ -l root -n -x" # -f: ssh to background self.CommandnoBlock = "@SSH@ -f -l root -n -x" # -B: batch mode, -q: no stats (quiet) self.CpCommand = "@SCP@ -B -q" + self.OurNode=string.lower(os.uname()[1]) + def setcmd(self, rshcommand): '''Set the name of the remote shell command''' self.Command = rshcommand def _fixcmd(self, cmd): return re.sub("\'", "'\\''", cmd) def _cmd(self, *args): '''Compute the string that will run the given command on the given remote system''' args= args[0] sysname = args[0] command = args[1] - ret = self.Command + " " + sysname + " '" + self._fixcmd(command) + "'" + + if sysname == self.OurNode or sysname == "localhost": + ret = self._fixcmd(command) + else: + ret = self.Command + " " + sysname + " '" + self._fixcmd(command) + "'" #print ("About to run %s\n" % ret) return ret def __call__(self, *args): '''Run the given command on the given remote system If you call this class like a function, this is the function that gets called. It just runs it roughly as though it were a system() call on the remote machine. The first argument is name of the machine to run it on. ''' count=0; rc = 0; while count < 3: rc = os.system(self._cmd(args)) if rc == 0: return rc print "Retrying command %s" % self._cmd(args) count=count+1 return rc def popen(self, *args): '''popen the given remote command on the remote system. As in __call__, the first argument is name of the machine to run it on. ''' #print "Now running %s\n" % self._cmd(args) return Popen3(self._cmd(args), None) def readaline(self, *args): '''Run a command on the remote machine and capture 1 line of stdout from the given remote command As in __call__, the first argument is name of the machine to run it on. ''' p = self.popen(args[0], args[1]) p.tochild.close() result = p.fromchild.readline() p.fromchild.close() self.lastrc = p.wait() return result def readlines(self, *args): p = self.popen(args[0], args[1]) p.tochild.close() result = p.fromchild.readlines() p.fromchild.close() self.lastrc = p.wait() return result def cp(self, *args): '''Perform a remote copy''' cpstring=self.CpCommand for arg in args: cpstring = cpstring + " \'" + arg + "\'" count=0; rc = 0; for i in range(3): rc = os.system(cpstring) if rc == 0: return rc print "Retrying command %s" % cpstring return rc def noBlock(self, *args): '''Perform a remote execution without waiting for it to finish''' sshnoBlock = self.CommandnoBlock for arg in args: sshnoBlock = sshnoBlock + " \'" + arg + "\'" count=0; rc = 0; for i in range(3): rc = os.system(sshnoBlock) if rc == 0: return rc print "Retrying command %s" % sshnoBlock return rc - - def remote_python_call(self, node, module, func, *args): + + def remote_py(self, node, module, func, *args): '''Execute a remote python function If the call success, lastrc == 0 and return result. If the call fail, lastrc == 1 and return the reason (string) ''' encode_args = binascii.b2a_base64(pickle.dumps(args)) - result = self.readaline(node, \ + result = self.readlines(node, \ string.join(["@libdir@/heartbeat/cts/CTSproxy.py",module,func,encode_args])) - if self.lastrc == 0 : - return pickle.loads(binascii.a2b_base64(result)) + + if result != None: + result.pop() + + if self.lastrc == 0: + last_line="" + if result != None: + array_len = len(result) + if array_len > 0: + last_line=result.pop() + #print "result: %s" % repr(last_line) + return pickle.loads(binascii.a2b_base64(last_line)), result - return result + return -1, result class LogWatcher: '''This class watches logs for messages that fit certain regular expressions. Watching logs for events isn't the ideal way to do business, but it's better than nothing :-) On the other hand, this class is really pretty cool ;-) The way you use this class is as follows: Construct a LogWatcher object Call setwatch() when you want to start watching the log Call look() to scan the log looking for the patterns ''' def __init__(self, log, regexes, timeout=10, debug=None): '''This is the constructor for the LogWatcher class. It takes a log name to watch, and a list of regular expressions to watch for." ''' # Validate our arguments. Better sooner than later ;-) for regex in regexes: assert re.compile(regex) self.regexes = regexes self.filename = log self.debug=debug self.whichmatch = -1 self.unmatched = None if self.debug: print "Debug now on for for log", log self.Timeout = int(timeout) self.returnonlymatch = None if not os.access(log, os.R_OK): raise ValueError("File [" + log + "] not accessible (r)") def setwatch(self, frombeginning=None): '''Mark the place to start watching the log from. ''' self.file = open(self.filename, "r") self.size = os.path.getsize(self.filename) if not frombeginning: self.file.seek(0,2) def ReturnOnlyMatch(self, onlymatch=1): '''Mark the place to start watching the log from. ''' self.returnonlymatch = onlymatch def look(self, timeout=None): '''Examine the log looking for the given patterns. It starts looking from the place marked by setwatch(). This function looks in the file in the fashion of tail -f. It properly recovers from log file truncation, but not from removing and recreating the log. It would be nice if it recovered from this as well :-) We return the first line which matches any of our patterns. ''' last_line=None first_line=None if timeout == None: timeout = self.Timeout done=time.time()+timeout+1 if self.debug: print "starting search: timeout=%d" % timeout for regex in self.regexes: print "Looking for regex: ", regex while (timeout <= 0 or time.time() <= done): newsize=os.path.getsize(self.filename) if self.debug > 4: print "newsize = %d" % newsize if newsize < self.size: # Somebody truncated the log! if self.debug: print "Log truncated!" self.setwatch(frombeginning=1) continue if newsize > self.file.tell(): line=self.file.readline() if self.debug > 2: print "Looking at line:", line if line: last_line=line if not first_line: first_line=line if self.debug: print "First line: "+ line which=-1 for regex in self.regexes: which=which+1 if self.debug > 3: print "Comparing line to ", regex matchobj = re.search(regex, line) if matchobj: self.whichmatch=which if self.returnonlymatch: return matchobj.group(self.returnonlymatch) else: if self.debug: print "Returning line" return line newsize=os.path.getsize(self.filename) if self.file.tell() == newsize: if timeout > 0: time.sleep(0.025) else: if self.debug: print "End of file" if self.debug: print "Last line: "+last_line return None if self.debug: print "Timeout" if self.debug: print "Last line: "+last_line return None def lookforall(self, timeout=None): '''Examine the log looking for ALL of the given patterns. It starts looking from the place marked by setwatch(). We return when the timeout is reached, or when we have found ALL of the regexes that were part of the watch ''' if timeout == None: timeout = self.Timeout save_regexes = self.regexes returnresult = [] while (len(self.regexes) > 0): oneresult = self.look(timeout) if not oneresult: self.unmatched = self.regexes self.regexes = save_regexes return None returnresult.append(oneresult) del self.regexes[self.whichmatch] self.unmatched = None self.regexes = save_regexes return returnresult # In case we ever want multiple regexes to match a single line... #- del self.regexes[self.whichmatch] #+ tmp_regexes = self.regexes #+ self.regexes = [] #+ which = 0 #+ for regex in tmp_regexes: #+ matchobj = re.search(regex, oneresult) #+ if not matchobj: #+ self.regexes.append(regex) class NodeStatus: def __init__(self, Env): self.Env = Env def IsNodeBooted(self, node): '''Return TRUE if the given node is booted (responds to pings''' return os.system("@PING@ -nq -c1 @PING_TIMEOUT_OPT@ %s >/dev/null 2>&1" % node) == 0 def WaitForNodeToComeUp(self, node, Timeout=300): '''Return TRUE when given node comes up, or None/FALSE if timeout''' timeout=Timeout anytimeouts=0 while timeout > 0: if self.IsNodeBooted(node): if anytimeouts: # Fudge to wait for the system to finish coming up time.sleep(30) self.Env.log("Node %s now up" % node) return 1 time.sleep(1) if (not anytimeouts): self.Env.log("Waiting for node %s to come up" % node) anytimeouts=1 timeout = timeout - 1 self.Env.log("%s did not come up within %d tries" % (node, Timeout)) def WaitForAllNodesToComeUp(self, nodes, timeout=300): '''Return TRUE when all nodes come up, or FALSE if timeout''' for node in nodes: if not self.WaitForNodeToComeUp(node, timeout): return None return 1 class