No OneTemporary
Actions

Size

85 KB

Referenced Files

None

Subscribers

None

View Options

	diff --git a/cts/CIB.py b/cts/CIB.py
	index 8fbae393eb..d212e52d1f 100644
	--- a/cts/CIB.py
	+++ b/cts/CIB.py
	@@ -1,565 +1,561 @@
	'''CTS: Cluster Testing System: CIB generator
	'''
	__copyright__='''
	Author: Andrew Beekhof <abeekhof@suse.de>
	Copyright (C) 2008 Andrew Beekhof
	'''

	from UserDict import UserDict
	import sys, time, types, syslog, os, struct, string, signal, traceback, warnings

	from cts.CTSvars import *
	from cts.CTS import ClusterManager

	class CibBase:
	cts_cib = None
	cib_tmpfile = None
	version = "unknown"
	feature_set = "unknown"
	target = None

	def __init__(self, CM, tmpfile=None):
	self.CM = CM
	#self.target = self.CM.Env["nodes"][0]

	if not tmpfile:
	warnings.filterwarnings("ignore")
	self.cib_tmpfile=os.tmpnam()
	warnings.resetwarnings()
	else:
	self.cib_tmpfile = tmpfile

	def version(self):
	return self.version

	def NextIP(self):
	fields = string.split(self.CM.Env["IPBase"], '.')
	fields[3] = str(int(fields[3])+1)
	ip = string.join(fields, '.')
	self.CM.Env["IPBase"] = ip
	return ip

	class CIB06(CibBase):
	version = "transitional-0.6"
	coloc_template = """<rsc_colocation id="%s" from="%s" to="%s" to_role="%s" score="%s"/>"""

	cib_template ='''
	<cib admin_epoch="1" epoch="0" num_updates="0" remote_access_port="9898">
	<configuration>
	<crm_config> %s
	</crm_config>
	<nodes/>
	<resources> %s
	</resources>
	<constraints> %s
	</constraints>
	</configuration>
	<status/>
	</cib> '''

	cib_option_template = '''
	<cluster_property_set id="cib-bootstrap-options"><attributes>
	<nvpair id="cib-bootstrap-1" name="start-failure-is-fatal" value="false"/>
	<nvpair id="cib-bootstrap-2" name="stonith-enabled" value="%d"/>
	<nvpair id="cib-bootstrap-3" name="pe-input-series-max" value="30000"/>
	<nvpair id="cib-bootstrap-4" name="shutdown-escalation" value="5min"/>
	<nvpair id="cib-bootstrap-5" name="startup-fencing" value="false"/>
	<nvpair id="cib-bootstrap-6" name="batch-limit" value="10"/>
	<nvpair id="cib-bootstrap-7" name="no-quorum-policy" value="%s"/>
	</attributes></cluster_property_set>'''

	lsb_resource = '''
	<primitive id="lsb_dummy" class="lsb" type="''' +CTSvars.CTS_home+ '''/LSBDummy">
	<operations>
	<op id="ocf_lsb_monitor" name="monitor" interval="5s"/>
	</operations>
	</primitive> '''

	clustermon_location_constraint = '''
	<rsc_location id="run_cluster_mon" rsc="cluster_mon">
	<rule id="cant_run_cluster_mon" score="-INFINITY" boolean_op="and">
	<expression id="mon_expr" attribute="#is_dc" operation="eq" value="false"/>
	</rule>
	</rsc_location> '''

	resource_group_template = '''<group id="group-1">%s %s %s</group>'''

	per_node_constraint_template = '''
	<rsc_location id="preferred-%s" rsc="%s" node="%s" score="100"/>'''

	pingd_constraint_template = '''
	<rsc_location id="%s-is-connected" rsc="%s">
	<rule id="%s-connected-rule" role="%s" score="-INFINITY">
	<expression id="%s-connected-expr" attribute="connected" operation="lt" value="%d"/>
	</rule>
	</rsc_location>'''

	dummy_resource_template = '''
	<primitive id="%s" class="ocf" type="Dummy" provider="heartbeat">
	<operations>
	<op id="mon-%s" name="monitor" interval="P10S"/>
	</operations>
	<instance_attributes id="%s-attrs"><attributes>
	<nvpair id="migrate-%s" name="allow_migrate" value="1"/>
	</attributes></instance_attributes>
	</primitive> '''

	clustermon_resource_template = '''
	<primitive id="cluster_mon" class="ocf" type="ClusterMon" provider="heartbeat">
	<operations>
	<op id="cluster_mon-1" name="monitor" interval="5s" prereq="nothing"/>
	<op id="cluster_mon-2" name="start" prereq="nothing"/>
	</operations>
	<instance_attributes id="cluster_mon-attrs">
	<attributes>
	<nvpair id="cluster_mon-1" name="htmlfile" value="/suse/abeekhof/Export/cluster.html"/>
	<nvpair id="cluster_mon-2" name="update" value="10"/>
	<nvpair id="cluster_mon-3" name="extra_options" value="-n -r"/>
	<nvpair id="cluster_mon-4" name="user" value="abeekhof"/>
	</attributes>
	</instance_attributes>
	</primitive> '''

	master_slave_resource = '''
	<master_slave id="master-1">
	<instance_attributes id="master_rsc">
	<attributes>
	<nvpair id="clone_max_1" name="clone_max" value="%d"/>
	<nvpair id="clone_node_max_2" name="clone_node_max" value="%d"/>
	<nvpair id="master_max_3" name="master_max" value="%d"/>
	<nvpair id="master_node_max_4" name="master_node_max" value="%d"/>
	</attributes>
	</instance_attributes>
	<primitive id="ocf_msdummy" class="ocf" type="Stateful" provider="heartbeat">
	<operations>
	<op id="ocf_msdummy_monitor" name="monitor" interval="15s"/>
	<op id="ocf_msdummy_monitor_master" name="monitor" interval="16s" role="Master"/>
	</operations>
	</primitive>
	</master_slave>'''

	pingd_resource_template = """
	<clone id="Connectivity">
	<meta_attributes id="pingd-opts">
	<attributes>
	<nvpair id="pingd-opt-1" name="globally_unique" value="false"/>
	</attributes>
	</meta_attributes>
	<primitive id="pingd" class="ocf" provider="pacemaker" type="pingd">
	<operations>
	<op id="pingd-op-1" name="monitor" interval="120s"/>
	</operations>
	<instance_attributes id="pingd-attrs">
	<attributes>
	<nvpair id="pingd-attr-1" name="host_list" value="%s"/>
	<nvpair id="pingd-attr-2" name="name" value="connected"/>
	</attributes>
	</instance_attributes>
	</primitive>
	</clone>"""

