diff --git a/cts/CIB.py b/cts/CIB.py
index 82d02d77cc..1f1fea2779 100644
--- a/cts/CIB.py
+++ b/cts/CIB.py
@@ -1,452 +1,452 @@
'''CTS: Cluster Testing System: CIB generator
'''
__copyright__ = '''
Author: Andrew Beekhof <abeekhof@suse.de>
Copyright (C) 2008 Andrew Beekhof
'''
import os, string, warnings
from cts.CTSvars import *
from cts.CTS import ClusterManager
class CibBase:
def __init__(self, Factory, tag, _id, **kwargs):
self.tag = tag
self.name = _id
self.kwargs = kwargs
self.values = []
self.children = []
self.Factory = Factory
def __repr__(self):
return "%s-%s" % (self.tag, self.name)
def add_child(self, child):
self.children.append(child)
def __setitem__(self, key, value):
if value:
self.kwargs[key] = value
else:
self.values.append(key)
from cib_xml import *
class ConfigBase:
cts_cib = None
version = "unknown"
feature_set = "unknown"
Factory = None
def __init__(self, CM, factory, tmpfile=None):
self.CM = CM
self.Factory = factory
if not tmpfile:
warnings.filterwarnings("ignore")
tmpfile = os.tmpnam()
warnings.resetwarnings()
self.Factory.tmpfile = tmpfile
def version(self):
return self.version
def NextIP(self):
ip = self.CM.Env["IPBase"]
if ":" in ip:
(prefix, sep, suffix) = ip.rpartition(":")
suffix = str(hex(int(suffix, 16)+1)).lstrip("0x")
else:
(prefix, sep, suffix) = ip.rpartition(".")
suffix = str(int(suffix)+1)
ip = prefix + sep + suffix
self.CM.Env["IPBase"] = ip
return ip.strip()
class CIB11(ConfigBase):
feature_set = "3.0"
version = "pacemaker-1.1"
counter = 1
def _show(self, command=""):
output = ""
(rc, result) = self.Factory.rsh(self.Factory.target, "HOME=/root CIB_file="+self.Factory.tmpfile+" cibadmin -Ql "+command, None, )
for line in result:
output += line
self.Factory.debug("Generated Config: "+line)
return output
def NewIP(self, name=None, standard="ocf"):
if self.CM.Env["IPagent"] == "IPaddr2":
ip = self.NextIP()
if not name:
if ":" in ip:
(prefix, sep, suffix) = ip.rpartition(":")
name = "r"+suffix
else:
name = "r"+ip
r = Resource(self.Factory, name, self.CM.Env["IPagent"], standard)
r["ip"] = ip
if ":" in ip:
r["cidr_netmask"] = "64"
r["nic"] = "eth0"
else:
r["cidr_netmask"] = "32"
else:
if not name:
name = "r%s%d" % (self.CM.Env["IPagent"], self.counter)
self.counter = self.counter + 1
r = Resource(self.Factory, name, self.CM.Env["IPagent"], standard)
r.add_op("monitor", "5s")
return r
def install(self, target):
old = self.Factory.tmpfile
# Force a rebuild
self.cts_cib = None
self.Factory.tmpfile = CTSvars.CRM_CONFIG_DIR+"/cib.xml"
self.contents(target)
self.Factory.rsh(self.Factory.target, "chown "+CTSvars.CRM_DAEMON_USER+" "+self.Factory.tmpfile)
self.Factory.tmpfile = old
def contents(self, target=None):
# fencing resource
if self.cts_cib:
return self.cts_cib
if target:
self.Factory.target = target
self.Factory.rsh(self.Factory.target, "HOME=/root cibadmin --empty %s > %s" % (self.version, self.Factory.tmpfile))
#cib_base = self.cib_template % (self.feature_set, self.version, ''' remote-tls-port='9898' remote-clear-port='9999' ''')
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.Factory.log("Cluster only has %d nodes, configuring: no-quroum-policy=ignore" % self.num_nodes)
# Fencing resource
# Define first so that the shell doesn't reject every update
if self.CM.Env["DoFencing"]:
st = Resource(self.Factory, "Fencing", ""+self.CM.Env["stonith-type"], "stonith")
# Set a threshold for unreliable stonith devices such as the vmware one
st.add_meta("migration-threshold", "5")
st.add_op("monitor", "120s", timeout="120s")
st.add_op("stop", "0", timeout="60s")
st.add_op("start", "0", timeout="60s")
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"], " ")
elif name == "pcmk_host_list" and value == "all":
value = string.join(self.CM.Env["nodes"], " ")
st[name] = value
st.commit()
# Test advanced fencing logic
if True:
stf_nodes = []
stt_nodes = []
# Create the levels
stl = FencingTopology(self.Factory)
for node in self.CM.Env["nodes"]:
ftype = self.CM.Env.RandomGen.choice(["levels-and", "levels-or ", "broadcast "])
self.CM.log(" - Using %s fencing for node: %s" % (ftype, node))
# for baremetal remote node tests
stt_nodes.append("remote_%s" % node)
if ftype == "levels-and":
stl.level(1, node, "FencingPass,Fencing")
stt_nodes.append(node)
elif ftype == "levels-or ":
stl.level(1, node, "FencingFail")
stl.level(2, node, "Fencing")
stf_nodes.append(node)
# Create a Dummy agent that always passes for levels-and
if len(stt_nodes):
self.CM.install_helper("fence_dummy", destdir="/usr/sbin", sourcedir=CTSvars.Fencing_home)
stt = Resource(self.Factory, "FencingPass", "fence_dummy", "stonith")
stt["pcmk_host_list"] = string.join(stt_nodes, " ")
# Wait this many seconds before doing anything, handy for letting disks get flushed too
stt["random_sleep_range"] = "30"
stt["mode"] = "pass"
stt.commit()
# Create a Dummy agent that always fails for levels-or
if len(stf_nodes):
self.CM.install_helper("fence_dummy", destdir="/usr/sbin", sourcedir=CTSvars.Fencing_home)
stf = Resource(self.Factory, "FencingFail", "fence_dummy", "stonith")
stf["pcmk_host_list"] = string.join(stf_nodes, " ")
# Wait this many seconds before doing anything, handy for letting disks get flushed too
stf["random_sleep_range"] = "30"
stf["mode"] = "fail"
stf.commit()
# Now commit the levels themselves
stl.commit()
o = Option(self.Factory, "stonith-enabled", self.CM.Env["DoFencing"])
o["start-failure-is-fatal"] = "false"
o["pe-input-series-max"] = "5000"
o["default-action-timeout"] = "90s"
o["shutdown-escalation"] = "5min"
o["batch-limit"] = "10"
o["dc-deadtime"] = "5s"
o["no-quorum-policy"] = no_quorum
o["expected-quorum-votes"] = self.num_nodes
if self.CM.Env["DoBSC"] == 1:
o["ident-string"] = "Linux-HA TEST configuration file - REMOVEME!!"
o.commit()
# Add resources?
if self.CM.Env["CIBResource"] == 1:
self.add_resources()
if self.CM.cluster_monitor == 1:
mon = Resource(self.Factory, "cluster_mon", "ocf", "ClusterMon", "pacemaker")
mon.add_op("start", "0", requires="nothing")
mon.add_op("monitor", "5s", requires="nothing")
mon["update"] = "10"
mon["extra_options"] = "-r -n"
mon["user"] = "abeekhof"
mon["htmlfile"] = "/suse/abeekhof/Export/cluster.html"
mon.commit()
#self._create('''location prefer-dc cluster_mon rule -INFINITY: \#is_dc eq false''')
# generate cib
self.cts_cib = self._show()
if self.Factory.tmpfile != CTSvars.CRM_CONFIG_DIR+"/cib.xml":
self.Factory.rsh(self.Factory.target, "rm -f "+self.Factory.tmpfile)
return self.cts_cib
def add_resources(self):
# Per-node resources
for node in self.CM.Env["nodes"]:
name = "rsc_"+node
r = self.NewIP(name)
r.prefer(node, "100")
r.commit()
# Migrator
# Make this slightly sticky (since we have no other location constraints) to avoid relocation during Reattach
m = Resource(self.Factory, "migrator","Dummy", "ocf", "pacemaker")
m["passwd"] = "whatever"
m.add_meta("resource-stickiness","1")
m.add_meta("allow-migrate", "1")
m.add_op("monitor", "P10S")
m.commit()
# Ping the test master
p = Resource(self.Factory, "ping-1","ping", "ocf", "pacemaker")
p.add_op("monitor", "60s")
p["host_list"] = self.CM.Env["cts-master"]
p["name"] = "connected"
p["debug"] = "true"
c = Clone(self.Factory, "Connectivity", p)
c["globally-unique"] = "false"
c.commit()
#master slave resource
s = Resource(self.Factory, "stateful-1", "Stateful", "ocf", "pacemaker")
s.add_op("monitor", "15s", timeout="60s")
s.add_op("monitor", "16s", timeout="60s", role="Master")
ms = Master(self.Factory, "master-1", s)
ms["clone-max"] = self.num_nodes
ms["master-max"] = 1
ms["clone-node-max"] = 1
ms["master-node-max"] = 1
# Require conectivity to run the master
r = Rule(self.Factory, "connected", "-INFINITY", op="or")
r.add_child(Expression(self.Factory, "m1-connected-1", "connected", "lt", "1"))
r.add_child(Expression(self.Factory, "m1-connected-2", "connected", "not_defined", None))
ms.prefer("connected", rule=r)
ms.commit()
# Group Resource
g = Group(self.Factory, "group-1")
g.add_child(self.NewIP())
if self.CM.Env["have_systemd"]:
# It would be better to put the python in a separate file, so we
# could loop "while True" rather than sleep for 24 hours. We can't
# put a loop in a single-line python command; only simple commands
# may be separated by semicolon in python.
dummy_service_file = """
[Unit]
Description=Dummy resource that takes a while to start
[Service]
Type=notify
ExecStart=/usr/bin/python -c 'import time, systemd.daemon; time.sleep(10); systemd.daemon.notify("READY=1"); time.sleep(86400)'
ExecStop=/bin/sleep 10
ExecStop=/bin/kill -s KILL \$MAINPID
"""
os.system("cat <<-END >/tmp/DummySD.service\n%s\nEND" % (dummy_service_file))
self.CM.install_helper("DummySD.service", destdir="/usr/lib/systemd/system/", sourcedir="/tmp")
sysd = Resource(self.Factory, "petulant", "DummySD", "service")
sysd.add_op("monitor", "P10S")
g.add_child(sysd)
else:
g.add_child(self.NewIP())
g.add_child(self.NewIP())
# Group with the master
g.after("master-1", first="promote", then="start")
g.colocate("master-1", "INFINITY", withrole="Master")
g.commit()
# LSB resource
lsb_agent = self.CM.install_helper("LSBDummy")
lsb = Resource(self.Factory, "lsb-dummy",lsb_agent, "lsb")
lsb.add_op("monitor", "5s")
# LSB with group
lsb.after("group-1")
lsb.colocate("group-1")
lsb.commit()
class CIB12(CIB11):
feature_set = "3.0"
version = "pacemaker-1.2"
class CIB20(CIB11):
feature_set = "3.0"
version = "pacemaker-2.0"
#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.rsh = self.CM.rsh
self.register("pacemaker11", CIB11, CM, self)
self.register("pacemaker12", CIB12, CM, self)
self.register("pacemaker20", CIB20, CM, self)
# self.register("hae", HASI, CM, self)
self.target = self.CM.Env["nodes"][0]
self.tmpfile = None
def log(self, args):
self.CM.log("cib: %s" % args)
def debug(self, args):
self.CM.debug("cib: %s" % args)
def register(self, methodName, constructor, *args, **kargs):
"""register a constructor"""
_args = [constructor]
_args.extend(args)
setattr(self, methodName, apply(ConfigFactoryItem,_args, kargs))
def unregister(self, methodName):
"""unregister a constructor"""
delattr(self, methodName)
def createConfig(self, name="pacemaker-1.0"):
if name == "pacemaker-1.0":
name = "pacemaker10";
elif name == "pacemaker-1.1":
name = "pacemaker11";
elif name == "pacemaker-1.2":
name = "pacemaker12";
elif name == "pacemaker-2.0":
name = "pacemaker20";
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.pacemaker12()
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)
# Basic Sanity Testing
if __name__ == '__main__':
import CTSlab
env = CTSlab.LabEnvironment()
env["nodes"] = []
env["nodes"].append("pcmk-1")
env["nodes"].append("pcmk-2")
env["nodes"].append("pcmk-3")
env["nodes"].append("pcmk-4")
env["CIBResource"] = 1
env["IPBase"] = "fe80::1234:56:7890:1000"
env["DoStonith"] = 1
env["stonith-type"] = "fence_xvm"
env["stonith-params"] = "pcmk_arg_map=domain:uname"
manager = ClusterManager(env)
manager.cluster_monitor = False
CibFactory = ConfigFactory(manager)
cib = CibFactory.createConfig("pacemaker-1.1")
- print cib.contents()
+ print(cib.contents())
diff --git a/cts/CTS.py b/cts/CTS.py
index 9f9a291ef2..6d5fd09eb6 100644
--- a/cts/CTS.py
+++ b/cts/CTS.py
@@ -1,1001 +1,1001 @@
'''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., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
import string, sys, time, re, os, traceback
from UserDict import UserDict
from cts.CTSvars import *
from cts.logging import LogFactory
from cts.watcher import LogWatcher
from cts.remote import RemoteFactory
from cts.environment import EnvFactory
from cts.patterns import PatternSelector
has_log_stats = {}
log_stats_bin = CTSvars.CRM_DAEMON_DIR + "/cts_log_stats.sh"
log_stats = """
#!/bin/bash
# Tool for generating system load reports while CTS runs
trap "" 1
f=$1; shift
action=$1; shift
base=`basename $0`
if [ ! -e $f ]; then
echo "Time, Load 1, Load 5, Load 15, Test Marker" > $f
fi
function killpid() {
if [ -e $f.pid ]; then
kill -9 `cat $f.pid`
rm -f $f.pid
fi
}
function status() {
if [ -e $f.pid ]; then
kill -0 `cat $f.pid`
return $?
else
return 1
fi
}
function start() {
# Is it already running?
if
status
then
return
fi
echo Active as $$
echo $$ > $f.pid
while [ 1 = 1 ]; do
uptime | sed s/up.*:/,/ | tr '\\n' ',' >> $f
#top -b -c -n1 | grep -e usr/libexec/pacemaker | grep -v -e grep -e python | head -n 1 | sed s@/usr/libexec/pacemaker/@@ | awk '{print " 0, "$9", "$10", "$12}' | tr '\\n' ',' >> $f
echo 0 >> $f
sleep 5
done
}
case $action in
start)
start
;;
start-bg|bg)
# Use c --ssh -- ./stats.sh file start-bg
nohup $0 $f start >/dev/null 2>&1 </dev/null &
;;
stop)
killpid
;;
delete)
killpid
rm -f $f
;;
mark)
uptime | sed s/up.*:/,/ | tr '\\n' ',' >> $f
echo " $*" >> $f
start
;;
*)
echo "Unknown action: $action."