ClusterManager(UserDict): '''The Cluster Manager class. This is an subclass of the Python dictionary class. (this is because it contains lots of {name,value} pairs, not because it's behavior is that terribly similar to a dictionary in other ways.) This is an abstract class which class implements high-level operations on the cluster and/or its cluster managers. Actual cluster managers classes are subclassed from this type. One of the things we do is track the state we think every node should be in. ''' def __InitialConditions(self): #if os.geteuid() != 0: # raise ValueError("Must Be Root!") None def _finalConditions(self): for key in self.keys(): if self[key] == None: raise ValueError("Improper derivation: self[" + key + "] must be overridden by subclass.") def __init__(self, Environment, randseed=None): self.Env = Environment self.__InitialConditions() self.data = { "up" : "up", # Status meaning up "down" : "down", # Status meaning down "StonithCmd" : "@sbindir@/stonith -t baytech -p '10.10.10.100 admin admin' %s", "DeadTime" : 30, # Max time to detect dead node... "StartTime" : 90, # Max time to start up # # These next values need to be overridden in the derived class. # "Name" : None, "StartCmd" : None, "StopCmd" : None, "StatusCmd" : None, "RereadCmd" : None, "StartDRBDCmd" : None, "StopDRBDCmd" : None, "StatusDRBDCmd" : None, "DRBDCheckconf" : None, "BreakCommCmd" : None, "FixCommCmd" : None, "TestConfigDir" : None, "LogFileName" : None, "Pat:We_started" : None, "Pat:They_started" : None, "Pat:We_stopped" : None, "Pat:They_stopped" : None, "BadRegexes" : None, # A set of "bad news" regexes # to apply to the log } self.rsh = RemoteExec() self.ShouldBeStatus={} self.OurNode=string.lower(os.uname()[1]) self.ShouldBeStatus={} self.ns = NodeStatus(self.Env) def errorstoignore(self): '''Return list of errors which are 'normal' and should be ignored''' return [] def log(self, args): self.Env.log(args) def debug(self, args): self.Env.debug(args) def prepare(self): '''Finish the Initialization process. Prepare to test...''' for node in self.Env["nodes"]: if self.StataCM(node): self.ShouldBeStatus[node]=self["up"] else: self.ShouldBeStatus[node]=self["down"] def upcount(self): '''How many nodes are up?''' count=0 for node in self.Env["nodes"]: if self.ShouldBeStatus[node]==self["up"]: count=count+1 return count def TruncLogs(self): '''Truncate the log for the cluster manager so we can start clean''' if self["LogFileName"] != None: os.system("cp /dev/null " + self["LogFileName"]) def install_config(self, node): return None def StartaCM(self, node): '''Start up the cluster manager on a given node''' self.debug("Starting %s on node %s" %(self["Name"], node)) if not self.ShouldBeStatus.has_key(node): self.ShouldBeStatus[node] = self["down"] if self.ShouldBeStatus[node] != self["down"]: return 1 if not self.ns.WaitForNodeToComeUp(node): self.log("WaitForNodeToComeUp %s failed" % node) return 0 patterns = [] # Technically we should always be able to notice ourselves starting if self.upcount() == 0: patterns.append(self["Pat:We_started"] % node) else: patterns.append(self["Pat:They_started"] % node) watch = LogWatcher( self["LogFileName"], patterns, timeout=self["StartTime"]+10) watch.setwatch() self.install_config(node) # Clear out the host cache to prevent the PE from shooting nodes # during StartOnebyOne # self.rsh.remote_python_call( # node, "os", "system", "rm -f @HA_VARLIBDIR@/heartbeat/hostcache") if not self.ns.WaitForNodeToComeUp(node): self.log("Second WaitForNodeToComeUp %s failed" % node) return 0 if self.StataCM(node) and self.cluster_stable(self["DeadTime"]): self.log ("%s was already started" %(node)) return 1 if self.rsh(node, self["StartCmd"]) != 0: self.log ("Warn: Start command failed on node %s" %(node)) return None self.ShouldBeStatus[node]=self["up"] watch_result = watch.lookforall() if watch.unmatched: for regex in watch.unmatched: self.log ("Warn: Startup pattern not found: %s" %(regex)) if watch_result: #self.debug("Found match: "+ repr(watch_result)) self.cluster_stable(self["DeadTime"]) return 1 if self.StataCM(node) and self.cluster_stable(self["DeadTime"]): return 1 self.log ("Warn: Start failed for node %s" %(node)) return None def