	stonith_resource_template = """
	<clone id="DoFencing">
	<meta_attributes id="fencing">
	<attributes>
	<nvpair id="DoFencing-attr-1" name="resource_failure_stickiness" value="-1"/>
	<nvpair id="DoFencing-attr-2" name="globally_unique" value="false"/>
	</attributes>
	</meta_attributes>
	<primitive id="child_DoFencing" class="stonith" type="%s">
	<operations>
	<op id="DoFencing-op-1" name="monitor" interval="120s" prereq="nothing" timeout="300s"/>
	<op id="DoFencing-op-2" name="start" prereq="nothing" timeout="180s"/>
	<op id="DoFencing-op-3" name="stop" timeout="180s"/>
	</operations>
	<instance_attributes id="fencing-child">
	<attributes>
	<nvpair id="child_DoFencing-1" name="%s" value="%s"/>
	<nvpair id="child_DoFencing-2" name="livedangerously" value="yes"/>
	</attributes>
	</instance_attributes>
	</primitive>
	</clone>"""

	bsc_template = '''
	<cluster_property_set id="bsc-options">
	<attributes>
	<nvpair id="bsc-options-ident-string" name="ident-string" value="Linux-HA TEST configuration file - REMOVEME!!"/>
	</attributes>
	</cluster_property_set>'''

	def NewIP(self, name=None):
	template = '''
	<primitive id="%s" class="ocf" type="IPaddr" provider="heartbeat">
	<operations>
	<op id="mon-%s" name="monitor" interval="5s"/>
	</operations>
	<instance_attributes id="attrs-%s"><attributes>
	<nvpair id="netmask-%s" name="cidr_netmask" value="32"/>
	<nvpair id="ip-%s" name="ip" value="%s"/>
	</attributes></instance_attributes>
	</primitive> '''

	ip = self.NextIP()
	if not name:
	name = "r"+ip

	return template % (name, name, name, name, name, ip)

	def NewHBIP(self, name=None):
	template = '''
	<primitive id="%s" class="heartbeat" type="IPaddr">
	<operations>
	<op id="mon-%s" name="monitor" interval="5s"/>
	</operations>
	<instance_attributes id="attrs-%s"><attributes>
	<nvpair id="ip-%s" name="1" value="%s/32"/>
	</attributes></instance_attributes>
	</primitive> '''

	ip = self.NextIP()
	if not name:
	name = "r"+ip

	return template % (name, name, name, name, ip)

	def NewDummy(self, name):
	return self.dummy_resource_template % (name, name, name, name)

	def install(self, target):
	self.CM.rsh("localhost", "echo \'" + self.contents(target) + "\' > " + self.cib_tmpfile)
	rc = self.CM.rsh.cp(cib_file, "root@%s:%s/cib.xml" + (target, CTSvars.CRM_CONFIG_DIR))
	if rc != 0:
	raise ValueError("Can not copy %s to %s (%d)"%(self.cib_tmpfile, target, rc))

	self.CM.rsh(target, "chown "+CTSvars.CRM_DAEMON_USER+" "+CTSvars.CRM_CONFIG_DIR+"/cib.xml")
	self.CM.rsh("localhost", "rm -f "+self.cib_tmpfile)

	def contents(self, target=None):
	# fencing resource
	if self.cts_cib:
	return self.cts_cib

	nodelist = ""
	num_nodes = 0
	for node in self.CM.Env["nodes"]:
	nodelist += node + " "
	num_nodes = num_nodes + 1

	no_quorum = "stop"
	if num_nodes < 3:
	no_quorum = "ignore"
	self.CM.debug("Cluster only has %d nodes, ignoring quorum" % num_nodes)

	#make up crm config
	cib_options = self.cib_option_template % (self.CM.Env["DoFencing"], no_quorum)

	#create resources and their constraints
	resources = ""
	constraints = ""

	if self.CM.Env["DoBSC"] == 1:
	cib_options = cib_options + self.bsc_template

	if self.CM.Env["CIBResource"] != 1:
	# generate cib
	self.cts_cib = self.cib_template % (cib_options, resources, constraints)
	return self.cts_cib

	if self.CM.cluster_monitor == 1:
	resources += self.clustermon_resource_template
	constraints += self.clustermon_location_constraint

	ip1_rsc = self.NewIP()
	ip2_rsc = self.NewHBIP()
	ip3_rsc = self.NewIP()
	resources += self.resource_group_template % (ip1_rsc, ip2_rsc, ip3_rsc)

	# lsb resource
	resources += self.lsb_resource

	# Migrator
	resources += self.NewDummy("migrator")
	constraints += self.coloc_template % ("group-with-master", "group-1", "master-1", "Master", "INFINITY")
	constraints += self.coloc_template % ("lsb-with-group", "lsb_dummy", "group-1", "Started", "INFINITY")

	# per node resource
	for node in self.CM.Env["nodes"]:
	per_node_resources = self.NewIP("rsc_"+node)
	per_node_constraint = self.per_node_constraint_template % (node, "rsc_"+node, node)

	resources += per_node_resources
	constraints += per_node_constraint

	# Ping the test master
	resources += self.pingd_resource_template % os.uname()[1]

	# Require conectivity to run
	constraints += self.pingd_constraint_template % ("master-1", "master-1", "m", "Started", "m", 1)

	if self.CM.Env["DoFencing"]:
	p_name = None
	p_value = None
	entries = string.split(self.CM.Env["stonith-params"], ',')
	for entry in entries:
	(p_name, p_value) = string.split(entry, '=')
	if p_name == "hostlist" and p_value == "all":
	p_value = string.join(self.CM.Env["nodes"], " ")

	stonith_resource = self.stonith_resource_template % (self.CM.Env["stonith-type"], p_name, p_value)
	resources += stonith_resource

	#master slave resource
	resources += self.master_slave_resource % (num_nodes, 1, 1, 1)

	# generate cib
	self.cts_cib = self.cib_template % (cib_options, resources, constraints)
	return self.cts_cib


	class CIB10(CibBase):
	feature_set = "3.0"
	version = "pacemaker-1.0"
	cib_template = '''
	<cib crm_feature_set='%s' admin_epoch='1' epoch='0' num_updates='0' validate-with='%s' %s>
	<configuration>
	<crm_config/>
	<nodes/>
	<resources/>
	<constraints/>
	</configuration>
	<status/>
	</cib>'''

	def _create(self, command):
	fixed = "HOME=/root CIB_file="+self.cib_tmpfile+" crm --force configure " + command
	rc = self.CM.rsh(self.target, fixed)
	if rc != 0:
	self.CM.log("Configure call failed: "+fixed)
	sys.exit(1)

	def _show(self, command=""):
	output = ""
	(rc, result) = self.CM.rsh(self.target, "HOME=/root CIB_file="+self.cib_tmpfile+" crm configure show "+command, None, )
	for line in result:
	output += line
	self.CM.debug("Generated Config: "+line)
	return output

	def NewIP(self, name=None, standard="ocf:heartbeat"):
	ip = self.NextIP()
	if not name:
	name = "r"+ip

	if not standard:
	standard = ""
	else:
	standard += ":"

	self._create('''primitive %s %sIPaddr params ip=%s cidr_netmask=32 op monitor interval=5s'''
	% (name, standard, ip))
	return name