;;
esac
"""
class CtsLab:
'''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, args=None):
self.Env = EnvFactory().getInstance(args)
self.Scenario = None
self.logger = LogFactory()
self.rsh = RemoteFactory().getInstance()
def dump(self):
self.Env.dump()
def has_key(self, key):
return self.Env.has_key(key)
def __getitem__(self, key):
return self.Env[key]
def __setitem__(self, key, value):
self.Env[key] = value
def run(self, Scenario, Iterations):
if not Scenario:
self.logger.log("No scenario was defined")
return 1
self.logger.log("Cluster nodes: ")
for node in self.Env["nodes"]:
self.logger.log(" * %s" % (node))
if not Scenario.SetUp():
return 1
try :
Scenario.run(Iterations)
except :
self.logger.log("Exception by %s" % sys.exc_info()[0])
self.logger.traceback(traceback)
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.logger.log("No failure count but success != requested iterations")
return 1
return 0
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)
class NodeStatus:
def __init__(self, env):
self.Env = env
def IsNodeBooted(self, node):
'''Return TRUE if the given node is booted (responds to pings)'''
if self.Env["docker"]:
return RemoteFactory().getInstance()("localhost", "docker inspect --format {{.State.Running}} %s | grep -q true" % node, silent=True) == 0
return RemoteFactory().getInstance()("localhost", "ping -nq -c1 -w1 %s" % node, silent=True) == 0
def IsSshdUp(self, node):
rc = RemoteFactory().getInstance()(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)
LogFactory().debug("Node %s now up" % node)
return 1
time.sleep(30)
if (not anytimeouts):
LogFactory().debug("Waiting for node %s to come up" % node)
anytimeouts = 1
timeout = timeout - 1
LogFactory().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 = EnvFactory().getInstance()
self.templates = PatternSelector(self.Env["Name"])
self.__InitialConditions()
self.logger = LogFactory()
self.clear_cache = 0
self.TestLoggingLevel=0
self.data = {}
self.name = self.Env["Name"]
self.rsh = RemoteFactory().getInstance()
self.ShouldBeStatus={}
self.ns = NodeStatus(self.Env)
self.OurNode = string.lower(os.uname()[1])
self.__instance_errorstoignore = []
def __getitem__(self, key):
if key == "Name":
return self.name
- print "FIXME: Getting %s from %s" % (key, repr(self))
+ print("FIXME: Getting %s from %s" % (key, repr(self)))
if self.data.has_key(key):
return self.data[key]
return self.templates.get_patterns(self.Env["Name"], key)
def __setitem__(self, key, value):
- print "FIXME: Setting %s=%s on %s" % (key, value, repr(self))
+ print("FIXME: Setting %s=%s on %s" % (key, value, repr(self)))
self.data[key] = value
def key_for_node(self, node):
return node
def instance_errorstoignore_clear(self):
'''Allows the test scenario to reset instance errors to ignore on each iteration.'''
self.__instance_errorstoignore = []
def instance_errorstoignore(self):
'''Return list of errors which are 'normal' for a specific test instance'''
return self.__instance_errorstoignore
def errorstoignore(self):
'''Return list of errors which are 'normal' and should be ignored'''
return []
def log(self, args):
self.logger.log(args)
def debug(self, args):
self.logger.debug(args)
def prepare(self):
'''Finish the Initialization process. Prepare to test...'''
- print repr(self)+"prepare"
+ print(repr(self)+"prepare")
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, destdir=None, nodes=None, sourcedir=None):
if sourcedir == None:
sourcedir = CTSvars.CTS_home
file_with_path = "%s/%s" % (sourcedir, filename)
if not nodes:
nodes = self.Env["nodes"]
if not destdir:
destdir = CTSvars.CTS_home
self.debug("Installing %s to %s on %s" % (filename, destdir, repr(self.Env["nodes"])))
for node in nodes:
self.rsh(node, "mkdir -p %s" % destdir)
self.rsh.cp(file_with_path, "root@%s:%s/%s" % (node, destdir, filename))
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 prepare_fencing_watcher(self, name):
# If we don't have quorum now but get it as a result of starting this node,
# then a bunch of nodes might get fenced
upnode = None
if self.HasQuorum(None):
self.debug("Have quorum")
return None
if not self.templates["Pat:Fencing_start"]:
- print "No start pattern"
+ print("No start pattern")
return None
if not self.templates["Pat:Fencing_ok"]:
- print "No ok pattern"
+ print("No ok pattern")
return None
stonith = None
stonithPats = []
for peer in self.Env["nodes"]:
if self.ShouldBeStatus[peer] != "up":
stonithPats.append(self.templates["Pat:Fencing_ok"] % peer)
stonithPats.append(self.templates["Pat:Fencing_start"] % peer)
elif self.Env["Stack"] == "corosync (cman)":
# There is a delay between gaining quorum and CMAN starting fencing
# This can mean that even nodes that are fully up get fenced
# There is no use fighting it, just look for everyone so that CTS doesn't get confused
stonithPats.append(self.templates["Pat:Fencing_ok"] % peer)
stonithPats.append(self.templates["Pat:Fencing_start"] % peer)
stonith = LogWatcher(self.Env["LogFileName"], stonithPats, "StartupFencing", 0, hosts=self.Env["nodes"], kind=self.Env["LogWatcher"])
stonith.setwatch()
return stonith
def fencing_cleanup(self, node, stonith):
peer_list = []
peer_state = {}
self.debug("Looking for nodes that were fenced as a result of %s starting" % node)
# If we just started a node, we may now have quorum (and permission to fence)
if not stonith:
self.debug("Nothing to do")
return peer_list
q = self.HasQuorum(None)
if not q and len(self.Env["nodes"]) > 2:
# We didn't gain quorum - we shouldn't have shot anyone
self.debug("Quorum: %d Len: %d" % (q, len(self.Env["nodes"])))
return peer_list
for n in self.Env["nodes"]:
peer_state[n] = "unknown"
# Now see if any states need to be updated
self.debug("looking for: " + repr(stonith.regexes))
shot = stonith.look(0)
while shot:
line = repr(shot)
self.debug("Found: " + line)
del stonith.regexes[stonith.whichmatch]
# Extract node name
for n in self.Env["nodes"]:
if re.search(self.templates["Pat:Fencing_ok"] % n, shot):
peer = n
peer_state[peer] = "complete"
self.__instance_errorstoignore.append(self.templates["Pat:Fencing_ok"] % peer)
elif peer_state[n] != "complete" and re.search(self.templates["Pat:Fencing_start"] % n, shot):
# TODO: Correctly detect multiple fencing operations for the same host
peer = n
peer_state[peer] = "in-progress"
self.__instance_errorstoignore.append(self.templates["Pat:Fencing_start"] % peer)
if not peer:
self.logger.log("ERROR: Unknown stonith match: %s" % line)
elif not peer in peer_list:
self.debug("Found peer: " + peer)
peer_list.append(peer)
# Get the next one
shot = stonith.look(60)
for peer in peer_list:
self.debug(" Peer %s was fenced as a result of %s starting: %s" % (peer, node, peer_state[peer]))
if self.Env["at-boot"]:
self.ShouldBeStatus[peer] = "up"
else:
self.ShouldBeStatus[peer] = "down"
if peer_state[peer] == "in-progress":
# Wait for any in-progress operations to complete
shot = stonith.look(60)
while len(stonith.regexes) and shot:
line = repr(shot)
self.debug("Found: " + line)
del stonith.regexes[stonith.whichmatch]
shot = stonith.look(60)
# Now make sure the node is alive too
self.ns.WaitForNodeToComeUp(peer, self.Env["DeadTime"])
# Poll until it comes up
if self.Env["at-boot"]:
if not self.StataCM(peer):
time.sleep(self.Env["StartTime"])
if not self.StataCM(peer):
self.logger.log("ERROR: Peer %s failed to restart after being fenced" % peer)
return None
return peer_list
def StartaCM(self, node, verbose=False):
'''Start up the cluster manager on a given node'''
if verbose: self.logger.log("Starting %s on node %s" % (self.templates["Name"], node))
else: self.debug("Starting %s on node %s" % (self.templates["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.templates["Pat:Local_started"] % node)
if self.upcount() == 0:
patterns.append(self.templates["Pat:Master_started"] % node)
else:
patterns.append(self.templates["Pat:Slave_started"] % node)
watch = LogWatcher(
self.Env["LogFileName"], patterns, "StartaCM", self.Env["StartTime"]+10, hosts=self.Env["nodes"], kind=self.Env["LogWatcher"])
self.install_config(node)
self.ShouldBeStatus[node] = "any"
if self.StataCM(node) and self.cluster_stable(self.Env["DeadTime"]):
self.logger.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.templates["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.templates["StartCmd"])
stonith = self.prepare_fencing_watcher(node)
watch.setwatch()
if self.rsh(node, startCmd) != 0:
self.logger.log ("Warn: Start command failed on node %s" % (node))
self.fencing_cleanup(node, stonith)
return None
self.ShouldBeStatus[node] = "up"
watch_result = watch.lookforall()
if watch.unmatched:
for regex in watch.unmatched:
self.logger.log ("Warn: Startup pattern not found: %s" % (regex))
if watch_result and self.cluster_stable(self.Env["DeadTime"]):
#self.debug("Found match: "+ repr(watch_result))
self.fencing_cleanup(node, stonith)
return 1
elif self.StataCM(node) and self.cluster_stable(self.Env["DeadTime"]):
self.fencing_cleanup(node, stonith)
return 1
self.logger.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.logger.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")
self.install_config(node)
if not(self.Env["valgrind-tests"]):
startCmd = self.templates["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.templates["StartCmd"])
self.rsh(node, startCmd, synchronous=0)
self.ShouldBeStatus[node] = "up"
return 1
def StopaCM(self, node, verbose=False, force=False):
'''Stop the cluster manager on a given node'''
if verbose: self.logger.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" and force == False:
return 1
if self.rsh(node, self.templates["StopCmd"]) == 0:
# Make sure we can continue even if corosync leaks
# fdata-* is the old name
#self.rsh(node, "rm -f /dev/shm/qb-* /dev/shm/fdata-*")
self.ShouldBeStatus[node] = "down"
self.cluster_stable(self.Env["DeadTime"])
return 1
else:
self.logger.log ("ERROR: 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.templates["StopCmd"], synchronous=0)
self.ShouldBeStatus[node] = "down"
return 1
def cluster_stable(self, timeout = None):
time.sleep(self.Env["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.templates["RereadCmd"])
if rc == 0:
return 1
else:
self.logger.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.templates["StatusCmd"] % node, 1)
ret= (string.find(out, 'stopped') == -1)
try:
if ret:
if self.ShouldBeStatus[node] == "down":
self.logger.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.logger.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, quick=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.
'''
map = {}
if not nodelist:
nodelist = self.Env["nodes"]
for node in nodelist:
if self.ShouldBeStatus[node] == "down":
self.ns.WaitForAllNodesToComeUp(nodelist, 300)
if not quick:
if not self.StartaCM(node, verbose=verbose):
return 0
return 1
# Approximation of SimulStartList for --boot
watchpats = [ ]
watchpats.append(self.templates["Pat:DC_IDLE"])
for node in nodelist:
watchpats.append(self.templates["Pat:Local_started"] % node)
watchpats.append(self.templates["Pat:InfraUp"] % node)
watchpats.append(self.templates["Pat:PacemakerUp"] % node)
# Start all the nodes - at about the same time...
watch = LogWatcher(self.Env["LogFileName"], watchpats, "fast-start", self.Env["DeadTime"]+10, hosts=self.Env["nodes"], kind=self.Env["LogWatcher"])
watch.setwatch()
if not self.StartaCM(nodelist[0], verbose=verbose):
return 0
for node in nodelist:
self.StartaCMnoBlock(node, verbose=verbose)
watch.lookforall()
if watch.unmatched:
for regex in watch.unmatched:
self.logger.log ("Warn: Startup pattern not found: %s" % (regex))
if not self.cluster_stable():
self.logger.log("Cluster did not stabilize")
return 0
return 1
def stopall(self, nodelist=None, verbose=False, force=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" or force == True:
if not self.StopaCM(node, verbose=verbose, force=force):
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.templates["BreakCommCmd"] % self.key_for_node(node))
if rc != 0:
self.logger.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.templates["FixCommCmd"] % self.key_for_node(node), synchronous=0)
self.rsh(node, self.templates["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.templates["ReduceCommCmd"]%(self.Env["XmitLoss"],self.Env["RecvLoss"]))
if rc == 0:
return 1
else:
self.logger.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.templates["RestoreCommCmd"]);
if rc == 0:
return 1
else:
self.logger.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")
def StatsExtract(self):
if not self.Env["stats"]:
return
for host in self.Env["nodes"]:
log_stats_file = "%s/cts-stats.csv" % CTSvars.CRM_DAEMON_DIR
if has_log_stats.has_key(host):
self.rsh(host, '''bash %s %s stop''' % (log_stats_bin, log_stats_file))
(rc, lines) = self.rsh(host, '''cat %s''' % log_stats_file, stdout=2)
self.rsh(host, '''bash %s %s delete''' % (log_stats_bin, log_stats_file))
fname = "cts-stats-%d-nodes-%s.csv" % (len(self.Env["nodes"]), host)
- print "Extracted stats: %s" % fname
+ print("Extracted stats: %s" % fname)
fd = open(fname, "a")
fd.writelines(lines)
fd.close()
def StatsMark(self, testnum):
'''Mark the test number in the stats log'''
global has_log_stats
if not self.Env["stats"]:
return
for host in self.Env["nodes"]:
log_stats_file = "%s/cts-stats.csv" % CTSvars.CRM_DAEMON_DIR
if not has_log_stats.has_key(host):
global log_stats
global log_stats_bin
script=log_stats
#script = re.sub("\\\\", "\\\\", script)
script = re.sub('\"', '\\\"', script)
script = re.sub("'", "\'", script)
script = re.sub("`", "\`", script)
script = re.sub("\$", "\\\$", script)
self.debug("Installing %s on %s" % (log_stats_bin, host))
self.rsh(host, '''echo "%s" > %s''' % (script, log_stats_bin), silent=True)
self.rsh(host, '''bash %s %s delete''' % (log_stats_bin, log_stats_file))
has_log_stats[host] = 1
# Now mark it
self.rsh(host, '''bash %s %s mark %s''' % (log_stats_bin, log_stats_file, testnum), synchronous=0)
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=[], common_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.badnews_ignore.extend(common_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 2f3a69bdee..989bf849e4 100644
--- a/cts/CTSscenarios.py
+++ b/cts/CTSscenarios.py
@@ -1,579 +1,579 @@
from CTS import *
from CTStests import CTSTest
from CTSaudits import ClusterAudit
from cts.watcher import LogWatcher
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.audit() # Also detects remote/local log config
self.ClusterManager.StatsMark(0)
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["LogFileName"],
self.ClusterManager.templates.get_patterns(
self.ClusterManager.Env["Name"], "BadNews"), "BadNews", 0,
kind=self.ClusterManager.Env["LogWatcher"],
hosts=self.ClusterManager.Env["nodes"])
self.BadNews.setwatch() # Call after we've figured out what type of log watching to do in LogAudit
j = 0
while j < len(self.Components):
if not self.Components[j].SetUp(self.ClusterManager):
# OOPS! We failed. Tear partial setups down.
self.audit()
self.ClusterManager.log("Tearing down partial setup")
self.TearDown(j)
return None
j = j + 1
self.audit()
return 1
def TearDown(self, max=None):
'''Tear Down the Scenario - in reverse order.'''