StartaCMnoBlock(self, node): '''Start up the cluster manager on a given node with none-block mode''' self.debug("Starting %s on node %s" %(self["Name"], node)) if not self.ns.IsNodeBooted(node): self.log("node %s not booted, StartaCMnoBlock failed." % node) return 0 self.rsh.noBlock(node, self["StartCmd"]) self.ShouldBeStatus[node]=self["up"] return 1 def StopaCM(self, node): '''Stop the cluster manager on a given node''' self.debug("Stopping %s on node %s" %(self["Name"], node)) if self.ShouldBeStatus[node] != self["up"]: return 1 if not self.ns.WaitForNodeToComeUp(node): self.log("WaitForNodeToComeUp %s filed" % node) return 0 if self.rsh(node, self["StopCmd"]) == 0: self.ShouldBeStatus[node]=self["down"] self.cluster_stable(self["DeadTime"]) return 1 else: self.log ("Could not stop %s on node %s" %(self["Name"], node)) return None def StopaCMnoBlock(self, node): '''Stop the cluster manager on a given node with none-block mode''' self.debug("Stopping %s on node %s" %(self["Name"], node)) if not self.ns.IsNodeBooted(node): self.log("node %s not booted, StopaCNnoBlock failed." % node) return 0 self.rsh.noBlock(node, self["StopCmd"]) self.ShouldBeStatus[node]=self["down"] return 1 def cluster_stable(self, timeout = None): time.sleep(self["StableTime"]) return 1 def node_stable(self, node): return 1 def RereadCM(self, node): '''Force the cluster manager on a given node to reread its config This may be a no-op on certain cluster managers. ''' if not self.ns.WaitForNodeToComeUp(node): self.log("WaitForNodeToComeUp %s filed" % node) return 0 rc=self.rsh(node, self["RereadCmd"]) if rc == 0: return 1 else: self.log ("Could not force %s on node %s to reread its config" % (self["Name"], node)) return None def StataCM(self, node): '''Report the status of the cluster manager on a given node''' out=self.rsh.readaline(node, self["StatusCmd"]) ret= (string.find(out, 'stopped') == -1) try: if ret: if self.ShouldBeStatus[node] != self["up"]: self.log( "Node status for %s is %s but we think it should be %s" % (node, self["up"], self.ShouldBeStatus[node])) else: if self.ShouldBeStatus[node] != self["down"]: self.log( "Node status for %s is %s but we think it should be %s" % (node, self["down"], self.ShouldBeStatus[node])) except KeyError: pass if ret: self.ShouldBeStatus[node]=self["up"] else: self.ShouldBeStatus[node]=self["down"] return ret def startall(self, nodelist=None): '''Start the cluster manager on every node in the cluster. We can do it on a subset of the cluster if nodelist is not None. ''' map = {} if not nodelist: nodelist=self.Env["nodes"] for node in nodelist: if self.ShouldBeStatus[node] == self["down"]: self.StartaCM(node) def stopall(self, nodelist=None): '''Stop the cluster managers on every node in the cluster. We can do it on a subset of the cluster if nodelist is not None. ''' map = {} if not nodelist: nodelist=self.Env["nodes"] for node in self.Env["nodes"]: if self.ShouldBeStatus[node] == self["up"]: self.StopaCM(node) def rereadall(self, nodelist=None): '''Force the cluster managers on every node in the cluster to reread their config files. We can do it on a subset of the cluster if nodelist is not None. ''' map = {} if not nodelist: nodelist=self.Env["nodes"] for node in self.Env["nodes"]: if self.ShouldBeStatus[node] == self["up"]: self.RereadCM(node) def statall(self, nodelist=None): '''Return the status of the cluster managers in the cluster. We can do it on a subset of the cluster if nodelist is not None. ''' result={} if not nodelist: nodelist=self.Env["nodes"] for node in nodelist: if self.StataCM(node): result[node] = self["up"] else: result[node] = self["down"] return result def isolate_node(self, node): '''isolate the communication between the nodes''' if not self.ns.WaitForNodeToComeUp(node): self.log("WaitForNodeToComeUp %s filed in isolate_node" % node) return None rc = self.rsh(node, self["BreakCommCmd"]) if rc == 0: return 1 else: self.log("Could not break the communication between the nodes frome node: %s" % node) return None def unisolate_node(self, node): '''fix the communication between the nodes''' if not self.ns.WaitForNodeToComeUp(node): self.log("WaitForNodeToComeUp %s filed in unisolate_node" % node) return