	def install(self, target):
	old = self.cib_tmpfile

	# Force a rebuild
	self.cts_cib = None

	self.cib_tmpfile = CTSvars.CRM_CONFIG_DIR+"/cib.xml"
	self.contents(target)
	self.CM.rsh(self.target, "chown "+CTSvars.CRM_DAEMON_USER+" "+self.cib_tmpfile)

	self.cib_tmpfile = old

	def contents(self, target=None):
	# fencing resource
	if self.cts_cib:
	return self.cts_cib

	if not target:
	self.target = self.CM.Env["nodes"][0]
	else:
	self.target = target

	cib_base = self.cib_template % (self.feature_set, self.version, ''' remote-tls-port='9898' remote-clear-port='9999' ''')
	self.CM.rsh(self.target, '''echo "%s" > %s''' % (cib_base, self.cib_tmpfile))
	#self.CM.rsh.cp(self.cib_tmpfile, "root@%s:%s" % (self.target, self.cib_tmpfile))

	nodelist = ""
	self.num_nodes = 0
	for node in self.CM.Env["nodes"]:
	nodelist += node + " "
	self.num_nodes = self.num_nodes + 1

	no_quorum = "stop"
	if self.num_nodes < 3:
	no_quorum = "ignore"
	self.CM.debug("Cluster only has %d nodes, ignoring quorum" % self.num_nodes)


	# The shell no longer functions when the lrmd isn't running, how wonderful
	# Start one here and let the cluster clean it up when the full stack starts
	# Just hope target has the same location for lrmd
	self.CM.rsh(self.target, CTSvars.CRM_DAEMON_DIR+"/lrmd", synchronous=0)

	# Tell the shell to mind its own business, we know what we're doing
	self.CM.rsh(self.target, "crm options check-mode relaxed")

	# Now stop the shell from rejecting every update because we've not defined stonith resources yet
	self._create('''property stonith-enabled=false''')

	self._create('''property start-failure-is-fatal=false pe-input-series-max=5000''')
	self._create('''property shutdown-escalation=5min startup-fencing=false batch-limit=10 dc-deadtime=5s''')
	self._create('''property no-quorum-policy=%s expected-quorum-votes=%d''' % (no_quorum, self.num_nodes))

	if self.CM.Env["DoBSC"] == 1:
	self._create('''property ident-string="Linux-HA TEST configuration file - REMOVEME!!"''')

	# Add resources?
	if self.CM.Env["CIBResource"] == 1:
	self.add_resources()

	# Fencing resource
	if self.CM.Env["DoFencing"]:
	params = None
	entries = string.split(self.CM.Env["stonith-params"], ',')
	for entry in entries:
	(name, value) = string.split(entry, '=')
	if name == "hostlist" and value == "all":
	value = string.join(self.CM.Env["nodes"], " ")

	if params:
	params = ("""%s '%s="%s"' """ % (params, name, value))
	else:
	params = ("""'%s="%s"' """ % (name, value))

	if params:
	params = "params %s" % params
	else:
	params = ""

	self._create('''primitive FencingChild stonith::%s %s op monitor interval=120s timeout=300 op start interval=0 timeout=180s op stop interval=0 timeout=180s''' % (self.CM.Env["stonith-type"], params))
	# Set a threshold for unreliable stonith devices such as the vmware one
	self._create('''clone Fencing FencingChild meta globally-unique=false migration-threshold=5''')

	if self.CM.cluster_monitor == 1:
	self._create('''primitive cluster_mon ocf:pacemaker:ClusterMon params update=10 extra_options="-r -n" user=abeekhof htmlfile=/suse/abeekhof/Export/cluster.html op start interval=0 requires=nothing op monitor interval=5s requires=nothing''')
	self._create('''location prefer-dc cluster_mon rule -INFINITY: \#is_dc eq false''')

	self._create('''property stonith-enabled=%s''' % (self.CM.Env["DoFencing"]))

	# generate cib
	self.cts_cib = self._show("xml")

	if self.cib_tmpfile != CTSvars.CRM_CONFIG_DIR+"/cib.xml":
	self.CM.rsh(self.target, "rm -f "+self.cib_tmpfile)

	return self.cts_cib

	def add_resources(self):
	# Group Resource
	r1 = self.NewIP()
	ip = self.NextIP()
	r2 = "r"+ip
	self._create('''primitive %s heartbeat::IPaddr params 1=%s/32 op monitor interval=5s''' % (r2, ip))
	r3 = self.NewIP()
	self._create('''group group-1 %s %s %s''' % (r1, r2, r3))

	# Per-node resources
	for node in self.CM.Env["nodes"]:
	r = self.NewIP("rsc_"+node)
	self._create('''location prefer-%s %s rule 100: \#uname eq %s''' % (node, r, node))

	# LSB resource
	- lsb_agent=CTSvars.CTS_home+'''/LSBDummy'''
	- self.CM.log("Installing LSB agent %s on %s" % (lsb_agent, repr(self.CM.Env["nodes"])))
	- for node in self.CM.Env["nodes"]:
	- self.CM.rsh(node, "mkdir -p %s" % CTSvars.CTS_home)
	- self.CM.rsh.cp(lsb_agent, "root@%s:%s" % (node, lsb_agent))
	+ lsb_agent = self.CM.install_helper("LSBDummy")

	self._create('''primitive lsb-dummy lsb::''' +lsb_agent+ ''' op monitor interval=5s''')
	self._create('''colocation lsb-with-group INFINITY: lsb-dummy group-1''')
	self._create('''order lsb-after-group mandatory: group-1 lsb-dummy symmetrical=true''')

	# Migrator
	self._create('''primitive migrator ocf:pacemaker:Dummy meta allow-migrate=1 op monitor interval=P10S''')

	# Ping the test master
	self._create('''primitive ping-1 ocf:pacemaker:ping params host_list=%s name=connected debug=true op monitor interval=120s''' % os.uname()[1])
	self._create('''clone Connectivity ping-1 meta globally-unique=false''')

	#master slave resource
	self._create('''primitive stateful-1 ocf:pacemaker:Stateful op monitor interval=15s op monitor interval=16s role=Master''')
	self._create('''ms master-1 stateful-1 meta clone-max=%d clone-node-max=%d master-max=%d master-node-max=%d'''
	% (self.num_nodes, 1, 1, 1))

	# Require conectivity to run the master
	self._create('''location %s-is-connected %s rule -INFINITY: connected lt %d or not_defined connected''' % ("m1", "master-1", 1))

	# Group with the master
	self._create('''colocation group-with-master INFINITY: group-1 master-1:Master''')
	self._create('''order group-after-master mandatory: master-1:promote group-1:start symmetrical=true''')