if max == None:
max = len(self.Components)-1
j = max
while j >= 0:
self.Components[j].TearDown(self.ClusterManager)
j = j - 1
self.audit()
self.ClusterManager.StatsExtract()
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.StatsMark(testcount)
self.ClusterManager.instance_errorstoignore_clear()
self.ClusterManager.log(("Running test %s" % test.name).ljust(35) + (" (%s) " % nodechoice).ljust(15) + "[" + ("%d" % testcount).rjust(3) + "]")
starttime = test.set_timer()
if not test.setup(nodechoice):
self.ClusterManager.log("Setup failed")
ret = 0
elif not test.canrunnow(nodechoice):
self.ClusterManager.log("Skipped")
test.skipped()
else:
did_run = 1
ret = test(nodechoice)
if not test.teardown(nodechoice):
self.ClusterManager.log("Teardown failed")
answer = raw_input('Continue? [nY] ')
if answer and answer == "n":
raise ValueError("Teardown of %s on %s failed" % (test.name, nodechoice))
ret = 0
stoptime = time.time()
self.ClusterManager.oprofileSave(testcount)
elapsed_time = stoptime - starttime
test_time = stoptime - test.get_timer()
if not test.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())
ignorelist.extend(self.ClusterManager.instance_errorstoignore())
# This makes sure everything is stabilized before starting...
failed = 0
for audit in self.Audits:
if not audit():
self.ClusterManager.log("Audit " + audit.name() + " FAILED.")
failed += 1
else:
self.ClusterManager.debug("Audit " + audit.name() + " passed.")
while errcount < 1000:
match = None
if self.BadNews:
match = self.BadNews.look(0)
if match:
add_err = 1
for ignore in ignorelist:
if add_err == 1 and re.search(ignore, match):
add_err = 0
if add_err == 1:
self.ClusterManager.log("BadNews: " + match)
self.incr("BadNews")
errcount = errcount + 1
else:
break
else:
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!")
if self.BadNews:
self.BadNews.end()
return failed
class AllOnce(Scenario):
'''Every Test Once''' # Accessable as __doc__
def run_loop(self, Iterations):
testcount = 1
for test in self.Tests:
self.run_test(test, testcount)
testcount += 1
class RandomTests(Scenario):
'''Random Test Execution'''
def run_loop(self, Iterations):
testcount = 1
while testcount <= Iterations:
test = self.ClusterManager.Env.RandomGen.choice(self.Tests)
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 Boot(Scenario):
'''Start the Cluster'''
def run_loop(self, Iterations):
testcount = 0
class BootCluster(ScenarioComponent):
(
'''BootCluster is the most basic of ScenarioComponents.
This ScenarioComponent simply starts the cluster manager on all the nodes.
It is fairly robust as it waits for all nodes to come up before starting
as they might have been rebooted or crashed for some reason beforehand.
''')
def __init__(self, Env):
pass
def IsApplicable(self):
'''BootCluster is so generic it is always Applicable'''
return 1
def SetUp(self, CM):
'''Basic Cluster Manager startup. Start everything'''
CM.prepare()
# Clear out the cobwebs ;-)
CM.stopall(verbose=True, force=True)
# Now start the Cluster Manager on all the nodes.
CM.log("Starting Cluster Manager on all nodes.")
return CM.startall(verbose=True, quick=True)
def TearDown(self, CM, force=False):
'''Set up the given ScenarioComponent'''
# Stop the cluster manager everywhere
CM.log("Stopping Cluster Manager on all nodes")
return CM.stopall(verbose=True, force=force)
class LeaveBooted(BootCluster):
def TearDown(self, CM):
'''Set up the given ScenarioComponent'''
# Stop the cluster manager everywhere
CM.log("Leaving Cluster running on all nodes")
return 1
class PingFest(ScenarioComponent):
(
'''PingFest does a flood ping to each node in the cluster from the test machine.
If the LabEnvironment Parameter PingSize is set, it will be used as the size
of ping packet requested (via the -s option). If it is not set, it defaults
to 1024 bytes.
According to the manual page for ping:
Outputs packets as fast as they come back or one hundred times per
second, whichever is more. For every ECHO_REQUEST sent a period ``.''
is printed, while for every ECHO_REPLY received a backspace is printed.
This provides a rapid display of how many packets are being dropped.
Only the super-user may use this option. This can be very hard on a net-
work and should be used with caution.
''' )
def __init__(self, Env):
self.Env = Env
def IsApplicable(self):
'''PingFests are always applicable ;-)
'''
return 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, force=True)
# Now start the Cluster Manager on all the nodes.
CM.log("Starting Cluster Manager on all node(s).")
return CM.startall()
def TearDown(self, CM):
CM.log("Stopping Cluster Manager on all node(s).")
return CM.stopall()
class RollingUpgrade(ScenarioComponent):
(
'''
Test a rolling upgrade between two versions of the stack
''')
def __init__(self, Env):
self.Env = Env
def IsApplicable(self):
if not self.Env["rpm-dir"]:
return None
if not self.Env["current-version"]:
return None
if not self.Env["previous-version"]:
return None
return 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):
- print repr(self)+"prepare"
+ print(repr(self)+"prepare")
CM.prepare()
# Clear out the cobwebs
CM.stopall(force=True)
CM.log("Downgrading all nodes to %s." % self.Env["previous-version"])
for node in self.Env["nodes"]:
if not self.downgrade(node):
CM.log("Couldn't downgrade %s" % node)
return None
return 1
def TearDown(self, CM):
# Stop everything
CM.log("Stopping Cluster Manager on Upgrade nodes.")
CM.stopall()
CM.log("Upgrading all nodes to %s." % self.Env["current-version"])
for node in self.Env["nodes"]:
if not self.upgrade(node):
CM.log("Couldn't upgrade %s" % node)
return None
return 1
diff --git a/cts/CTStests.py b/cts/CTStests.py
index 4d9bd8aa97..ef36916d0c 100644
--- a/cts/CTStests.py
+++ b/cts/CTStests.py
@@ -1,3225 +1,3225 @@
'''CTS: Cluster Testing System: Tests module
There are a few things we want to do here:
'''
__copyright__ = '''
Copyright (C) 2000, 2001 Alan Robertson <alanr@unix.sh>
Licensed under the GNU GPL.
Add RecourceRecover testcase Zhao Kai <zhaokai@cn.ibm.com>
'''
#
# 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., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
#
# SPECIAL NOTE:
#
# Tests may NOT implement any cluster-manager-specific code in them.
# EXTEND the ClusterManager object to provide the base capabilities
# the test needs if you need to do something that the current CM classes
# do not. Otherwise you screw up the whole point of the object structure
# in CTS.
#
# Thank you.
#
import time, os, re, string, tempfile
from stat import *
from cts import CTS
from cts.CTSaudits import *
from cts.CTSvars import *
from cts.patterns import PatternSelector
from cts.logging import LogFactory
from cts.remote import RemoteFactory
from cts.watcher import LogWatcher
from cts.environment import EnvFactory
AllTestClasses = [ ]
class CTSTest:
'''
A Cluster test.
We implement the basic set of properties and behaviors for a generic
cluster test.
Cluster tests track their own statistics.
We keep each of the kinds of counts we track as separate {name,value}
pairs.
'''
def __init__(self, cm):
#self.name="the unnamed test"
self.Stats = {"calls":0
, "success":0
, "failure":0
, "skipped":0
, "auditfail":0}
# if not issubclass(cm.__class__, ClusterManager):
# raise ValueError("Must be a ClusterManager object")
self.CM = cm
self.Env = EnvFactory().getInstance()
self.rsh = RemoteFactory().getInstance()
self.logger = LogFactory()
self.templates = PatternSelector(cm["Name"])
self.Audits = []
self.timeout = 120
self.passed = 1
self.is_loop = 0
self.is_unsafe = 0
self.is_docker_unsafe = 0
self.is_experimental = 0
self.is_container = 0
self.is_valgrind = 0
self.benchmark = 0 # which tests to benchmark
self.timer = {} # timers
def log(self, args):
self.logger.log(args)
def debug(self, args):
self.logger.debug(args)
def has_key(self, key):
return self.Stats.has_key(key)
def __setitem__(self, key, value):
self.Stats[key] = value
def __getitem__(self, key):
return self.Stats[key]
def log_mark(self, msg):
self.debug("MARK: test %s %s %d" % (self.name,msg,time.time()))
return
def get_timer(self,key = "test"):
try: return self.timer[key]
except: return 0
def set_timer(self,key = "test"):
self.timer[key] = time.time()
return self.timer[key]
def log_timer(self,key = "test"):
elapsed = 0
if key in self.timer:
elapsed = time.time() - self.timer[key]
s = key == "test" and self.name or "%s:%s" % (self.name,key)
self.debug("%s runtime: %.2f" % (s, elapsed))
del self.timer[key]
return elapsed
def incr(self, name):
'''Increment (or initialize) the value associated with the given name'''
if not self.Stats.has_key(name):
self.Stats[name] = 0
self.Stats[name] = self.Stats[name]+1
# Reset the test passed boolean
if name == "calls":
self.passed = 1
def failure(self, reason="none"):
'''Increment the failure count'''
self.passed = 0
self.incr("failure")
self.logger.log(("Test %s" % self.name).ljust(35) + " FAILED: %s" % reason)
return None
def success(self):
'''Increment the success count'''
self.incr("success")
return 1
def skipped(self):
'''Increment the skipped count'''
self.incr("skipped")
return 1
def __call__(self, node):
'''Perform the given test'''
raise ValueError("Abstract Class member (__call__)")
self.incr("calls")
return self.failure()
def audit(self):
passed = 1
if len(self.Audits) > 0:
for audit in self.Audits:
if not audit():
self.logger.log("Internal %s Audit %s FAILED." % (self.name, audit.name()))
self.incr("auditfail")
passed = 0
return passed
def setup(self, node):
'''Setup the given test'''
return self.success()
def teardown(self, node):
'''Tear down the given test'''
return self.success()
def create_watch(self, patterns, timeout, name=None):
if not name:
name = self.name
return LogWatcher(self.Env["LogFileName"], patterns, name, timeout, kind=self.Env["LogWatcher"], hosts=self.Env["nodes"])
def local_badnews(self, prefix, watch, local_ignore=[]):
errcount = 0
if not prefix:
prefix = "LocalBadNews:"
ignorelist = []
ignorelist.append(" CTS: ")
ignorelist.append(prefix)
ignorelist.extend(local_ignore)
while errcount < 100:
match = watch.look(0)
if match:
add_err = 1
for ignore in ignorelist:
if add_err == 1 and re.search(ignore, match):
add_err = 0
if add_err == 1:
self.logger.log(prefix + " " + match)
errcount = errcount + 1
else:
break
else:
self.logger.log("Too many errors!")
watch.end()
return errcount
def is_applicable(self):
return self.is_applicable_common()
def is_applicable_common(self):
'''Return TRUE if we are applicable in the current test configuration'''
#raise ValueError("Abstract Class member (is_applicable)")
if self.is_loop and not self.Env["loop-tests"]:
return 0
elif self.is_unsafe and not self.Env["unsafe-tests"]:
return 0
elif self.is_valgrind and not self.Env["valgrind-tests"]:
return 0
elif self.is_experimental and not self.Env["experimental-tests"]:
return 0
elif self.is_docker_unsafe and self.Env["docker"]:
return 0
elif self.is_container and not self.Env["container-tests"]:
return 0
elif self.Env["benchmark"] and self.benchmark == 0:
return 0
return 1
def find_ocfs2_resources(self, node):
self.r_o2cb = None
self.r_ocfs2 = []
(rc, lines) = self.rsh(node, "crm_resource -c", None)
for line in lines:
if re.search("^Resource", line):
r = AuditResource(self.CM, line)
if r.rtype == "o2cb" and r.parent != "NA":
self.debug("Found o2cb: %s" % self.r_o2cb)
self.r_o2cb = r.parent
if re.search("^Constraint", line):
c = AuditConstraint(self.CM, line)
if c.type == "rsc_colocation" and c.target == self.r_o2cb:
self.r_ocfs2.append(c.rsc)
self.debug("Found ocfs2 filesystems: %s" % repr(self.r_ocfs2))
return len(self.r_ocfs2)
def canrunnow(self, node):
'''Return TRUE if we can meaningfully run right now'''
return 1
def errorstoignore(self):
'''Return list of errors which are 'normal' and should be ignored'''
return []
class StopTest(CTSTest):
'''Stop (deactivate) the cluster manager on a node'''
def __init__(self, cm):
CTSTest.__init__(self, cm)
self.name = "Stop"
def __call__(self, node):
'''Perform the 'stop' test. '''
self.incr("calls")
if self.CM.ShouldBeStatus[node] != "up":
return self.skipped()
patterns = []
# Technically we should always be able to notice ourselves stopping
patterns.append(self.templates["Pat:We_stopped"] % node)
#if self.Env["use_logd"]:
# patterns.append(self.templates["Pat:Logd_stopped"] % node)
# Any active node needs to notice this one left
# NOTE: This wont work if we have multiple partitions
for other in self.Env["nodes"]:
if self.CM.ShouldBeStatus[other] == "up" and other != node:
patterns.append(self.templates["Pat:They_stopped"] %(other, self.CM.key_for_node(node)))
#self.debug("Checking %s will notice %s left"%(other, node))
watch = self.create_watch(patterns, self.Env["DeadTime"])
watch.setwatch()
if node == self.CM.OurNode:
self.incr("us")
else:
if self.CM.upcount() <= 1:
self.incr("all")
else:
self.incr("them")
self.CM.StopaCM(node)
watch_result = watch.lookforall()
failreason = None
UnmatchedList = "||"
if watch.unmatched:
(rc, output) = self.rsh(node, "/bin/ps axf", None)
for line in output:
self.debug(line)
(rc, output) = self.rsh(node, "/usr/sbin/dlm_tool dump", None)
for line in output:
self.debug(line)
for regex in watch.unmatched:
self.logger.log ("ERROR: Shutdown pattern not found: %s" % (regex))
UnmatchedList += regex + "||";
failreason = "Missing shutdown pattern"
self.CM.cluster_stable(self.Env["DeadTime"])
if not watch.unmatched or self.CM.upcount() == 0:
return self.success()
if len(watch.unmatched) >= self.CM.upcount():
return self.failure("no match against (%s)" % UnmatchedList)
if failreason == None:
return self.success()
else:
return self.failure(failreason)
#
# We don't register StopTest because it's better when called by
# another test...
#
class StartTest(CTSTest):
'''Start (activate) the cluster manager on a node'''
def __init__(self, cm, debug=None):
CTSTest.__init__(self,cm)
self.name = "start"
self.debug = debug
def __call__(self, node):
'''Perform the 'start' test. '''
self.incr("calls")
if self.CM.upcount() == 0:
self.incr("us")
else:
self.incr("them")
if self.CM.ShouldBeStatus[node] != "down":
return self.skipped()
elif self.CM.StartaCM(node):
return self.success()
else:
return self.failure("Startup %s on node %s failed"
% (self.Env["Name"], node))
#
# We don't register StartTest because it's better when called by
# another test...
#
class FlipTest(CTSTest):
'''If it's running, stop it. If it's stopped start it.