None rc = self.rsh(node, self["FixCommCmd"]) if rc == 0: return 1 else: self.log("Could not fix the communication between the nodes from node: %s" % node) return None def reducecomm_node(self,node): '''reduce the communication between the nodes''' if not self.ns.WaitForNodeToComeUp(node): self.log("WaitForNodeToComeUp %s filed in reducecomm_node" % node) return None rc = self.rsh(node, self["ReduceCommCmd"]%(self.Env["XmitLoss"],self.Env["RecvLoss"])) if rc == 0: return 1 else: self.log("Could not reduce the communication between the nodes from node: %s" % node) return None def savecomm_node(self,node): '''save current the communication between the nodes''' rc = 0 if not self.ns.WaitForNodeToComeUp(node): self.log("WaitForNodeToComeUp %s filed in savecomm_node" % node) return None if float(self.Env["XmitLoss"])!=0 or float(self.Env["RecvLoss"])!=0 : rc = self.rsh(node, self["SaveFileCmd"]); if rc == 0: return 1 else: self.log("Could not save the communication between the nodes from node: %s" % node) return None def restorecomm_node(self,node): '''restore the saved communication between the nodes''' rc = 0 if not self.ns.WaitForNodeToComeUp(node): self.log("WaitForNodeToComeUp %s filed in restoreomm_node" % node) return None if float(self.Env["XmitLoss"])!=0 or float(self.Env["RecvLoss"])!=0 : rc = self.rsh(node, self["RestoreCommCmd"]); if rc == 0: return 1 else: self.log("Could not restore the communication between the nodes from node: %s" % node) return None def SyncTestConfigs(self): '''Synchronize test configurations throughout the cluster. This one's a no-op for FailSafe, since it does that by itself. ''' fromdir=self["TestConfigDir"] if not os.access(fromdir, os.F_OK | os.R_OK | os.W_OK): raise ValueError("Directory [" + fromdir + "] not accessible (rwx)") for node in self.Env["nodes"]: if node == self.OurNode: continue self.log("Syncing test configurations on " + node) # Perhaps I ought to use rsync... self.rsh.cp("-r", fromdir, node + ":" + fromdir) def SetClusterConfig(self, configpath="default", nodelist=None): '''Activate the named test configuration throughout the cluster. It would be useful to implement this :-) ''' pass return 1 def ResourceGroups(self): "Return a list of resource type/instance pairs for the cluster" raise ValueError("Abstract Class member (ResourceGroups)") def InternalCommConfig(self): "Return a list of paths: each patch consists of a tuple" raise ValueError("Abstract Class member (InternalCommConfig)") def HasQuorum(self, node_list): "Return TRUE if the cluster currently has quorum" # If we are auditing a partition, then one side will # have quorum and the other not. # So the caller needs to tell us which we are checking # If no value for node_list is specified... assume all nodes raise ValueError("Abstract Class member (HasQuorum)") def Components(self): raise ValueError("Abstract Class member (Components)") def TestLogging(self): patterns= [] prefix="Test message from " for node in self.Env["nodes"]: patterns.append(prefix + node) watch = LogWatcher(self["LogFileName"], patterns, 30 + len(self.Env["nodes"])) watch.setwatch() logpri = self.Env["logfacility"] + ".info" for node in self.Env["nodes"]: cmd="logger -p %s %s%s" % (logpri, prefix, node) if self.rsh.noBlock(node, cmd) != 0: self.log ("ERROR: Cannot execute remote command [%s] on %s" % (cmd, node)) watch_result = watch.lookforall() if watch.unmatched: self.log("ERROR: Remote logging is not working. Please fix before continuing.") for regex in watch.unmatched: self.log ("ERROR: Test message [%s] not found in logs." % (regex)) return None return 1 def CheckDf(self): dfcmd="df -B1000000 /var/log | tail -1 | tr -s ' ' | cut -d' ' -f2" dfmin=500 result=1 for node in self.Env["nodes"]: dfout=self.rsh.readaline(node, dfcmd) if not dfout: self.log ("ERROR: Cannot execute remote df command [%s] on %s" % (dfcmd, node)) else: try: idfout = int(dfout) except (ValueError, TypeError): self.log("Warning: df output from %s was invalid [%s]" % (node, dfout)) else: if idfout == 0: self.log("CRIT: Completely out of log disk space on %s" % node) result=None elif idfout <= 500: self.log("WARN: Low on log disk space (%d Mbytes) on %s" % (idfout, node)) return