	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:
	def __init__(self, CM):
	self.CM = CM
	self.register("pacemaker06", CIB06, CM)
	self.register("pacemaker10", CIB10, CM)
	self.register("hae", HASI, CM)


	def register(self, methodName, constructor, args, *kargs):
	"""register a constructor"""
	_args = [constructor]
	_args.extend(args)
	setattr(self, methodName, apply(ConfigFactoryItem,_args, kargs))

	def unregister(self, methodName):
	"""unregister a constructor"""
	delattr(self, methodName)

	def createConfig(self, name="pacemaker-1.0"):
	if name == "pacemaker-0.6":
	name = "pacemaker06";
	elif name == "pacemaker-1.0":
	name = "pacemaker10";
	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.pacemaker10()

	class ConfigFactoryItem:
	def __init__(self, function, args, *kargs):
	assert callable(function), "function should be a callable obj"
	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 apply(self._function,_args,_kargs)


	#CibFactory = ConfigFactory()
	diff --git a/cts/CTS.py b/cts/CTS.py
	index a107cb5ad1..feed0c98f2 100644
	--- a/cts/CTS.py
	+++ b/cts/CTS.py
	@@ -1,1353 +1,1364 @@
	'''CTS: Cluster Testing System: Main module

	Classes related to testing high-availability clusters...
	'''

	__copyright__='''
	Copyright (C) 2000, 2001 Alan Robertson <alanr@unix.sh>
	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, signal
	import time, syslog, random, traceback, base64, pickle, binascii, fcntl

	from socket import gethostbyname_ex
	from UserDict import UserDict
	from subprocess import Popen,PIPE
	from cts.CTSvars import *

	class CtsLab(UserDict):
	'''This class defines the Lab Environment for the Cluster Test System.
	It defines those things which are expected to change from test
	environment to test environment for the same cluster manager.

	It is where you define the set of nodes that are in your test lab
	what kind of reset mechanism you use, etc.

	This class is derived from a UserDict because we hold many
	different parameters of different kinds, and this provides
	provide a uniform and extensible interface useful for any kind of
	communication between the user/administrator/tester and CTS.

	At this point in time, it is the intent of this class to model static
	configuration and/or environmental data about the environment which
	doesn't change as the tests proceed.

	Well-known names (keys) are an important concept in this class.
	The HasMinimalKeys member function knows the minimal set of
	well-known names for the class.

	The following names are standard (well-known) at this time:

	nodes An array of the nodes in the cluster
	reset A ResetMechanism object
	logger An array of objects that log strings...
	CMclass The type of ClusterManager we are running
	(This is a class object, not a class instance)
	RandSeed Random seed. It is a triple of bytes. (optional)

	The CTS code ignores names it doesn't know about/need.
	The individual tests have access to this information, and it is
	perfectly acceptable to provide hints, tweaks, fine-tuning
	directions or other information to the tests through this mechanism.
	'''

	def __init__(self):
	self.data = {}
	self.rsh = RemoteExec(self)
	self.RandomGen = random.Random()
	self.Scenario = None

	# Get a random seed for the random number generator.
	self["LogWatcher"] = "any"
	self["LogFileName"] = "/var/log/messages"
	self["OutputFile"] = None
	self["SyslogFacility"] = None
	self["CMclass"] = None
	self["logrestartcmd"] = "/etc/init.d/syslog-ng restart 2>&1 > /dev/null"
	self["logger"] = ([StdErrLog(self)])

	self.SeedRandom()

	def SeedRandom(self, seed=None):
	if not seed:
	seed = int(time.time())

	if self.has_key("RandSeed"):
	self.log("New random seed is: " + str(seed))
	else:
	self.log("Random seed is: " + str(seed))

	self["RandSeed"] = seed
	self.RandomGen.seed(str(seed))

	def HasMinimalKeys(self):
	'Return TRUE if our object has the minimal set of keys/values in it'
	result = 1
	for key in self.MinimalKeys:
	if not self.has_key(key):
	result = None
	return result

	def log(self, args):
	"Log using each of the supplied logging methods"
	for logfcn in self._logfunctions:
	logfcn(string.strip(args))

	def debug(self, args):
	"Log using each of the supplied logging methods"
	for logfcn in self._logfunctions:
	if logfcn.name() != "StdErrLog":
	logfcn("debug: %s" % string.strip(args))

	def dump(self):
	keys = []
	for key in self.keys():
	keys.append(key)

	keys.sort()
	for key in keys:
	self.debug("Environment["+key+"]:\t"+str(self[key]))

	def run(self, Scenario, Iterations):
	if not Scenario:
	self.log("No scenario was defined")
	return 1

	self.log("Cluster nodes: ")
	for node in self["nodes"]:
	self.log(" * %s" % (node))

	if not Scenario.SetUp():
	return 1

	try :
	Scenario.run(Iterations)
	except :
	self.log("Exception by %s" % sys.exc_info()[0])
	for logmethod in self["logger"]:
	traceback.print_exc(50, logmethod)

	Scenario.summarize()
	Scenario.TearDown()
	return 1

	#ClusterManager.oprofileSave(Iterations)
	Scenario.TearDown()

	Scenario.summarize()
	if Scenario.Stats["failure"] > 0:
	return Scenario.Stats["failure"]

	elif Scenario.Stats["success"] != Iterations:
	self.log("No failure count but success != requested iterations")
	return 1

	return 0

	def __setitem__(self, key, value):
	'''Since this function gets called whenever we modify the
	dictionary (object), we can (and do) validate those keys that we
	know how to validate. For the most part, we know how to validate
	the "MinimalKeys" elements.
	'''

	#
	# List of nodes in the system
	#
	if key == "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:
	self.Nodes[node] = gethostbyname_ex(node)
	except:
	print node+" not found in DNS... aborting"
	raise
	#
	# List of Logging Mechanism(s)
	#
	elif key == "logger":
	if len(value) < 1:
	raise ValueError("Must have at least one logging mechanism")
	for logger in value:
	if not callable(logger):
	raise ValueError("'logger' elements must be callable")
	self._logfunctions = value
	#
	# Cluster Manager Class
	#
	elif key == "CMclass":
	if value and not issubclass(value, ClusterManager):
	raise ValueError("'CMclass' must be a subclass of"
	" ClusterManager")
	#
	# Initial Random seed...
	#
	#elif key == "RandSeed":
	# if len(value) != 3:
	# raise ValueError("'Randseed' must be a 3-element list/tuple")
	# for elem in value:
	# if not isinstance(elem, types.IntType):
	# raise ValueError("'Randseed' list must all be ints")

	self.data[key] = value

	def IsValidNode(self, node):
	'Return TRUE if the given node is valid'
	return self.Nodes.has_key(node)

	def __CheckNode(self, node):
	"Raise a ValueError if the given node isn't valid"

	if not self.IsValidNode(node):
	raise ValueError("Invalid node [%s] in CheckNode" % node)

	def RandomNode(self):
	'''Choose a random node from the cluster'''
	return self.RandomGen.choice(self["nodes"])

	class Logger:
	TimeFormat = "%b %d %H:%M:%S\t"