Overthrow the status quo...
'''
def __init__(self, cm):
CTSTest.__init__(self,cm)
self.name = "Flip"
self.start = StartTest(cm)
self.stop = StopTest(cm)
def __call__(self, node):
'''Perform the 'Flip' test. '''
self.incr("calls")
if self.CM.ShouldBeStatus[node] == "up":
self.incr("stopped")
ret = self.stop(node)
type = "up->down"
# Give the cluster time to recognize it's gone...
time.sleep(self.Env["StableTime"])
elif self.CM.ShouldBeStatus[node] == "down":
self.incr("started")
ret = self.start(node)
type = "down->up"
else:
return self.skipped()
self.incr(type)
if ret:
return self.success()
else:
return self.failure("%s failure" % type)
# Register FlipTest as a good test to run
AllTestClasses.append(FlipTest)
class RestartTest(CTSTest):
'''Stop and restart a node'''
def __init__(self, cm):
CTSTest.__init__(self,cm)
self.name = "Restart"
self.start = StartTest(cm)
self.stop = StopTest(cm)
self.benchmark = 1
def __call__(self, node):
'''Perform the 'restart' test. '''
self.incr("calls")
self.incr("node:" + node)
ret1 = 1
if self.CM.StataCM(node):
self.incr("WasStopped")
if not self.start(node):
return self.failure("start (setup) failure: "+node)
self.set_timer()
if not self.stop(node):
return self.failure("stop failure: "+node)
if not self.start(node):
return self.failure("start failure: "+node)
return self.success()
# Register RestartTest as a good test to run
AllTestClasses.append(RestartTest)
class StonithdTest(CTSTest):
def __init__(self, cm):
CTSTest.__init__(self, cm)
self.name = "Stonithd"
self.startall = SimulStartLite(cm)
self.benchmark = 1
def __call__(self, node):
self.incr("calls")
if len(self.Env["nodes"]) < 2:
return self.skipped()
ret = self.startall(None)
if not ret:
return self.failure("Setup failed")
is_dc = self.CM.is_node_dc(node)
watchpats = []
watchpats.append(self.templates["Pat:FenceOpOK"] % node)
watchpats.append(self.templates["Pat:NodeFenced"] % node)
if self.Env["at-boot"] == 0:
self.debug("Expecting %s to stay down" % node)
self.CM.ShouldBeStatus[node] = "down"
else:
self.debug("Expecting %s to come up again %d" % (node, self.Env["at-boot"]))
watchpats.append("%s.* S_STARTING -> S_PENDING" % node)
watchpats.append("%s.* S_PENDING -> S_NOT_DC" % node)
watch = self.create_watch(watchpats, 30 + self.Env["DeadTime"] + self.Env["StableTime"] + self.Env["StartTime"])
watch.setwatch()
origin = self.Env.RandomGen.choice(self.Env["nodes"])
rc = self.rsh(origin, "stonith_admin --reboot %s -VVVVVV" % node)
if rc == 194:
# 194 - 256 = -62 = Timer expired
#
# Look for the patterns, usually this means the required
# device was running on the node to be fenced - or that
# the required devices were in the process of being loaded
# and/or moved
#
# Effectively the node committed suicide so there will be
# no confirmation, but pacemaker should be watching and
# fence the node again
self.logger.log("Fencing command on %s to fence %s timed out" % (origin, node))
elif origin != node and rc != 0:
self.debug("Waiting for the cluster to recover")
self.CM.cluster_stable()
self.debug("Waiting STONITHd node to come back up")
self.CM.ns.WaitForAllNodesToComeUp(self.Env["nodes"], 600)
self.logger.log("Fencing command on %s failed to fence %s (rc=%d)" % (origin, node, rc))
elif origin == node and rc != 255:
# 255 == broken pipe, ie. the node was fenced as expected
self.logger.log("Locally originated fencing returned %d" % rc)
self.set_timer("fence")
matched = watch.lookforall()
self.log_timer("fence")
self.set_timer("reform")
if watch.unmatched:
self.logger.log("Patterns not found: " + repr(watch.unmatched))
self.debug("Waiting for the cluster to recover")
self.CM.cluster_stable()
self.debug("Waiting STONITHd node to come back up")
self.CM.ns.WaitForAllNodesToComeUp(self.Env["nodes"], 600)
self.debug("Waiting for the cluster to re-stabilize with all nodes")
is_stable = self.CM.cluster_stable(self.Env["StartTime"])
if not matched:
return self.failure("Didn't find all expected patterns")
elif not is_stable:
return self.failure("Cluster did not become stable")
self.log_timer("reform")
return self.success()
def errorstoignore(self):
return [
self.templates["Pat:Fencing_start"] % ".*",
self.templates["Pat:Fencing_ok"] % ".*",
r"error.*: Resource .*stonith::.* is active on 2 nodes attempting recovery",
r"error.*: Operation reboot of .*by .* for stonith_admin.*: Timer expired",
]
def is_applicable(self):
if not self.is_applicable_common():
return 0
if self.Env.has_key("DoFencing"):
return self.Env["DoFencing"]
return 1
AllTestClasses.append(StonithdTest)
class StartOnebyOne(CTSTest):
'''Start all the nodes ~ one by one'''
def __init__(self, cm):
CTSTest.__init__(self,cm)
self.name = "StartOnebyOne"
self.stopall = SimulStopLite(cm)
self.start = StartTest(cm)
self.ns = CTS.NodeStatus(cm.Env)
def __call__(self, dummy):
'''Perform the 'StartOnebyOne' test. '''
self.incr("calls")
# We ignore the "node" parameter...
# Shut down all the nodes...
ret = self.stopall(None)
if not ret:
return self.failure("Test setup failed")
failed = []
self.set_timer()
for node in self.Env["nodes"]:
if not self.start(node):
failed.append(node)
if len(failed) > 0:
return self.failure("Some node failed to start: " + repr(failed))
return self.success()
# Register StartOnebyOne as a good test to run
AllTestClasses.append(StartOnebyOne)
class SimulStart(CTSTest):
'''Start all the nodes ~ simultaneously'''
def __init__(self, cm):
CTSTest.__init__(self,cm)
self.name = "SimulStart"
self.stopall = SimulStopLite(cm)
self.startall = SimulStartLite(cm)
def __call__(self, dummy):
'''Perform the 'SimulStart' test. '''
self.incr("calls")
# We ignore the "node" parameter...
# Shut down all the nodes...
ret = self.stopall(None)
if not ret:
return self.failure("Setup failed")
self.CM.clear_all_caches()
if not self.startall(None):
return self.failure("Startall failed")
return self.success()
# Register SimulStart as a good test to run
AllTestClasses.append(SimulStart)
class SimulStop(CTSTest):
'''Stop all the nodes ~ simultaneously'''
def __init__(self, cm):
CTSTest.__init__(self,cm)
self.name = "SimulStop"
self.startall = SimulStartLite(cm)
self.stopall = SimulStopLite(cm)
def __call__(self, dummy):
'''Perform the 'SimulStop' test. '''
self.incr("calls")
# We ignore the "node" parameter...
# Start up all the nodes...
ret = self.startall(None)
if not ret:
return self.failure("Setup failed")
if not self.stopall(None):
return self.failure("Stopall failed")
return self.success()
# Register SimulStop as a good test to run
AllTestClasses.append(SimulStop)
class StopOnebyOne(CTSTest):
'''Stop all the nodes in order'''
def __init__(self, cm):
CTSTest.__init__(self,cm)
self.name = "StopOnebyOne"
self.startall = SimulStartLite(cm)
self.stop = StopTest(cm)
def __call__(self, dummy):
'''Perform the 'StopOnebyOne' test. '''
self.incr("calls")
# We ignore the "node" parameter...
# Start up all the nodes...
ret = self.startall(None)
if not ret:
return self.failure("Setup failed")
failed = []
self.set_timer()
for node in self.Env["nodes"]:
if not self.stop(node):
failed.append(node)
if len(failed) > 0:
return self.failure("Some node failed to stop: " + repr(failed))
self.CM.clear_all_caches()
return self.success()
# Register StopOnebyOne as a good test to run
AllTestClasses.append(StopOnebyOne)
class RestartOnebyOne(CTSTest):
'''Restart all the nodes in order'''
def __init__(self, cm):
CTSTest.__init__(self,cm)
self.name = "RestartOnebyOne"
self.startall = SimulStartLite(cm)
def __call__(self, dummy):
'''Perform the 'RestartOnebyOne' test. '''
self.incr("calls")
# We ignore the "node" parameter...
# Start up all the nodes...
ret = self.startall(None)
if not ret:
return self.failure("Setup failed")
did_fail = []
self.set_timer()
self.restart = RestartTest(self.CM)
for node in self.Env["nodes"]:
if not self.restart(node):
did_fail.append(node)
if did_fail:
return self.failure("Could not restart %d nodes: %s"
% (len(did_fail), repr(did_fail)))
return self.success()
# Register StopOnebyOne as a good test to run
AllTestClasses.append(RestartOnebyOne)
class PartialStart(CTSTest):
'''Start a node - but tell it to stop before it finishes starting up'''
def __init__(self, cm):
CTSTest.__init__(self,cm)
self.name = "PartialStart"
self.startall = SimulStartLite(cm)
self.stopall = SimulStopLite(cm)
self.stop = StopTest(cm)
#self.is_unsafe = 1
def __call__(self, node):
'''Perform the 'PartialStart' test. '''
self.incr("calls")
ret = self.stopall(None)
if not ret:
return self.failure("Setup failed")
# FIXME! This should use the CM class to get the pattern
# then it would be applicable in general
watchpats = []
watchpats.append("crmd.*Connecting to cluster infrastructure")
watch = self.create_watch(watchpats, self.Env["DeadTime"]+10)
watch.setwatch()
self.CM.StartaCMnoBlock(node)
ret = watch.lookforall()
if not ret:
self.logger.log("Patterns not found: " + repr(watch.unmatched))
return self.failure("Setup of %s failed" % node)
ret = self.stop(node)
if not ret:
return self.failure("%s did not stop in time" % node)
return self.success()
def errorstoignore(self):
'''Return list of errors which should be ignored'''
# We might do some fencing in the 2-node case if we make it up far enough
return [
"""Executing reboot fencing operation""",
]
# Register StopOnebyOne as a good test to run
AllTestClasses.append(PartialStart)
class StandbyTest(CTSTest):
def __init__(self, cm):
CTSTest.__init__(self,cm)
self.name = "Standby"
self.benchmark = 1
self.start = StartTest(cm)
self.startall = SimulStartLite(cm)