result class Resource: ''' This is an HA resource (not a resource group). A resource group is just an ordered list of Resource objects. ''' def __init__(self, cm, rsctype=None, instance=None): self.CM = cm self.ResourceType = rsctype self.Instance = instance self.needs_quorum = 1 def Type(self): return self.ResourceType def Instance(self, nodename): return self.Instance def IsRunningOn(self, nodename): ''' This member function returns true if our resource is running on the given node in the cluster. It is analagous to the "status" operation on SystemV init scripts and heartbeat scripts. FailSafe calls it the "exclusive" operation. ''' raise ValueError("Abstract Class member (IsRunningOn)") return None def IsWorkingCorrectly(self, nodename): ''' This member function returns true if our resource is operating correctly on the given node in the cluster. Heartbeat does not require this operation, but it might be called the Monitor operation, which is what FailSafe calls it. For remotely monitorable resources (like IP addresses), they *should* be monitored remotely for testing. ''' raise ValueError("Abstract Class member (IsWorkingCorrectly)") return None def Start(self, nodename): ''' This member function starts or activates the resource. ''' raise ValueError("Abstract Class member (Start)") return None def Stop(self, nodename): ''' This member function stops or deactivates the resource. ''' raise ValueError("Abstract Class member (Stop)") return None def __repr__(self): if (self.Instance and len(self.Instance) > 1): return "{" + self.ResourceType + "::" + self.Instance + "}" else: return "{" + self.ResourceType + "}" class Component: def kill(self, node): None class Process(Component): def __init__(self, name, cm): self.name = str(name) self.CM = cm self.KillCmd = "killall -9 " + self.name def kill(self, node): if not self.CM.ns.WaitForNodeToComeUp(node): self.CM.log("WaitForNodeToComeUp %s filed in kill" % node) return None if self.CM.rsh(node, self.KillCmd) != 0: self.log ("Warn: Kill %s failed on node %s" %(name,node)) return None return 1 class ScenarioComponent: 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: ( '''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, Components): "Initialize the Scenario from the list of ScenarioComponents" for comp in Components: if not issubclass(comp.__class__, ScenarioComponent): raise ValueError("Init value must be subclass of" " ScenarioComponent") self.Components = Components def IsApplicable(self): ( '''A Scenario IsApplicable() iff each of its ScenarioComponents IsApplicable() ''' ) for comp in self.Components: if not comp.IsApplicable(): return None return 1 def SetUp(self, CM): '''Set up the Scenario. Return TRUE on success.''' j=0 while j < len(self.Components): if not self.Components[j].SetUp(CM): # OOPS! We failed. Tear partial setups down. CM.log("Tearing down partial setup") self.TearDown(CM, j) return None j=j+1 return 1 def TearDown(self, CM, 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(CM) j=j-1 class InitClusterManager(ScenarioComponent): ( '''InitClusterManager 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): '''InitClusterManager is so generic it is always Applicable''' return 1 def SetUp(self, CM): '''Basic Cluster Manager startup. Start everything''' CM.prepare() # Clear out the cobwebs ;-) self.TearDown(CM) for node in CM.Env["nodes"]: CM.rsh(node, CM["DelFileCommCmd"]+ "; true") # Now start the Cluster Manager on all the nodes. CM.log("Starting Cluster Manager on all nodes.") CM.startall() return 1 def TearDown(self, CM): '''Set up the given ScenarioComponent''' # Stop the cluster manager everywhere CM.log("Stopping Cluster Manager on all nodes") CM.stopall() 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 1 def SetUp(self, CM): '''Start the PingFest!''' self.PingSize=1024 if CM.Env.has_key("PingSize"): 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 1 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): for node in CM.Env["nodes"]: CM.log("Starting BSC node") if not CM.StartaCM(node): return 0 return 1 def TearDown(self, CM): for node in CM.Env["nodes"]: CM.log("Shutting down BSC node") if not CM.StopaCM(node): return 0 return 1