	def __call__(self, lines):
	raise ValueError("Abstract class member (__call__)")
	def write(self, line):
	return self(line.rstrip())
	def writelines(self, lines):
	for s in lines:
	self.write(s)
	return 1
	def flush(self):
	return 1
	def isatty(self):
	return None

	class SysLog(Logger):
	# http://docs.python.org/lib/module-syslog.html
	defaultsource="CTS"
	map = {
	"kernel": syslog.LOG_KERN,
	"user": syslog.LOG_USER,
	"mail": syslog.LOG_MAIL,
	"daemon": syslog.LOG_DAEMON,
	"auth": syslog.LOG_AUTH,
	"lpr": syslog.LOG_LPR,
	"news": syslog.LOG_NEWS,
	"uucp": syslog.LOG_UUCP,
	"cron": syslog.LOG_CRON,
	"local0": syslog.LOG_LOCAL0,
	"local1": syslog.LOG_LOCAL1,
	"local2": syslog.LOG_LOCAL2,
	"local3": syslog.LOG_LOCAL3,
	"local4": syslog.LOG_LOCAL4,
	"local5": syslog.LOG_LOCAL5,
	"local6": syslog.LOG_LOCAL6,
	"local7": syslog.LOG_LOCAL7,
	}
	def __init__(self, labinfo):

	if labinfo.has_key("syslogsource"):
	self.source=labinfo["syslogsource"]
	else:
	self.source=SysLog.defaultsource

	self.facility="daemon"

	if labinfo.has_key("SyslogFacility") and labinfo["SyslogFacility"]:
	if SysLog.map.has_key(labinfo["SyslogFacility"]):
	self.facility=labinfo["SyslogFacility"]
	else:
	raise ValueError("%s: bad syslog facility"%labinfo["SyslogFacility"])

	self.facility=SysLog.map[self.facility]
	syslog.openlog(self.source, 0, self.facility)

	def setfacility(self, facility):
	self.facility = facility
	if SysLog.map.has_key(self.facility):
	self.facility=SysLog.map[self.facility]
	syslog.closelog()
	syslog.openlog(self.source, 0, self.facility)


	def __call__(self, lines):
	if isinstance(lines, types.StringType):
	syslog.syslog(lines)
	else:
	for line in lines:
	syslog.syslog(line)

	def name(self):
	return "Syslog"

	class StdErrLog(Logger):

	def __init__(self, labinfo):
	pass

	def __call__(self, lines):
	t = time.strftime(Logger.TimeFormat, time.localtime(time.time()))
	if isinstance(lines, types.StringType):
	sys.__stderr__.writelines([t, lines, "\n"])
	else:
	for line in lines:
	sys.__stderr__.writelines([t, line, "\n"])
	sys.__stderr__.flush()

	def name(self):
	return "StdErrLog"

	class FileLog(Logger):
	def __init__(self, labinfo, filename=None):

	if filename == None:
	filename=labinfo["LogFileName"]

	self.logfile=filename
	import os
	self.hostname = os.uname()[1]+" "
	self.source = "CTS: "
	def __call__(self, lines):

	fd = open(self.logfile, "a")
	t = time.strftime(Logger.TimeFormat, time.localtime(time.time()))

	if isinstance(lines, types.StringType):
	fd.writelines([t, self.hostname, self.source, lines, "\n"])
	else:
	for line in lines:
	fd.writelines([t, self.hostname, self.source, line, "\n"])
	fd.close()

	def name(self):
	return "FileLog"

	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, Env=None, silent=False):
	self.Env = Env
	self.silent = silent

	# -n: no stdin, -x: no X11,
	# -o ServerAliveInterval=5 disconnect after 3*5s if the server stops responding
	self.Command = "ssh -l root -n -x -o ServerAliveInterval=5 -o ConnectTimeout=10 -o TCPKeepAlive=yes -o ServerAliveCountMax=3 "
	# -B: batch mode, -q: no stats (quiet)
	self.CpCommand = "scp -B -q"

	self.OurNode=string.lower(os.uname()[1])

	def enable_qarsh(self):
	# http://nstraz.wordpress.com/2008/12/03/introducing-qarsh/
	self.log("Using QARSH for connections to cluster nodes")

	self.Command = "qarsh -l root"
	self.CpCommand = "qacp"

	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]

	#print "sysname: %s, us: %s" % (sysname, self.OurNode)
	if sysname == None or string.lower(sysname) == self.OurNode or sysname == "localhost":
	ret = command
	else:
	ret = self.Command + " " + sysname + " '" + self._fixcmd(command) + "'"
	#print ("About to run %s\n" % ret)
	return ret

	def log(self, args):
	if not self.silent:
	if not self.Env:
	print (args)
	else:
	self.Env.log(args)

	def debug(self, args):
	if not self.silent:
	if not self.Env:
	print (args)
	else:
	self.Env.debug(args)

	def __call__(self, node, command, stdout=0, synchronous=1, silent=False, blocking=True):
	'''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.
	'''

	rc = 0
	result = None
	if not synchronous:
	proc = Popen(self._cmd([node, command]),
	stdout = PIPE, stderr = PIPE, close_fds = True, shell = True)

	if not silent: self.debug("cmd: async: target=%s, rc=%d: %s" % (node, proc.pid, command))
	if proc.pid > 0:
	return 0
	return -1

	proc = Popen(self._cmd([node, command]),
	stdout = PIPE, stderr = PIPE, close_fds = True, shell = True)

	#if not blocking:
	# fcntl.fcntl(proc.stdout.fileno(), fcntl.F_SETFL, os.O_NONBLOCK)

	if proc.stdout:
	if stdout == 1:
	result = proc.stdout.readline()
	else:
	result = proc.stdout.readlines()
	proc.stdout.close()
	else:
	self.log("No stdout stream")

	rc = proc.wait()

	if not silent: self.debug("cmd: target=%s, rc=%d: %s" % (node, rc, command))

	if stdout == 1:
	return result

	if proc.stderr:
	errors = proc.stderr.readlines()
	proc.stderr.close()
	if not silent:
	for err in errors:
	self.debug("cmd: stderr: %s" % err)

	if stdout == 0:
	if not silent and result:
	for line in result:
	self.debug("cmd: stdout: %s" % line)
	return rc

	return (rc, result)

	def cp(self, source, target, silent=False):
	'''Perform a remote copy'''
	cpstring = self.CpCommand + " \'" + source + "\'" + " \'" + target + "\'"
	rc = os.system(cpstring)
	if not silent: self.debug("cmd: rc=%d: %s" % (rc, cpstring))

	return rc

	has_log_watcher = {}
	log_watcher_bin = "/tmp/cts_log_watcher.py"
	log_watcher = """
	import sys, os, fcntl

	'''
	Remote logfile reader for CTS
	Reads a specified number of lines from the supplied offset
	Returns the current offset