# make sure the node is active
# set the node to standby mode
# check resources, none resource should be running on the node
# set the node to active mode
# check resouces, resources should have been migrated back (SHOULD THEY?)
def __call__(self, node):
self.incr("calls")
ret = self.startall(None)
if not ret:
return self.failure("Start all nodes failed")
self.debug("Make sure node %s is active" % node)
if self.CM.StandbyStatus(node) != "off":
if not self.CM.SetStandbyMode(node, "off"):
return self.failure("can't set node %s to active mode" % node)
self.CM.cluster_stable()
status = self.CM.StandbyStatus(node)
if status != "off":
return self.failure("standby status of %s is [%s] but we expect [off]" % (node, status))
self.debug("Getting resources running on node %s" % node)
rsc_on_node = self.CM.active_resources(node)
watchpats = []
watchpats.append(r"State transition .* -> S_POLICY_ENGINE")
watch = self.create_watch(watchpats, self.Env["DeadTime"]+10)
watch.setwatch()
self.debug("Setting node %s to standby mode" % node)
if not self.CM.SetStandbyMode(node, "on"):
return self.failure("can't set node %s to standby mode" % node)
self.set_timer("on")
ret = watch.lookforall()
if not ret:
self.logger.log("Patterns not found: " + repr(watch.unmatched))
self.CM.SetStandbyMode(node, "off")
return self.failure("cluster didn't react to standby change on %s" % node)
self.CM.cluster_stable()
status = self.CM.StandbyStatus(node)
if status != "on":
return self.failure("standby status of %s is [%s] but we expect [on]" % (node, status))
self.log_timer("on")
self.debug("Checking resources")
bad_run = self.CM.active_resources(node)
if len(bad_run) > 0:
rc = self.failure("%s set to standby, %s is still running on it" % (node, repr(bad_run)))
self.debug("Setting node %s to active mode" % node)
self.CM.SetStandbyMode(node, "off")
return rc
self.debug("Setting node %s to active mode" % node)
if not self.CM.SetStandbyMode(node, "off"):
return self.failure("can't set node %s to active mode" % node)
self.set_timer("off")
self.CM.cluster_stable()
status = self.CM.StandbyStatus(node)
if status != "off":
return self.failure("standby status of %s is [%s] but we expect [off]" % (node, status))
self.log_timer("off")
return self.success()
AllTestClasses.append(StandbyTest)
class ValgrindTest(CTSTest):
'''Check for memory leaks'''
def __init__(self, cm):
CTSTest.__init__(self,cm)
self.name = "Valgrind"
self.stopall = SimulStopLite(cm)
self.startall = SimulStartLite(cm)
self.is_valgrind = 1
self.is_loop = 1
def setup(self, node):
self.incr("calls")
ret = self.stopall(None)
if not ret:
return self.failure("Stop all nodes failed")
# Enable valgrind
self.logger.logPat = "/tmp/%s-*.valgrind" % self.name
self.Env["valgrind-prefix"] = self.name
self.rsh(node, "rm -f %s" % self.logger.logPat, None)
ret = self.startall(None)
if not ret:
return self.failure("Start all nodes failed")
for node in self.Env["nodes"]:
(rc, output) = self.rsh(node, "ps u --ppid `pidofproc aisexec`", None)
for line in output:
self.debug(line)
return self.success()
def teardown(self, node):
# Disable valgrind
self.Env["valgrind-prefix"] = None
# Return all nodes to normal
ret = self.stopall(None)
if not ret:
return self.failure("Stop all nodes failed")
return self.success()
def find_leaks(self):
# Check for leaks
leaked = []
self.stop = StopTest(self.CM)
for node in self.Env["nodes"]:
(rc, ps_out) = self.rsh(node, "ps u --ppid `pidofproc aisexec`", None)
rc = self.stop(node)
if not rc:
self.failure("Couldn't shut down %s" % node)
rc = self.rsh(node, "grep -e indirectly.*lost:.*[1-9] -e definitely.*lost:.*[1-9] -e (ERROR|error).*SUMMARY:.*[1-9].*errors %s" % self.logger.logPat, 0)
if rc != 1:
leaked.append(node)
self.failure("Valgrind errors detected on %s" % node)
for line in ps_out:
self.logger.log(line)
(rc, output) = self.rsh(node, "grep -e lost: -e SUMMARY: %s" % self.logger.logPat, None)
for line in output:
self.logger.log(line)
(rc, output) = self.rsh(node, "cat %s" % self.logger.logPat, None)
for line in output:
self.debug(line)
self.rsh(node, "rm -f %s" % self.logger.logPat, None)
return leaked
def __call__(self, node):
leaked = self.find_leaks()
if len(leaked) > 0:
return self.failure("Nodes %s leaked" % repr(leaked))
return self.success()
def errorstoignore(self):
'''Return list of errors which should be ignored'''
return [
r"cib.*: \*\*\*\*\*\*\*\*\*\*\*\*\*",
r"cib.*: .* avoid confusing Valgrind",
r"HA_VALGRIND_ENABLED",
]
class StandbyLoopTest(ValgrindTest):
'''Check for memory leaks by putting a node in and out of standby for an hour'''
def __init__(self, cm):
ValgrindTest.__init__(self,cm)
self.name = "StandbyLoop"
def __call__(self, node):
lpc = 0
delay = 2
failed = 0
done = time.time() + self.Env["loop-minutes"] * 60
while time.time() <= done and not failed:
lpc = lpc + 1
time.sleep(delay)
if not self.CM.SetStandbyMode(node, "on"):
self.failure("can't set node %s to standby mode" % node)
failed = lpc
time.sleep(delay)
if not self.CM.SetStandbyMode(node, "off"):
self.failure("can't set node %s to active mode" % node)
failed = lpc
leaked = self.find_leaks()
if failed:
return self.failure("Iteration %d failed" % failed)
elif len(leaked) > 0:
return self.failure("Nodes %s leaked" % repr(leaked))
return self.success()
AllTestClasses.append(StandbyLoopTest)
class BandwidthTest(CTSTest):
# Tests should not be cluster-manager-specific
# If you need to find out cluster manager configuration to do this, then
# it should be added to the generic cluster manager API.
'''Test the bandwidth which heartbeat uses'''
def __init__(self, cm):
CTSTest.__init__(self, cm)
self.name = "Bandwidth"
self.start = StartTest(cm)
self.__setitem__("min",0)
self.__setitem__("max",0)
self.__setitem__("totalbandwidth",0)
self.tempfile = tempfile.mktemp(".cts")
self.startall = SimulStartLite(cm)
def __call__(self, node):
'''Perform the Bandwidth test'''
self.incr("calls")
if self.CM.upcount() < 1:
return self.skipped()
Path = self.CM.InternalCommConfig()
if "ip" not in Path["mediatype"]:
return self.skipped()
port = Path["port"][0]
port = int(port)
ret = self.startall(None)
if not ret:
return self.failure("Test setup failed")
time.sleep(5) # We get extra messages right after startup.
fstmpfile = "/var/run/band_estimate"
dumpcmd = "tcpdump -p -n -c 102 -i any udp port %d > %s 2>&1" \
% (port, fstmpfile)
rc = self.rsh(node, dumpcmd)
if rc == 0:
farfile = "root@%s:%s" % (node, fstmpfile)
self.rsh.cp(farfile, self.tempfile)
Bandwidth = self.countbandwidth(self.tempfile)
if not Bandwidth:
self.logger.log("Could not compute bandwidth.")
return self.success()
intband = int(Bandwidth + 0.5)
self.logger.log("...bandwidth: %d bits/sec" % intband)
self.Stats["totalbandwidth"] = self.Stats["totalbandwidth"] + Bandwidth
if self.Stats["min"] == 0:
self.Stats["min"] = Bandwidth
if Bandwidth > self.Stats["max"]:
self.Stats["max"] = Bandwidth
if Bandwidth < self.Stats["min"]:
self.Stats["min"] = Bandwidth
self.rsh(node, "rm -f %s" % fstmpfile)
os.unlink(self.tempfile)
return self.success()
else:
return self.failure("no response from tcpdump command [%d]!" % rc)
def countbandwidth(self, file):
fp = open(file, "r")
fp.seek(0)
count = 0
sum = 0
while 1:
line = fp.readline()
if not line:
return None
if re.search("udp",line) or re.search("UDP,", line):
count = count + 1
linesplit = string.split(line," ")
for j in range(len(linesplit)-1):
if linesplit[j] == "udp": break
if linesplit[j] == "length:": break
try:
sum = sum + int(linesplit[j+1])
except ValueError:
self.logger.log("Invalid tcpdump line: %s" % line)
return None
T1 = linesplit[0]
timesplit = string.split(T1,":")
time2split = string.split(timesplit[2],".")
time1 = (long(timesplit[0])*60+long(timesplit[1]))*60+long(time2split[0])+long(time2split[1])*0.000001
break
while count < 100:
line = fp.readline()
if not line:
return None
if re.search("udp",line) or re.search("UDP,", line):
count = count+1
linessplit = string.split(line," ")
for j in range(len(linessplit)-1):
if linessplit[j] == "udp": break
if linesplit[j] == "length:": break
try:
sum = int(linessplit[j+1]) + sum
except ValueError:
self.logger.log("Invalid tcpdump line: %s" % line)
return None
T2 = linessplit[0]
timesplit = string.split(T2,":")
time2split = string.split(timesplit[2],".")
time2 = (long(timesplit[0])*60+long(timesplit[1]))*60+long(time2split[0])+long(time2split[1])*0.000001
time = time2-time1
if (time <= 0):
return 0
return (sum*8)/time
def is_applicable(self):
'''BandwidthTest never applicable'''
return 0
AllTestClasses.append(BandwidthTest)
###################################################################
class MaintenanceMode(CTSTest):
###################################################################
def __init__(self, cm):
CTSTest.__init__(self,cm)
self.name = "MaintenanceMode"
self.start = StartTest(cm)
self.startall = SimulStartLite(cm)
self.max = 30
#self.is_unsafe = 1
self.benchmark = 1
self.action = "asyncmon"
self.interval = 0
self.rid = "maintenanceDummy"
def toggleMaintenanceMode(self, node, action):
pats = []
pats.append(self.templates["Pat:DC_IDLE"])
# fail the resource right after turning Maintenance mode on
# verify it is not recovered until maintenance mode is turned off
if action == "On":
pats.append(r"pengine.*:\s+warning:.*Processing failed op %s for %s on" % (self.action, self.rid))
else:
pats.append(self.templates["Pat:RscOpOK"] % (self.rid, "stop_0"))
pats.append(self.templates["Pat:RscOpOK"] % (self.rid, "start_0"))
watch = self.create_watch(pats, 60)
watch.setwatch()
self.debug("Turning maintenance mode %s" % action)
self.rsh(node, self.templates["MaintenanceMode%s" % (action)])
if (action == "On"):
self.rsh(node, "crm_resource -V -F -r %s -H %s &>/dev/null" % (self.rid, node))
self.set_timer("recover%s" % (action))
watch.lookforall()
self.log_timer("recover%s" % (action))
if watch.unmatched:
self.debug("Failed to find patterns when turning maintenance mode %s" % action)
return repr(watch.unmatched)
return ""
def insertMaintenanceDummy(self, node):
pats = []
pats.append(("%s.*" % node) + (self.templates["Pat:RscOpOK"] % (self.rid, "start_0")))
watch = self.create_watch(pats, 60)
watch.setwatch()
self.CM.AddDummyRsc(node, self.rid)
self.set_timer("addDummy")
watch.lookforall()
self.log_timer("addDummy")
if watch.unmatched:
self.debug("Failed to find patterns when adding maintenance dummy resource")
return repr(watch.unmatched)
return ""
def removeMaintenanceDummy(self, node):
pats = []
pats.append(self.templates["Pat:RscOpOK"] % (self.rid, "stop_0"))
watch = self.create_watch(pats, 60)
watch.setwatch()
self.CM.RemoveDummyRsc(node, self.rid)
self.set_timer("removeDummy")
watch.lookforall()
self.log_timer("removeDummy")
if watch.unmatched:
self.debug("Failed to find patterns when removing maintenance dummy resource")
return repr(watch.unmatched)
return ""
def managedRscList(self, node):
rscList = []
(rc, lines) = self.rsh(node, "crm_resource -c", None)
for line in lines:
if re.search("^Resource", line):
tmp = AuditResource(self.CM, line)
if tmp.managed():
rscList.append(tmp.id)
return rscList
def verifyResources(self, node, rscList, managed):
managedList = list(rscList)
managed_str = "managed"
if not managed:
managed_str = "unmanaged"
(rc, lines) = self.rsh(node, "crm_resource -c", None)
for line in lines:
if re.search("^Resource", line):
tmp = AuditResource(self.CM, line)
if managed and not tmp.managed():
continue
elif not managed and tmp.managed():
continue
elif managedList.count(tmp.id):
managedList.remove(tmp.id)
if len(managedList) == 0:
self.debug("Found all %s resources on %s" % (managed_str, node))
return True
self.logger.log("Could not find all %s resources on %s. %s" % (managed_str, node, managedList))
return False
def __call__(self, node):
'''Perform the 'MaintenanceMode' test. '''
self.incr("calls")
verify_managed = False
verify_unmanaged = False
failPat = ""
ret = self.startall(None)
if not ret:
return self.failure("Setup failed")
# get a list of all the managed resources. We use this list
# after enabling maintenance mode to verify all managed resources
# become un-managed. After maintenance mode is turned off, we use
# this list to verify all the resources become managed again.
managedResources = self.managedRscList(node)
if len(managedResources) == 0:
self.logger.log("No managed resources on %s" % node)
return self.skipped()
# insert a fake resource we can fail during maintenance mode
# so we can verify recovery does not take place until after maintenance
# mode is disabled.
failPat = failPat + self.insertMaintenanceDummy(node)
# toggle maintenance mode ON, then fail dummy resource.
failPat = failPat + self.toggleMaintenanceMode(node, "On")
# verify all the resources are now unmanaged
if self.verifyResources(node, managedResources, False):
verify_unmanaged = True
# Toggle maintenance mode OFF, verify dummy is recovered.
failPat = failPat + self.toggleMaintenanceMode(node, "Off")
# verify all the resources are now managed again
if self.verifyResources(node, managedResources, True):
verify_managed = True
# Remove our maintenance dummy resource.
failPat = failPat + self.removeMaintenanceDummy(node)
self.CM.cluster_stable()
if failPat != "":
return self.failure("Unmatched patterns: %s" % (failPat))
elif verify_unmanaged is False:
return self.failure("Failed to verify resources became unmanaged during maintenance mode")
elif verify_managed is False:
return self.failure("Failed to verify resources switched back to managed after disabling maintenance mode")
return self.success()
def errorstoignore(self):
'''Return list of errors which should be ignored'''
return [
r"Updating failcount for %s" % self.rid,
r"pengine.*: Recover %s\s*\(.*\)" % self.rid,
r"Unknown operation: fail",
r"(ERROR|error): sending stonithRA op to stonithd failed.",
self.templates["Pat:RscOpOK"] % (self.rid, ("%s_%d" % (self.action, self.interval))),
r"(ERROR|error).*: Action %s_%s_%d .* initiated outside of a transition" % (self.rid, self.action, self.interval),
]
AllTestClasses.append(MaintenanceMode)
class ResourceRecover(CTSTest):
def __init__(self, cm):
CTSTest.__init__(self,cm)
self.name = "ResourceRecover"
self.start = StartTest(cm)
self.startall = SimulStartLite(cm)
self.max = 30
self.rid = None
self.rid_alt = None
#self.is_unsafe = 1
self.benchmark = 1
# these are the values used for the new LRM API call
self.action = "asyncmon"
self.interval = 0
def __call__(self, node):
'''Perform the 'ResourceRecover' test. '''
self.incr("calls")
ret = self.startall(None)
if not ret:
return self.failure("Setup failed")
resourcelist = self.CM.active_resources(node)
# if there are no resourcelist, return directly
if len(resourcelist) == 0:
self.logger.log("No active resources on %s" % node)
return self.skipped()
self.rid = self.Env.RandomGen.choice(resourcelist)
self.rid_alt = self.rid
rsc = None
(rc, lines) = self.rsh(node, "crm_resource -c", None)
for line in lines:
if re.search("^Resource", line):
tmp = AuditResource(self.CM, line)
if tmp.id == self.rid:
rsc = tmp
# Handle anonymous clones that get renamed
self.rid = rsc.clone_id
break
if not rsc:
return self.failure("Could not find %s in the resource list" % self.rid)
self.debug("Shooting %s aka. %s" % (rsc.clone_id, rsc.id))
pats = []
pats.append(r"pengine.*:\s+warning:.*Processing failed op %s for (%s|%s) on" % (self.action,
rsc.id, rsc.clone_id))
if rsc.managed():
pats.append(self.templates["Pat:RscOpOK"] % (self.rid, "stop_0"))
if rsc.unique():
pats.append(self.templates["Pat:RscOpOK"] % (self.rid, "start_0"))
else:
# Anonymous clones may get restarted with a different clone number
pats.append(self.templates["Pat:RscOpOK"] % (".*", "start_0"))
watch = self.create_watch(pats, 60)
watch.setwatch()
self.rsh(node, "crm_resource -V -F -r %s -H %s &>/dev/null" % (self.rid, node))
self.set_timer("recover")
watch.lookforall()
self.log_timer("recover")
self.CM.cluster_stable()
recovered = self.CM.ResourceLocation(self.rid)
if watch.unmatched:
return self.failure("Patterns not found: %s" % repr(watch.unmatched))
elif rsc.unique() and len(recovered) > 1:
return self.failure("%s is now active on more than one node: %s"%(self.rid, repr(recovered)))
elif len(recovered) > 0:
self.debug("%s is running on: %s" % (self.rid, repr(recovered)))
elif rsc.managed():
return self.failure("%s was not recovered and is inactive" % self.rid)
return self.success()
def errorstoignore(self):
'''Return list of errors which should be ignored'''
return [
r"Updating failcount for %s" % self.rid,
r"pengine.*: Recover (%s|%s)\s*\(.*\)" % (self.rid, self.rid_alt),
r"Unknown operation: fail",
r"(ERROR|error): sending stonithRA op to stonithd failed.",
self.templates["Pat:RscOpOK"] % (self.rid, ("%s_%d" % (self.action, self.interval))),
r"(ERROR|error).*: Action %s_%s_%d .* initiated outside of a transition" % (self.rid, self.action, self.interval),
]
AllTestClasses.append(ResourceRecover)
class ComponentFail(CTSTest):
def __init__(self, cm):
CTSTest.__init__(self,cm)
self.name = "ComponentFail"