	Contains logic for handling truncation
	'''

	limit = 100
	offset = 0
	prefix = ''
	filename = '/var/log/messages'

	skipthis=None
	args=sys.argv[1:]
	for i in range(0, len(args)):
	if skipthis:
	skipthis=None
	continue

	elif args[i] == '-l' or args[i] == '--limit':
	skipthis=1
	limit = int(args[i+1])

	elif args[i] == '-f' or args[i] == '--filename':
	skipthis=1
	filename = args[i+1]

	elif args[i] == '-o' or args[i] == '--offset':
	skipthis=1
	offset = args[i+1]

	elif args[i] == '-p' or args[i] == '--prefix':
	skipthis=1
	prefix = args[i+1]

	logfile=open(filename, 'r')
	logfile.seek(0, os.SEEK_END)
	newsize=logfile.tell()

	if offset != 'EOF':
	offset = int(offset)
	if newsize >= offset:
	logfile.seek(offset)
	else:
	print prefix + ('File truncated from %d to %d' % (offset, newsize))
	if (newsize*1.05) < offset:
	logfile.seek(0)
	# else: we probably just lost a few logs after a fencing op
	# continue from the new end
	# TODO: accept a timestamp and discard all messages older than it

	# Don't block when we reach EOF
	fcntl.fcntl(logfile.fileno(), fcntl.F_SETFL, os.O_NONBLOCK)

	count = limit
	while count > 0:
	count -= 1
	line = logfile.readline()
	if line: print line.strip()
	else: break

	print prefix + 'Last read: %d %d' % (logfile.tell(), limit - count)
	logfile.close()
	"""

	class SearchObj:
	def __init__(self, Env, filename, host=None):

	self.Env = Env
	self.host = host
	self.filename = filename

	self.cache = []
	self.offset = "EOF"

	if host == None:
	host = "localhost"

	global has_log_watcher
	if not has_log_watcher.has_key(host):

	global log_watcher
	global log_watcher_bin
	self.debug("Installing %s on %s" % (log_watcher_bin, host))
	self.Env.rsh(host, '''echo "%s" > %s''' % (log_watcher, log_watcher_bin), silent=True)
	has_log_watcher[host] = 1

	self.next()

	def __str__(self):
	if self.host:
	return "%s:%s" % (self.host, self.filename)
	return self.filename

	def log(self, args):
	message = "lw: %s: %s" % (self, args)
	if not self.Env:
	print (message)
	else:
	self.Env.log(message)

	def debug(self, args):
	message = "lw: %s: %s" % (self, args)
	if not self.Env:
	print (message)
	else:
	self.Env.debug(message)

	def next(self):
	cache = []
	if not len(self.cache):
	global log_watcher_bin
	(rc, lines) = self.Env.rsh(
	self.host,
	"python %s -p CTSwatcher: -f %s -o %s" % (log_watcher_bin, self.filename, self.offset),
	stdout=None, silent=True, blocking=False)

	for line in lines:
	match = re.search("^CTSwatcher:Last read: (\d+)", line)
	if match:
	last_offset = self.offset
	self.offset = match.group(1)
	#if last_offset == "EOF": self.debug("Got %d lines, new offset: %s" % (len(lines), self.offset))

	elif re.search("^CTSwatcher:.*truncated", line):
	self.log(line)
	elif re.search("^CTSwatcher:", line):
	self.debug("Got control line: "+ line)
	else:
	cache.append(line)

	return cache

	class LogWatcher(RemoteExec):

	'''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, Env, log, regexes, name="Anon", timeout=10, debug_level=None, silent=False):
	'''This is the constructor for the LogWatcher class. It takes a
	log name to watch, and a list of regular expressions to watch for."
	'''
	RemoteExec.__init__(self, Env)

	# Validate our arguments. Better sooner than later ;-)
	for regex in regexes:
	assert re.compile(regex)

	self.name = name
	self.regexes = regexes
	self.filename = log
	self.debug_level = debug_level
	self.whichmatch = -1
	self.unmatched = None

	self.file_list = []
	self.line_cache = []

	if not silent:
	for regex in self.regexes:
	self.debug("Looking for regex: "+regex)

	self.Timeout = int(timeout)
	self.returnonlymatch = None

	def debug(self, args):
	message = "lw: %s: %s" % (self.name, args)
	if not self.Env:
	print (message)
	else:
	self.Env.debug(message)

	def setwatch(self):
	'''Mark the place to start watching the log from.
	'''

	if self.Env["LogWatcher"] == "remote":
	for node in self.Env["nodes"]:
	self.file_list.append(SearchObj(self.Env, self.filename, node))

	else:
	self.file_list.append(SearchObj(self.Env, self.filename))

	def __del__(self):
	if self.debug_level > 1: self.debug("Destroy")

	def ReturnOnlyMatch(self, onlymatch=1):
	'''Specify one or more subgroups of the match to return rather than the whole string
	http://www.python.org/doc/2.5.2/lib/match-objects.html
	'''
	self.returnonlymatch = onlymatch

	def __get_line(self):

	if not len(self.line_cache):
	if not len(self.file_list):
	raise ValueError("No sources to read from")

	for f in self.file_list:
	lines = f.next()
	self.line_cache.extend(lines)

	if self.line_cache:
	line = self.line_cache[0]
	self.line_cache.remove(line)
	return line

	return None

	def look(self, timeout=None, silent=False):
	'''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.
	'''
	if timeout == None: timeout = self.Timeout

	now=time.time()
	done=now+timeout+1
	if self.debug_level > 2: self.debug("starting single search: timeout=%d, now=%d, limit=%d" % (timeout, now, done))

	while (timeout <= 0 or time.time() <= done):
	line = self.__get_line()
	if line:
	which=-1
	if re.search("CTS:", line):
	continue
	if self.debug_level > 2: self.debug("Processing: "+ line)
	for regex in self.regexes:
	which=which+1
	if self.debug_level > 2: self.debug("Comparing line to: "+ regex)
	#matchobj = re.search(string.lower(regex), string.lower(line))
	matchobj = re.search(regex, line)
	if matchobj:
	self.whichmatch=which
	if self.returnonlymatch:
	return matchobj.group(self.returnonlymatch)
	else:
	self.debug("Matched: "+line)
	if self.debug_level > 1: self.debug("With: "+ regex)
	return line

	elif timeout > 0:
	time.sleep(1)
	#time.sleep(0.025)

	else:
	self.debug("End of file")
	return None

	self.debug("Single search timed out: timeout=%d, start=%d, limit=%d, now=%d" % (timeout, now, done, time.time()))
	return None

	def lookforall(self, timeout=None, allow_multiple_matches=None, silent=False):
	'''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 = []

	if not silent:
	self.debug("starting search: timeout=%d" % timeout)
	for regex in self.regexes:
	if self.debug_level > 2: self.debug("Looking for regex: "+regex)

	while (len(self.regexes) > 0):
	oneresult = self.look(timeout)
	if not oneresult:
	self.unmatched = self.regexes
	self.matched = returnresult
	self.regexes = save_regexes
	return None

	returnresult.append(oneresult)
	if not allow_multiple_matches:
	del self.regexes[self.whichmatch]

	else:
	# Allow multiple regexes to match a single line
	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)

	self.unmatched = None
	self.matched = returnresult
	self.regexes = save_regexes
	return returnresult

	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 self.Env.rsh("localhost", "ping -nq -c1 -w1 %s" % node, silent=True) == 0

	def IsSshdUp(self, node):
	rc = self.Env.rsh(node, "true", silent=True)
	return rc == 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) and self.IsSshdUp(node):
	if anytimeouts:
	# Fudge to wait for the system to finish coming up
	time.sleep(30)
	self.Env.debug("Node %s now up" % node)
	return 1

	time.sleep(30)
	if (not anytimeouts):
	self.Env.debug("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))
	answer = raw_input('Continue? [nY]')
	if answer and answer == "n":
	raise ValueError("%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.clear_cache = 0
	self.TestLoggingLevel=0
	self.data = {
	"up" : "up", # Status meaning up
	"down" : "down", # Status meaning down
	"StonithCmd" : "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,
	"BreakCommCmd" : None,
	"FixCommCmd" : None,
	#"TestConfigDir" : None,
	"LogFileName" : None,

	#"Pat:Master_started" : None,
	#"Pat:Slave_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 = self.Env.rsh
	self.ShouldBeStatus={}
	self.ns = NodeStatus(self.Env)
	self.OurNode=string.lower(os.uname()[1])

	def key_for_node(self, node):
	return node

	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]="up"
	else:
	self.ShouldBeStatus[node]="down"

	self.unisolate_node(node)

	def upcount(self):
	'''How many nodes are up?'''
	count=0
	for node in self.Env["nodes"]:
	if self.ShouldBeStatus[node]=="up":
	count=count+1
	return count

	+ def install_helper(self, filename, nodes=None):
	+ file_with_path="%s/%s" % (CTSvars.CTS_home, filename)
	+ if not nodes:
	+ nodes = self.Env["nodes"]
	+
	+ self.debug("Installing %s to %s on %s" % (filename, CTSvars.CTS_home, repr(self.Env["nodes"])))
	+ for node in nodes:
	+ self.rsh(node, "mkdir -p %s" % CTSvars.CTS_home)
	+ self.rsh.cp(file_with_path, "root@%s:%s" % (node, file_with_path))
	+ return file_with_path
	+
	def install_config(self, node):
	return None

	def clear_all_caches(self):
	if self.clear_cache:
	for node in self.Env["nodes"]:
	if self.ShouldBeStatus[node] == "down":
	self.debug("Removing cache file on: "+node)
	self.rsh(node, "rm -f "+CTSvars.HA_VARLIBHBDIR+"/hostcache")
	else:
	self.debug("NOT Removing cache file on: "+node)

	def StartaCM(self, node, verbose=False):

	'''Start up the cluster manager on a given node'''
	if verbose: self.log("Starting %s on node %s" %(self["Name"], node))
	else: self.debug("Starting %s on node %s" %(self["Name"], node))
	ret = 1

	if not self.ShouldBeStatus.has_key(node):
	self.ShouldBeStatus[node] = "down"

	if self.ShouldBeStatus[node] != "down":
	return 1

	patterns = []
	# Technically we should always be able to notice ourselves starting
	patterns.append(self["Pat:Local_started"] % node)
	if self.upcount() == 0:
	patterns.append(self["Pat:Master_started"] % node)
	else:
	patterns.append(self["Pat:Slave_started"] % node)

	watch = LogWatcher(
	self.Env, self["LogFileName"], patterns, "StartaCM", self["StartTime"]+10)

	watch.setwatch()

	self.install_config(node)

	self.ShouldBeStatus[node] = "any"
	if self.StataCM(node) and self.cluster_stable(self["DeadTime"]):
	self.log ("%s was already started" %(node))
	return 1

	# Clear out the host cache so autojoin can be exercised
	if self.clear_cache:
	self.debug("Removing cache file on: "+node)
	self.rsh(node, "rm -f "+CTSvars.HA_VARLIBHBDIR+"/hostcache")

	if not(self.Env["valgrind-tests"]):
	startCmd = self["StartCmd"]
	else:
	if self.Env["valgrind-prefix"]:
	prefix = self.Env["valgrind-prefix"]
	else:
	prefix = "cts"

	startCmd = """G_SLICE=always-malloc HA_VALGRIND_ENABLED='%s' VALGRIND_OPTS='%s --log-file=/tmp/%s-%s.valgrind' %s""" % (
	self.Env["valgrind-procs"], self.Env["valgrind-opts"], prefix, """%p""", self["StartCmd"])

	if self.rsh(node, startCmd) != 0:
	self.log ("Warn: Start command failed on node %s" %(node))
	return None

	self.ShouldBeStatus[node]="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, verbose=False):

	'''Start up the cluster manager on a given node with none-block mode'''

	if verbose: self.log("Starting %s on node %s" %(self["Name"], node))
	else: self.debug("Starting %s on node %s" %(self["Name"], node))

	# Clear out the host cache so autojoin can be exercised
	if self.clear_cache:
	self.debug("Removing cache file on: "+node)
	self.rsh(node, "rm -f "+CTSvars.HA_VARLIBHBDIR+"/hostcache")

	if not(self.Env["valgrind-tests"]):
	startCmd = self["StartCmd"]
	else:
	if self.Env["valgrind-prefix"]:
	prefix = self.Env["valgrind-prefix"]
	else:
	prefix = "cts"

	startCmd = """G_SLICE=always-malloc HA_VALGRIND_ENABLED='%s' VALGRIND_OPTS='%s --log-file=/tmp/%s-%s.valgrind' %s""" % (
	self.Env["valgrind-procs"], self.Env["valgrind-opts"], prefix, """%p""", self["StartCmd"])

	self.rsh(node, startCmd, synchronous=0)
	self.ShouldBeStatus[node]="up"
	return 1

	def StopaCM(self, node, verbose=False):

	'''Stop the cluster manager on a given node'''

	if verbose: self.log("Stopping %s on node %s" %(self["Name"], node))
	else: self.debug("Stopping %s on node %s" %(self["Name"], node))

	if self.ShouldBeStatus[node] != "up":
	return 1

	if self.rsh(node, self["StopCmd"]) == 0:
	self.ShouldBeStatus[node]="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))

	self.rsh(node, self["StopCmd"], synchronous=0)
	self.ShouldBeStatus[node]="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.
	'''
	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(node, self["StatusCmd"], 1)
	ret= (string.find(out, 'stopped') == -1)

	try:
	if ret:
	if self.ShouldBeStatus[node] == "down":
	self.log(
	"Node status for %s is %s but we think it should be %s"
	% (node, "up", self.ShouldBeStatus[node]))
	else:
	if self.ShouldBeStatus[node] == "up":
	self.log(
	"Node status for %s is %s but we think it should be %s"
	% (node, "down", self.ShouldBeStatus[node]))
	except KeyError: pass

	if ret: self.ShouldBeStatus[node]="up"
	else: self.ShouldBeStatus[node]="down"
	return ret

	def startall(self, nodelist=None, verbose=False):