# TODO make this work correctly in docker.
self.is_docker_unsafe = 1
self.startall = SimulStartLite(cm)
self.complist = cm.Components()
self.patterns = []
self.okerrpatterns = []
self.is_unsafe = 1
def __call__(self, node):
'''Perform the 'ComponentFail' test. '''
self.incr("calls")
self.patterns = []
self.okerrpatterns = []
# start all nodes
ret = self.startall(None)
if not ret:
return self.failure("Setup failed")
if not self.CM.cluster_stable(self.Env["StableTime"]):
return self.failure("Setup failed - unstable")
node_is_dc = self.CM.is_node_dc(node, None)
# select a component to kill
chosen = self.Env.RandomGen.choice(self.complist)
while chosen.dc_only == 1 and node_is_dc == 0:
chosen = self.Env.RandomGen.choice(self.complist)
self.debug("...component %s (dc=%d,boot=%d)" % (chosen.name, node_is_dc,chosen.triggersreboot))
self.incr(chosen.name)
if chosen.name != "aisexec" and chosen.name != "corosync":
if self.Env["Name"] != "crm-lha" or chosen.name != "pengine":
self.patterns.append(self.templates["Pat:ChildKilled"] %(node, chosen.name))
self.patterns.append(self.templates["Pat:ChildRespawn"] %(node, chosen.name))
self.patterns.extend(chosen.pats)
if node_is_dc:
self.patterns.extend(chosen.dc_pats)
# In an ideal world, this next stuff should be in the "chosen" object as a member function
if self.Env["Name"] == "crm-lha" and chosen.triggersreboot:
# Make sure the node goes down and then comes back up if it should reboot...
for other in self.Env["nodes"]:
if other != node:
self.patterns.append(self.templates["Pat:They_stopped"] %(other, self.CM.key_for_node(node)))
self.patterns.append(self.templates["Pat:Slave_started"] % node)
self.patterns.append(self.templates["Pat:Local_started"] % node)
if chosen.dc_only:
# Sometimes these will be in the log, and sometimes they won't...
self.okerrpatterns.append("%s .*Process %s:.* exited" % (node, chosen.name))
self.okerrpatterns.append("%s .*I_ERROR.*crmdManagedChildDied" % node)
self.okerrpatterns.append("%s .*The %s subsystem terminated unexpectedly" % (node, chosen.name))
self.okerrpatterns.append("(ERROR|error): Client .* exited with return code")
else:
# Sometimes this won't be in the log...
self.okerrpatterns.append(self.templates["Pat:ChildKilled"] %(node, chosen.name))
self.okerrpatterns.append(self.templates["Pat:ChildRespawn"] %(node, chosen.name))
self.okerrpatterns.append(self.templates["Pat:ChildExit"])
if chosen.name == "stonith":
# Ignore actions for STONITH resources
(rc, lines) = self.rsh(node, "crm_resource -c", None)
for line in lines:
if re.search("^Resource", line):
r = AuditResource(self.CM, line)
if r.rclass == "stonith":
self.okerrpatterns.append(self.templates["Pat:Fencing_recover"] % r.id)
# supply a copy so self.patterns doesnt end up empty
tmpPats = []
tmpPats.extend(self.patterns)
self.patterns.extend(chosen.badnews_ignore)
# Look for STONITH ops, depending on Env["at-boot"] we might need to change the nodes status
stonithPats = []
stonithPats.append(self.templates["Pat:Fencing_ok"] % node)
stonith = self.create_watch(stonithPats, 0)
stonith.setwatch()
# set the watch for stable
watch = self.create_watch(
tmpPats, self.Env["DeadTime"] + self.Env["StableTime"] + self.Env["StartTime"])
watch.setwatch()
# kill the component
chosen.kill(node)
self.debug("Waiting for the cluster to recover")
self.CM.cluster_stable()
self.debug("Waiting for any STONITHd node to come back up")
self.CM.ns.WaitForAllNodesToComeUp(self.Env["nodes"], 600)
self.debug("Waiting for the cluster to re-stabilize with all nodes")
self.CM.cluster_stable(self.Env["StartTime"])
self.debug("Checking if %s was shot" % node)
shot = stonith.look(60)
if shot:
self.debug("Found: " + repr(shot))
self.okerrpatterns.append(self.templates["Pat:Fencing_start"] % node)
if self.Env["at-boot"] == 0:
self.CM.ShouldBeStatus[node] = "down"
# If fencing occurred, chances are many (if not all) the expected logs
# will not be sent - or will be lost when the node reboots
return self.success()
# check for logs indicating a graceful recovery
matched = watch.lookforall(allow_multiple_matches=1)
if watch.unmatched:
self.logger.log("Patterns not found: " + repr(watch.unmatched))
self.debug("Waiting for the cluster to re-stabilize with all nodes")
is_stable = self.CM.cluster_stable(self.Env["StartTime"])
if not matched:
return self.failure("Didn't find all expected %s patterns" % chosen.name)
elif not is_stable:
return self.failure("Cluster did not become stable after killing %s" % chosen.name)
return self.success()
def errorstoignore(self):
'''Return list of errors which should be ignored'''
# Note that okerrpatterns refers to the last time we ran this test
# The good news is that this works fine for us...
self.okerrpatterns.extend(self.patterns)
return self.okerrpatterns
AllTestClasses.append(ComponentFail)
class SplitBrainTest(CTSTest):
'''It is used to test split-brain. when the path between the two nodes break
check the two nodes both take over the resource'''
def __init__(self,cm):
CTSTest.__init__(self,cm)
self.name = "SplitBrain"
self.start = StartTest(cm)
self.startall = SimulStartLite(cm)
self.is_experimental = 1
def isolate_partition(self, partition):
other_nodes = []
other_nodes.extend(self.Env["nodes"])
for node in partition:
try:
other_nodes.remove(node)
except ValueError:
self.logger.log("Node "+node+" not in " + repr(self.Env["nodes"]) + " from " +repr(partition))
if len(other_nodes) == 0:
return 1
self.debug("Creating partition: " + repr(partition))
self.debug("Everyone else: " + repr(other_nodes))
for node in partition:
if not self.CM.isolate_node(node, other_nodes):
self.logger.log("Could not isolate %s" % node)
return 0
return 1
def heal_partition(self, partition):
other_nodes = []
other_nodes.extend(self.Env["nodes"])
for node in partition:
try:
other_nodes.remove(node)
except ValueError:
self.logger.log("Node "+node+" not in " + repr(self.Env["nodes"]))
if len(other_nodes) == 0:
return 1
self.debug("Healing partition: " + repr(partition))
self.debug("Everyone else: " + repr(other_nodes))
for node in partition:
self.CM.unisolate_node(node, other_nodes)
def __call__(self, node):
'''Perform split-brain test'''
self.incr("calls")
self.passed = 1
partitions = {}
ret = self.startall(None)
if not ret:
return self.failure("Setup failed")
while 1:
# Retry until we get multiple partitions
partitions = {}
p_max = len(self.Env["nodes"])
for node in self.Env["nodes"]:
p = self.Env.RandomGen.randint(1, p_max)
if not partitions.has_key(p):
partitions[p] = []
partitions[p].append(node)
p_max = len(partitions.keys())
if p_max > 1:
break
# else, try again
self.debug("Created %d partitions" % p_max)
for key in partitions.keys():
self.debug("Partition["+str(key)+"]:\t"+repr(partitions[key]))
# Disabling STONITH to reduce test complexity for now
self.rsh(node, "crm_attribute -V -n stonith-enabled -v false")
for key in partitions.keys():
self.isolate_partition(partitions[key])
count = 30
while count > 0:
if len(self.CM.find_partitions()) != p_max:
time.sleep(10)
else:
break
else:
self.failure("Expected partitions were not created")
# Target number of partitions formed - wait for stability
if not self.CM.cluster_stable():
self.failure("Partitioned cluster not stable")
# Now audit the cluster state
self.CM.partitions_expected = p_max
if not self.audit():
self.failure("Audits failed")
self.CM.partitions_expected = 1
# And heal them again
for key in partitions.keys():
self.heal_partition(partitions[key])
# Wait for a single partition to form
count = 30
while count > 0:
if len(self.CM.find_partitions()) != 1:
time.sleep(10)
count -= 1
else:
break
else:
self.failure("Cluster did not reform")
# Wait for it to have the right number of members
count = 30
while count > 0:
members = []
partitions = self.CM.find_partitions()
if len(partitions) > 0:
members = partitions[0].split()
if len(members) != len(self.Env["nodes"]):
time.sleep(10)
count -= 1
else:
break
else:
self.failure("Cluster did not completely reform")
# Wait up to 20 minutes - the delay is more preferable than
# trying to continue with in a messed up state
if not self.CM.cluster_stable(1200):
self.failure("Reformed cluster not stable")
answer = raw_input('Continue? [nY]')
if answer and answer == "n":
raise ValueError("Reformed cluster not stable")
# Turn fencing back on
if self.Env["DoFencing"]:
self.rsh(node, "crm_attribute -V -D -n stonith-enabled")
self.CM.cluster_stable()
if self.passed:
return self.success()
return self.failure("See previous errors")
def errorstoignore(self):
'''Return list of errors which are 'normal' and should be ignored'''
return [
r"Another DC detected:",
r"(ERROR|error).*: .*Application of an update diff failed",
r"crmd.*:.*not in our membership list",
r"CRIT:.*node.*returning after partition",
]
def is_applicable(self):
if not self.is_applicable_common():
return 0
return len(self.Env["nodes"]) > 2
AllTestClasses.append(SplitBrainTest)
class Reattach(CTSTest):
def __init__(self, cm):
CTSTest.__init__(self,cm)
self.name = "Reattach"
self.startall = SimulStartLite(cm)
self.restart1 = RestartTest(cm)
self.stopall = SimulStopLite(cm)
self.is_unsafe = 0 # Handled by canrunnow()
def setup(self, node):
attempt = 0
if not self.startall(None):
return None
# Make sure we are really _really_ stable and that all
# resources, including those that depend on transient node
# attributes, are started
while not self.CM.cluster_stable(double_check=True):
if attempt < 5:
attempt += 1
self.debug("Not stable yet, re-testing")
else:
self.logger.log("Cluster is not stable")
return None
return 1
def teardown(self, node):
# Make sure 'node' is up
start = StartTest(self.CM)
start(node)
is_managed = self.rsh(node, "crm_attribute -Q -G -t crm_config -n is-managed-default -d true", 1)
is_managed = is_managed[:-1] # Strip off the newline
if is_managed != "true":
self.logger.log("Attempting to re-enable resource management on %s (%s)" % (node, is_managed))
managed = self.create_watch(["is-managed-default"], 60)
managed.setwatch()
self.rsh(node, "crm_attribute -V -D -n is-managed-default")
if not managed.lookforall():
self.logger.log("Patterns not found: " + repr(managed.unmatched))
self.logger.log("Could not re-enable resource management")
return 0
return 1
def canrunnow(self, node):
'''Return TRUE if we can meaningfully run right now'''
if self.find_ocfs2_resources(node):
self.logger.log("Detach/Reattach scenarios are not possible with OCFS2 services present")
return 0
return 1
def __call__(self, node):
self.incr("calls")
pats = []
managed = self.create_watch(["is-managed-default"], 60)
managed.setwatch()
self.debug("Disable resource management")
self.rsh(node, "crm_attribute -V -n is-managed-default -v false")
if not managed.lookforall():
self.logger.log("Patterns not found: " + repr(managed.unmatched))
return self.failure("Resource management not disabled")
pats = []
pats.append(self.templates["Pat:RscOpOK"] % (".*", "start"))
pats.append(self.templates["Pat:RscOpOK"] % (".*", "stop"))
pats.append(self.templates["Pat:RscOpOK"] % (".*", "promote"))
pats.append(self.templates["Pat:RscOpOK"] % (".*", "demote"))
pats.append(self.templates["Pat:RscOpOK"] % (".*", "migrate"))
watch = self.create_watch(pats, 60, "ShutdownActivity")
watch.setwatch()
self.debug("Shutting down the cluster")
ret = self.stopall(None)
if not ret:
self.debug("Re-enable resource management")
self.rsh(node, "crm_attribute -V -D -n is-managed-default")
return self.failure("Couldn't shut down the cluster")
self.debug("Bringing the cluster back up")
ret = self.startall(None)
time.sleep(5) # allow ping to update the CIB
if not ret:
self.debug("Re-enable resource management")
self.rsh(node, "crm_attribute -V -D -n is-managed-default")
return self.failure("Couldn't restart the cluster")
if self.local_badnews("ResourceActivity:", watch):
self.debug("Re-enable resource management")
self.rsh(node, "crm_attribute -V -D -n is-managed-default")
return self.failure("Resources stopped or started during cluster restart")
watch = self.create_watch(pats, 60, "StartupActivity")
watch.setwatch()
managed = self.create_watch(["is-managed-default"], 60)
managed.setwatch()
self.debug("Re-enable resource management")
self.rsh(node, "crm_attribute -V -D -n is-managed-default")
if not managed.lookforall():
self.logger.log("Patterns not found: " + repr(managed.unmatched))
return self.failure("Resource management not enabled")
self.CM.cluster_stable()
# Ignore actions for STONITH resources
ignore = []
(rc, lines) = self.rsh(node, "crm_resource -c", None)
for line in lines:
if re.search("^Resource", line):
r = AuditResource(self.CM, line)
if r.rclass == "stonith":
self.debug("Ignoring start actions for %s" % r.id)
ignore.append(self.templates["Pat:RscOpOK"] % (r.id, "start_0"))
if self.local_badnews("ResourceActivity:", watch, ignore):
return self.failure("Resources stopped or started after resource management was re-enabled")
return ret
def errorstoignore(self):
'''Return list of errors which should be ignored'''
return [
r"resources were active at shutdown",
]
def is_applicable(self):
if self.Env["Name"] == "crm-lha":
return None
return 1
AllTestClasses.append(Reattach)
class SpecialTest1(CTSTest):
'''Set up a custom test to cause quorum failure issues for Andrew'''
def __init__(self, cm):
CTSTest.__init__(self,cm)
self.name = "SpecialTest1"
self.startall = SimulStartLite(cm)
self.restart1 = RestartTest(cm)
self.stopall = SimulStopLite(cm)
def __call__(self, node):
'''Perform the 'SpecialTest1' test for Andrew. '''
self.incr("calls")