	'''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.
	'''
	ret = 1
	map = {}
	if not nodelist:
	nodelist=self.Env["nodes"]
	for node in nodelist:
	if self.ShouldBeStatus[node] == "down":
	if not self.StartaCM(node, verbose=verbose):
	ret = 0
	return ret

	def stopall(self, nodelist=None, verbose=False):

	'''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.
	'''

	ret = 1
	map = {}
	if not nodelist:
	nodelist=self.Env["nodes"]
	for node in self.Env["nodes"]:
	if self.ShouldBeStatus[node] == "up":
	if not self.StopaCM(node, verbose=verbose):
	ret = 0
	return ret

	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] == "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] = "up"
	else:
	result[node] = "down"
	return result

	def isolate_node(self, target, nodes=None):
	'''isolate the communication between the nodes'''
	if not nodes:
	nodes = self.Env["nodes"]

	for node in nodes:
	if node != target:
	rc = self.rsh(target, self["BreakCommCmd"] % self.key_for_node(node))
	if rc != 0:
	self.log("Could not break the communication between %s and %s: %d" % (target, node, rc))
	return None
	else:
	self.debug("Communication cut between %s and %s" % (target, node))
	return 1

	def unisolate_node(self, target, nodes=None):
	'''fix the communication between the nodes'''
	if not nodes:
	nodes = self.Env["nodes"]

	for node in nodes:
	if node != target:
	restored = 0

	# Limit the amount of time we have asynchronous connectivity for
	# Restore both sides as simultaneously as possible
	self.rsh(target, self["FixCommCmd"] % self.key_for_node(node), synchronous=0)
	self.rsh(node, self["FixCommCmd"] % self.key_for_node(target), synchronous=0)
	self.debug("Communication restored between %s and %s" % (target, node))

	def reducecomm_node(self,node):
	'''reduce the communication between the nodes'''
	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 restorecomm_node(self,node):
	'''restore the saved communication between the nodes'''
	rc = 0
	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 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 oprofileStart(self, node=None):
	if not node:
	for n in self.Env["oprofile"]:
	self.oprofileStart(n)

	elif node in self.Env["oprofile"]:
	self.debug("Enabling oprofile on %s" % node)
	self.rsh(node, "opcontrol --init")
	self.rsh(node, "opcontrol --setup --no-vmlinux --separate=lib --callgraph=20 --image=all")
	self.rsh(node, "opcontrol --start")
	self.rsh(node, "opcontrol --reset")

	def oprofileSave(self, test, node=None):
	if not node:
	for n in self.Env["oprofile"]:
	self.oprofileSave(test, n)

	elif node in self.Env["oprofile"]:
	self.rsh(node, "opcontrol --dump")
	self.rsh(node, "opcontrol --save=cts.%d" % test)
	# Read back with: opreport -l session:cts.0 image:/usr/lib/heartbeat/c*
	if None:
	self.rsh(node, "opcontrol --reset")
	else:
	self.oprofileStop(node)
	self.oprofileStart(node)

	def oprofileStop(self, node=None):
	if not node:
	for n in self.Env["oprofile"]:
	self.oprofileStop(n)

	elif node in self.Env["oprofile"]:
	self.debug("Stopping oprofile on %s" % node)
	self.rsh(node, "opcontrol --reset")
	self.rsh(node, "opcontrol --shutdown 2>&1 > /dev/null")


	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, cm, name, process=None, dc_only=0, pats=[], dc_pats=[], badnews_ignore=[], triggersreboot=0):
	self.name = str(name)
	self.dc_only = dc_only
	self.pats = pats
	self.dc_pats = dc_pats
	self.CM = cm
	self.badnews_ignore = badnews_ignore
	self.triggersreboot = triggersreboot
	if process:
	self.proc = str(process)
	else:
	self.proc = str(name)
	self.KillCmd = "killall -9 " + self.proc

	def kill(self, node):
	if self.CM.rsh(node, self.KillCmd) != 0:
	self.CM.log ("ERROR: Kill %s failed on node %s" %(self.name,node))
	return None
	return 1

	diff --git a/cts/CTSscenarios.py b/cts/CTSscenarios.py
	index b4eafb9e4b..3ee38dbe8f 100644
	--- a/cts/CTSscenarios.py
	+++ b/cts/CTSscenarios.py
	@@ -1,539 +1,541 @@
	from CTS import *
	from CTStests import CTSTest
	from CTSaudits import ClusterAudit
	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, 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 1

	def SetUp(self):
	'''Set up the Scenario. Return TRUE on success.'''

	self.ClusterManager.prepare()
	self.ClusterManager.ns.WaitForAllNodesToComeUp(self.ClusterManager.Env["nodes"])

	self.audit()
	+ if self.ClusterManager.Env["valgrind-tests"]:
	+ self.ClusterManager.install_helper("cts.supp")

	self.BadNews = LogWatcher(self.ClusterManager.Env,
	self.ClusterManager["LogFileName"],
	self.ClusterManager["BadRegexes"], "BadNews", 0)
	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(self.ClusterManager, 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()

	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 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()

	ret = 1
	where = ""
	did_run = 0

	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")
	ret = 0

	stoptime=time.time()
	self.ClusterManager.oprofileSave(testcount)

	elapsed_time = stoptime - starttime
	test_time = stoptime - test.get_timer()
	if not test.has_key("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 incrimented 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 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())

	# 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:
	answer = raw_input('Big problems. Continue? [nY]')
	if answer and answer == "n":
	self.ClusterManager.log("Shutting down.")
	self.summarize()
	self.TearDown()
	raise ValueError("Looks like we hit a BadNews jackpot!")

	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)
	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)
	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 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)

	# Now start the Cluster Manager on all the nodes.
	CM.log("Starting Cluster Manager on all nodes.")
	return CM.startall(verbose=True)

	def TearDown(self, CM):
	'''Set up the given ScenarioComponent'''

	# Stop the cluster manager everywhere

	CM.log("Stopping Cluster Manager on all nodes")
	return CM.stopall(verbose=True)

	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):

	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)

	# 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 1

	def install(self, node, version):

	target_dir = "/tmp/rpm-%s" % version
	src_dir = "%s/%s" % (self.CM.Env["rpm-dir"], version)

	rc = self.CM.rsh(node, "mkdir -p %s" % target_dir)
	rc = self.CM.cp("%s/*.rpm %s:%s" % (src_dir, node, target_dir))
	rc = 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):
	CM.prepare()

	# Clear out the cobwebs
	CM.stopall()

	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

File Metadata

Mime Type: text/x-diff
Expires: Tue, Jul 8, 6:23 PM (11 h, 53 m)
Storage Engine: blob
Storage Format: Raw Data
Storage Handle: 2002623
Default Alt Text: (85 KB)

No OneTemporaryActions

View Options

File Metadata

Event Timeline

No OneTemporary
Actions