# Shut down all the nodes...
ret = self.stopall(None)
if not ret:
return self.failure("Could not stop all nodes")
# Test config recovery when the other nodes come up
self.rsh(node, "rm -f "+CTSvars.CRM_CONFIG_DIR+"/cib*")
# Start the selected node
ret = self.restart1(node)
if not ret:
return self.failure("Could not start "+node)
# Start all remaining nodes
ret = self.startall(None)
if not ret:
return self.failure("Could not start the remaining nodes")
return self.success()
def errorstoignore(self):
'''Return list of errors which should be ignored'''
# Errors that occur as a result of the CIB being wiped
return [
r"error.*: v1 patchset error, patch failed to apply: Application of an update diff failed",
r"error.*: Resource start-up disabled since no STONITH resources have been defined",
r"error.*: Either configure some or disable STONITH with the stonith-enabled option",
r"error.*: NOTE: Clusters with shared data need STONITH to ensure data integrity",
]
AllTestClasses.append(SpecialTest1)
class HAETest(CTSTest):
'''Set up a custom test to cause quorum failure issues for Andrew'''
def __init__(self, cm):
CTSTest.__init__(self,cm)
self.name = "HAETest"
self.stopall = SimulStopLite(cm)
self.startall = SimulStartLite(cm)
self.is_loop = 1
def setup(self, node):
# Start all remaining nodes
ret = self.startall(None)
if not ret:
return self.failure("Couldn't start all nodes")
return self.success()
def teardown(self, node):
# Stop everything
ret = self.stopall(None)
if not ret:
return self.failure("Couldn't stop all nodes")
return self.success()
def wait_on_state(self, node, resource, expected_clones, attempts=240):
while attempts > 0:
active = 0
(rc, lines) = self.rsh(node, "crm_resource -r %s -W -Q" % resource, stdout=None)
# Hack until crm_resource does the right thing
if rc == 0 and lines:
active = len(lines)
if len(lines) == expected_clones:
return 1
elif rc == 1:
self.debug("Resource %s is still inactive" % resource)
elif rc == 234:
self.logger.log("Unknown resource %s" % resource)
return 0
elif rc == 246:
self.logger.log("Cluster is inactive")
return 0
elif rc != 0:
self.logger.log("Call to crm_resource failed, rc=%d" % rc)
return 0
else:
self.debug("Resource %s is active on %d times instead of %d" % (resource, active, expected_clones))
attempts -= 1
time.sleep(1)
return 0
def find_dlm(self, node):
self.r_dlm = None
(rc, lines) = self.rsh(node, "crm_resource -c", None)
for line in lines:
if re.search("^Resource", line):
r = AuditResource(self.CM, line)
if r.rtype == "controld" and r.parent != "NA":
self.debug("Found dlm: %s" % self.r_dlm)
self.r_dlm = r.parent
return 1
return 0
def find_hae_resources(self, node):
self.r_dlm = None
self.r_o2cb = None
self.r_ocfs2 = []
if self.find_dlm(node):
self.find_ocfs2_resources(node)
def is_applicable(self):
if not self.is_applicable_common():
return 0
if self.Env["Schema"] == "hae":
return 1
return None
class HAERoleTest(HAETest):
def __init__(self, cm):
'''Lars' mount/unmount test for the HA extension. '''
HAETest.__init__(self,cm)
self.name = "HAERoleTest"
def change_state(self, node, resource, target):
rc = self.rsh(node, "crm_resource -V -r %s -p target-role -v %s --meta" % (resource, target))
return rc
def __call__(self, node):
self.incr("calls")
lpc = 0
failed = 0
delay = 2
done = time.time() + self.Env["loop-minutes"]*60
self.find_hae_resources(node)
clone_max = len(self.Env["nodes"])
while time.time() <= done and not failed:
lpc = lpc + 1
self.change_state(node, self.r_dlm, "Stopped")
if not self.wait_on_state(node, self.r_dlm, 0):
self.failure("%s did not go down correctly" % self.r_dlm)
failed = lpc
self.change_state(node, self.r_dlm, "Started")
if not self.wait_on_state(node, self.r_dlm, clone_max):
self.failure("%s did not come up correctly" % self.r_dlm)
failed = lpc
if not self.wait_on_state(node, self.r_o2cb, clone_max):
self.failure("%s did not come up correctly" % self.r_o2cb)
failed = lpc
for fs in self.r_ocfs2:
if not self.wait_on_state(node, fs, clone_max):
self.failure("%s did not come up correctly" % fs)
failed = lpc
if failed:
return self.failure("iteration %d failed" % failed)
return self.success()
AllTestClasses.append(HAERoleTest)
class HAEStandbyTest(HAETest):
'''Set up a custom test to cause quorum failure issues for Andrew'''
def __init__(self, cm):
HAETest.__init__(self,cm)
self.name = "HAEStandbyTest"
def change_state(self, node, resource, target):
rc = self.rsh(node, "crm_standby -V -l reboot -v %s" % (target))
return rc
def __call__(self, node):
self.incr("calls")
lpc = 0
failed = 0
done = time.time() + self.Env["loop-minutes"]*60
self.find_hae_resources(node)
clone_max = len(self.Env["nodes"])
while time.time() <= done and not failed:
lpc = lpc + 1
self.change_state(node, self.r_dlm, "true")
if not self.wait_on_state(node, self.r_dlm, clone_max-1):
self.failure("%s did not go down correctly" % self.r_dlm)
failed = lpc
self.change_state(node, self.r_dlm, "false")
if not self.wait_on_state(node, self.r_dlm, clone_max):
self.failure("%s did not come up correctly" % self.r_dlm)
failed = lpc
if not self.wait_on_state(node, self.r_o2cb, clone_max):
self.failure("%s did not come up correctly" % self.r_o2cb)
failed = lpc
for fs in self.r_ocfs2:
if not self.wait_on_state(node, fs, clone_max):
self.failure("%s did not come up correctly" % fs)
failed = lpc
if failed:
return self.failure("iteration %d failed" % failed)
return self.success()
AllTestClasses.append(HAEStandbyTest)
class NearQuorumPointTest(CTSTest):
'''
This test brings larger clusters near the quorum point (50%).
In addition, it will test doing starts and stops at the same time.
Here is how I think it should work:
- loop over the nodes and decide randomly which will be up and which
will be down Use a 50% probability for each of up/down.
- figure out what to do to get into that state from the current state
- in parallel, bring up those going up and bring those going down.
'''
def __init__(self, cm):
CTSTest.__init__(self,cm)
self.name = "NearQuorumPoint"
def __call__(self, dummy):
'''Perform the 'NearQuorumPoint' test. '''
self.incr("calls")
startset = []
stopset = []
stonith = self.CM.prepare_fencing_watcher("NearQuorumPoint")
#decide what to do with each node
for node in self.Env["nodes"]:
action = self.Env.RandomGen.choice(["start","stop"])
#action = self.Env.RandomGen.choice(["start","stop","no change"])
if action == "start" :
startset.append(node)
elif action == "stop" :
stopset.append(node)
self.debug("start nodes:" + repr(startset))
self.debug("stop nodes:" + repr(stopset))
#add search patterns
watchpats = [ ]
for node in stopset:
if self.CM.ShouldBeStatus[node] == "up":
watchpats.append(self.templates["Pat:We_stopped"] % node)
for node in startset:
if self.CM.ShouldBeStatus[node] == "down":
#watchpats.append(self.templates["Pat:Slave_started"] % node)
watchpats.append(self.templates["Pat:Local_started"] % node)
else:
for stopping in stopset:
if self.CM.ShouldBeStatus[stopping] == "up":
watchpats.append(self.templates["Pat:They_stopped"] % (node, self.CM.key_for_node(stopping)))
if len(watchpats) == 0:
return self.skipped()
if len(startset) != 0:
watchpats.append(self.templates["Pat:DC_IDLE"])
watch = self.create_watch(watchpats, self.Env["DeadTime"]+10)
watch.setwatch()
#begin actions
for node in stopset:
if self.CM.ShouldBeStatus[node] == "up":
self.CM.StopaCMnoBlock(node)
for node in startset:
if self.CM.ShouldBeStatus[node] == "down":
self.CM.StartaCMnoBlock(node)
#get the result
if watch.lookforall():
self.CM.cluster_stable()
self.CM.fencing_cleanup("NearQuorumPoint", stonith)
return self.success()
self.logger.log("Warn: Patterns not found: " + repr(watch.unmatched))
#get the "bad" nodes
upnodes = []
for node in stopset:
if self.CM.StataCM(node) == 1:
upnodes.append(node)
downnodes = []
for node in startset:
if self.CM.StataCM(node) == 0:
downnodes.append(node)
self.CM.fencing_cleanup("NearQuorumPoint", stonith)
if upnodes == [] and downnodes == []:
self.CM.cluster_stable()
# Make sure they're completely down with no residule
for node in stopset:
self.rsh(node, self.templates["StopCmd"])
return self.success()
if len(upnodes) > 0:
self.logger.log("Warn: Unstoppable nodes: " + repr(upnodes))
if len(downnodes) > 0:
self.logger.log("Warn: Unstartable nodes: " + repr(downnodes))
return self.failure()
def is_applicable(self):
if self.Env["Name"] == "crm-cman":
return None
return 1
AllTestClasses.append(NearQuorumPointTest)
class RollingUpgradeTest(CTSTest):
'''Perform a rolling upgrade of the cluster'''
def __init__(self, cm):
CTSTest.__init__(self,cm)
self.name = "RollingUpgrade"
self.start = StartTest(cm)
self.stop = StopTest(cm)
self.stopall = SimulStopLite(cm)
self.startall = SimulStartLite(cm)
def setup(self, node):
# Start all remaining nodes
ret = self.stopall(None)
if not ret:
return self.failure("Couldn't stop all nodes")
for node in self.Env["nodes"]:
if not self.downgrade(node, None):
return self.failure("Couldn't downgrade %s" % node)
ret = self.startall(None)
if not ret:
return self.failure("Couldn't start all nodes")
return self.success()
def teardown(self, node):
# Stop everything
ret = self.stopall(None)
if not ret:
return self.failure("Couldn't stop all nodes")
for node in self.Env["nodes"]:
if not self.upgrade(node, None):
return self.failure("Couldn't upgrade %s" % node)
return self.success()
def install(self, node, version, start=1, flags="--force"):
target_dir = "/tmp/rpm-%s" % version
src_dir = "%s/%s" % (self.Env["rpm-dir"], version)
self.logger.log("Installing %s on %s with %s" % (version, node, flags))
if not self.stop(node):
return self.failure("stop failure: "+node)
rc = self.rsh(node, "mkdir -p %s" % target_dir)
rc = self.rsh(node, "rm -f %s/*.rpm" % target_dir)
(rc, lines) = self.rsh(node, "ls -1 %s/*.rpm" % src_dir, None)
for line in lines:
line = line[:-1]
rc = self.rsh.cp("%s" % (line), "%s:%s/" % (node, target_dir))
rc = self.rsh(node, "rpm -Uvh %s %s/*.rpm" % (flags, target_dir))
if start and not self.start(node):
return self.failure("start failure: "+node)
return self.success()
def upgrade(self, node, start=1):
return self.install(node, self.Env["current-version"], start)
def downgrade(self, node, start=1):
return self.install(node, self.Env["previous-version"], start, "--force --nodeps")
def __call__(self, node):
'''Perform the 'Rolling Upgrade' test. '''
self.incr("calls")
for node in self.Env["nodes"]:
if self.upgrade(node):
return self.failure("Couldn't upgrade %s" % node)
self.CM.cluster_stable()
return self.success()
def is_applicable(self):
if not self.is_applicable_common():
return None
if not self.Env.has_key("rpm-dir"):
return None
if not self.Env.has_key("current-version"):
return None
if not self.Env.has_key("previous-version"):
return None
return 1
# Register RestartTest as a good test to run
AllTestClasses.append(RollingUpgradeTest)
class BSC_AddResource(CTSTest):
'''Add a resource to the cluster'''
def __init__(self, cm):
CTSTest.__init__(self, cm)
self.name = "AddResource"
self.resource_offset = 0
self.cib_cmd = """cibadmin -C -o %s -X '%s' """
def __call__(self, node):
self.incr("calls")
self.resource_offset = self.resource_offset + 1
r_id = "bsc-rsc-%s-%d" % (node, self.resource_offset)
start_pat = "crmd.*%s_start_0.*confirmed.*ok"
patterns = []
patterns.append(start_pat % r_id)
watch = self.create_watch(patterns, self.Env["DeadTime"])
watch.setwatch()
ip = self.NextIP()
if not self.make_ip_resource(node, r_id, "ocf", "IPaddr", ip):
return self.failure("Make resource %s failed" % r_id)
failed = 0
watch_result = watch.lookforall()
if watch.unmatched:
for regex in watch.unmatched:
self.logger.log ("Warn: Pattern not found: %s" % (regex))
failed = 1
if failed:
return self.failure("Resource pattern(s) not found")
if not self.CM.cluster_stable(self.Env["DeadTime"]):
return self.failure("Unstable cluster")
return self.success()
def NextIP(self):
ip = self.Env["IPBase"]
if ":" in ip:
fields = ip.rpartition(":")
fields[2] = str(hex(int(fields[2], 16)+1))
- print str(hex(int(f[2], 16)+1))
+ print(str(hex(int(f[2], 16)+1)))
else:
fields = ip.rpartition('.')
fields[2] = str(int(fields[2])+1)
ip = fields[0] + fields[1] + fields[3];
self.Env["IPBase"] = ip
return ip.strip()
def make_ip_resource(self, node, id, rclass, type, ip):
self.logger.log("Creating %s::%s:%s (%s) on %s" % (rclass,type,id,ip,node))
rsc_xml="""
<primitive id="%s" class="%s" type="%s" provider="heartbeat">
<instance_attributes id="%s"><attributes>
<nvpair id="%s" name="ip" value="%s"/>
</attributes></instance_attributes>
</primitive>""" % (id, rclass, type, id, id, ip)
node_constraint = """
<rsc_location id="run_%s" rsc="%s">
<rule id="pref_run_%s" score="100">
<expression id="%s_loc_expr" attribute="#uname" operation="eq" value="%s"/>
</rule>
</rsc_location>""" % (id, id, id, id, node)
rc = 0
(rc, lines) = self.rsh(node, self.cib_cmd % ("constraints", node_constraint), None)
if rc != 0:
self.logger.log("Constraint creation failed: %d" % rc)
return None
(rc, lines) = self.rsh(node, self.cib_cmd % ("resources", rsc_xml), None)
if rc != 0:
self.logger.log("Resource creation failed: %d" % rc)
return None
return 1
def is_applicable(self):
if self.Env["DoBSC"]:
return 1
return None
AllTestClasses.append(BSC_AddResource)
class SimulStopLite(CTSTest):
'''Stop any active nodes ~ simultaneously'''
def __init__(self, cm):
CTSTest.__init__(self,cm)
self.name = "SimulStopLite"
def __call__(self, dummy):
'''Perform the 'SimulStopLite' setup work. '''
self.incr("calls")
self.debug("Setup: " + self.name)
# We ignore the "node" parameter...
watchpats = [ ]
for node in self.Env["nodes"]:
if self.CM.ShouldBeStatus[node] == "up":
self.incr("WasStarted")
watchpats.append(self.templates["Pat:We_stopped"] % node)
#if self.Env["use_logd"]:
# watchpats.append(self.templates["Pat:Logd_stopped"] % node)
if len(watchpats) == 0:
self.CM.clear_all_caches()
return self.success()
# Stop all the nodes - at about the same time...
watch = self.create_watch(watchpats, self.Env["DeadTime"]+10)
watch.setwatch()
self.set_timer()
for node in self.Env["nodes"]:
if self.CM.ShouldBeStatus[node] == "up":
self.CM.StopaCMnoBlock(node)
if watch.lookforall():
self.CM.clear_all_caches()
# Make sure they're completely down with no residule
for node in self.Env["nodes"]:
self.rsh(node, self.templates["StopCmd"])
return self.success()
did_fail = 0
up_nodes = []
for node in self.Env["nodes"]:
if self.CM.StataCM(node) == 1:
did_fail = 1
up_nodes.append(node)
if did_fail:
return self.failure("Active nodes exist: " + repr(up_nodes))
self.logger.log("Warn: All nodes stopped but CTS didnt detect: "
+ repr(watch.unmatched))
self.CM.clear_all_caches()
return self.failure("Missing log message: "+repr(watch.unmatched))
def is_applicable(self):
'''SimulStopLite is a setup test and never applicable'''
return 0
class SimulStartLite(CTSTest):
'''Start any stopped nodes ~ simultaneously'''
def __init__(self, cm):
CTSTest.__init__(self,cm)
self.name = "SimulStartLite"
def __call__(self, dummy):
'''Perform the 'SimulStartList' setup work. '''
self.incr("calls")
self.debug("Setup: " + self.name)
# We ignore the "node" parameter...
node_list = []
for node in self.Env["nodes"]:
if self.CM.ShouldBeStatus[node] == "down":
self.incr("WasStopped")
node_list.append(node)
self.set_timer()
while len(node_list) > 0:
# Repeat until all nodes come up
watchpats = [ ]
uppat = self.templates["Pat:Slave_started"]
if self.CM.upcount() == 0:
uppat = self.templates["Pat:Local_started"]
watchpats.append(self.templates["Pat:DC_IDLE"])
for node in node_list:
watchpats.append(uppat % node)
watchpats.append(self.templates["Pat:InfraUp"] % node)
watchpats.append(self.templates["Pat:PacemakerUp"] % node)
# Start all the nodes - at about the same time...
watch = self.create_watch(watchpats, self.Env["DeadTime"]+10)
watch.setwatch()
stonith = self.CM.prepare_fencing_watcher(self.name)
for node in node_list:
self.CM.StartaCMnoBlock(node)
watch.lookforall()
node_list = self.CM.fencing_cleanup(self.name, stonith)
# Remove node_list messages from watch.unmatched
for node in node_list:
self.logger.debug("Dealing with stonith operations for %s" % repr(node_list))
if watch.unmatched:
try:
watch.unmatched.remove(uppat % node)
except:
self.debug("Already matched: %s" % (uppat % node))
try:
watch.unmatched.remove(self.templates["Pat:InfraUp"] % node)
except:
self.debug("Already matched: %s" % (self.templates["Pat:InfraUp"] % node))
try:
watch.unmatched.remove(self.templates["Pat:PacemakerUp"] % node)
except:
self.debug("Already matched: %s" % (self.templates["Pat:PacemakerUp"] % node))
if watch.unmatched:
for regex in watch.unmatched:
self.logger.log ("Warn: Startup pattern not found: %s" %(regex))
if not self.CM.cluster_stable():
return self.failure("Cluster did not stabilize")
did_fail = 0
unstable = []
for node in self.Env["nodes"]:
if self.CM.StataCM(node) == 0:
did_fail = 1
unstable.append(node)
if did_fail:
return self.failure("Unstarted nodes exist: " + repr(unstable))
unstable = []
for node in self.Env["nodes"]:
if not self.CM.node_stable(node):
did_fail = 1
unstable.append(node)
if did_fail:
return self.failure("Unstable cluster nodes exist: " + repr(unstable))
return self.success()
def is_applicable(self):
'''SimulStartLite is a setup test and never applicable'''
return 0
def TestList(cm, audits):
result = []
for testclass in AllTestClasses:
bound_test = testclass(cm)
if bound_test.is_applicable():
bound_test.Audits = audits
result.append(bound_test)
return result
class RemoteLXC(CTSTest):
def __init__(self, cm):
CTSTest.__init__(self,cm)
self.name = "RemoteLXC"
self.start = StartTest(cm)
self.startall = SimulStartLite(cm)
self.num_containers = 2
self.is_container = 1
self.is_docker_unsafe = 1
self.failed = 0
self.fail_string = ""
def start_lxc_simple(self, node):
# restore any artifacts laying around from a previous test.
self.rsh(node, "/usr/share/pacemaker/tests/cts/lxc_autogen.sh -R &>/dev/null")
# generate the containers, put them in the config, add some resources to them
pats = [ ]
watch = self.create_watch(pats, 120)
watch.setwatch()
pats.append(self.templates["Pat:RscOpOK"] % ("lxc1", "start_0"))
pats.append(self.templates["Pat:RscOpOK"] % ("lxc2", "start_0"))
pats.append(self.templates["Pat:RscOpOK"] % ("lxc-ms", "start_0"))
pats.append(self.templates["Pat:RscOpOK"] % ("lxc-ms", "promote_0"))
self.rsh(node, "/usr/share/pacemaker/tests/cts/lxc_autogen.sh -g -a -m -s -c %d &>/dev/null" % self.num_containers)
self.set_timer("remoteSimpleInit")
watch.lookforall()
self.log_timer("remoteSimpleInit")
if watch.unmatched:
self.fail_string = "Unmatched patterns: %s" % (repr(watch.unmatched))
self.failed = 1
def cleanup_lxc_simple(self, node):
pats = [ ]
# if the test failed, attempt to clean up the cib and libvirt environment
# as best as possible
if self.failed == 1:
# restore libvirt and cib
self.rsh(node, "/usr/share/pacemaker/tests/cts/lxc_autogen.sh -R &>/dev/null")
self.rsh(node, "crm_resource -C -r container1 &>/dev/null")
self.rsh(node, "crm_resource -C -r container2 &>/dev/null")
self.rsh(node, "crm_resource -C -r lxc1 &>/dev/null")
self.rsh(node, "crm_resource -C -r lxc2 &>/dev/null")
self.rsh(node, "crm_resource -C -r lxc-ms &>/dev/null")
time.sleep(20)
return
watch = self.create_watch(pats, 120)
watch.setwatch()
pats.append(self.templates["Pat:RscOpOK"] % ("container1", "stop_0"))
pats.append(self.templates["Pat:RscOpOK"] % ("container2", "stop_0"))
self.rsh(node, "/usr/share/pacemaker/tests/cts/lxc_autogen.sh -p &>/dev/null")
self.set_timer("remoteSimpleCleanup")
watch.lookforall()
self.log_timer("remoteSimpleCleanup")
if watch.unmatched:
self.fail_string = "Unmatched patterns: %s" % (repr(watch.unmatched))
self.failed = 1
# cleanup libvirt
self.rsh(node, "/usr/share/pacemaker/tests/cts/lxc_autogen.sh -R &>/dev/null")
def __call__(self, node):
'''Perform the 'RemoteLXC' test. '''
self.incr("calls")
ret = self.startall(None)
if not ret:
return self.failure("Setup failed, start all nodes failed.")
rc = self.rsh(node, "/usr/share/pacemaker/tests/cts/lxc_autogen.sh -v &>/dev/null")
if rc == 1:
self.log("Environment test for lxc support failed.")
return self.skipped()
self.start_lxc_simple(node)
self.cleanup_lxc_simple(node)
self.debug("Waiting for the cluster to recover")
self.CM.cluster_stable()
if self.failed == 1:
return self.failure(self.fail_string)
return self.success()
def errorstoignore(self):
'''Return list of errors which should be ignored'''
return [
r"Updating failcount for ping",
r"pengine.*: Recover (ping|lxc-ms|container)\s*\(.*\)",
# The orphaned lxc-ms resource causes an expected transition error
# that is a result of the pengine not having knowledge that the
# ms resource used to be a clone. As a result it looks like that
# resource is running in multiple locations when it shouldn't... But in
# this instance we know why this error is occurring and that it is expected.
r"Calculated Transition .* /var/lib/pacemaker/pengine/pe-error",
r"Resource lxc-ms .* is active on 2 nodes attempting recovery",
r"Unknown operation: fail",
r"(ERROR|error): sending stonithRA op to stonithd failed.",
]
AllTestClasses.append(RemoteLXC)
###################################################################
class RemoteDriver(CTSTest):
###################################################################
def __init__(self, cm):
CTSTest.__init__(self,cm)
self.name = "RemoteDriver"
self.is_docker_unsafe = 1
self.start = StartTest(cm)
self.startall = SimulStartLite(cm)
self.stop = StopTest(cm)
self.pcmk_started = 0
self.failed = 0
self.fail_string = ""
self.remote_node_added = 0
self.remote_rsc_added = 0
self.remote_rsc = "remote-rsc"
self.remote_use_reconnect_interval = self.Env.RandomGen.choice(["true","false"])
self.cib_cmd = """cibadmin -C -o %s -X '%s' """
def get_othernode(self, node):
for othernode in self.Env["nodes"]:
if othernode == node:
# we don't want to try and use the cib that we just shutdown.
# find a cluster node that is not our soon to be remote-node.
continue
else:
return othernode
def del_rsc(self, node, rsc):
othernode = self.get_othernode(node)
rc = self.rsh(othernode, "crm_resource -D -r %s -t primitive" % (rsc))
if rc != 0:
self.fail_string = ("Removal of resource '%s' failed" % (rsc))
self.failed = 1
def add_rsc(self, node, rsc_xml):
othernode = self.get_othernode(node)
rc = self.rsh(othernode, self.cib_cmd % ("resources", rsc_xml))
if rc != 0:
self.fail_string = "resource creation failed"
self.failed = 1
def add_primitive_rsc(self, node):
rsc_xml = """
<primitive class="ocf" id="%s" provider="heartbeat" type="Dummy">
<operations>
<op id="remote-rsc-monitor-interval-10s" interval="10s" name="monitor"/>
</operations>
<meta_attributes id="remote-meta_attributes"/>
</primitive>""" % (self.remote_rsc)
self.add_rsc(node, rsc_xml)
if self.failed == 0:
self.remote_rsc_added = 1
def add_connection_rsc(self, node):
if self.remote_use_reconnect_interval == "true":
# use reconnect interval and make sure to set cluster-recheck-interval as well.
rsc_xml = """
<primitive class="ocf" id="%s" provider="pacemaker" type="remote">
<instance_attributes id="remote-instance_attributes"/>
<instance_attributes id="remote-instance_attributes">
<nvpair id="remote-instance_attributes-server" name="server" value="%s"/>
<nvpair id="remote-instance_attributes-reconnect_interval" name="reconnect_interval" value="60s"/>
</instance_attributes>
<operations>
<op id="remote-monitor-interval-60s" interval="60s" name="monitor"/>
<op id="remote-name-start-interval-0-timeout-120" interval="0" name="start" timeout="60"/>
</operations>
</primitive>""" % (self.remote_node, node)
self.rsh(self.get_othernode(node), self.templates["SetCheckInterval"] % ("45s"))
else:
# not using reconnect interval
rsc_xml = """
<primitive class="ocf" id="%s" provider="pacemaker" type="remote">
<instance_attributes id="remote-instance_attributes"/>
<instance_attributes id="remote-instance_attributes">
<nvpair id="remote-instance_attributes-server" name="server" value="%s"/>
</instance_attributes>
<operations>
<op id="remote-monitor-interval-60s" interval="60s" name="monitor"/>
<op id="remote-name-start-interval-0-timeout-120" interval="0" name="start" timeout="120"/>
</operations>
</primitive>""" % (self.remote_node, node)
self.add_rsc(node, rsc_xml)
if self.failed == 0:
self.remote_node_added = 1
def stop_pcmk_remote(self, node):
# disable pcmk remote
for i in range(10):
rc = self.rsh(node, "service pacemaker_remote stop")
if rc != 0:
time.sleep(6)
else:
break
def start_pcmk_remote(self, node):
for i in range(10):
rc = self.rsh(node, "service pacemaker_remote start")
if rc != 0:
time.sleep(6)
else:
self.pcmk_started = 1
break
def start_metal(self, node):
pcmk_started = 0
# make sure the resource doesn't already exist for some reason
self.rsh(node, "crm_resource -D -r %s -t primitive" % (self.remote_rsc))
self.rsh(node, "crm_resource -D -r %s -t primitive" % (self.remote_node))
if not self.stop(node):
self.failed = 1
self.fail_string = "Failed to shutdown cluster node %s" % (node)
return
self.start_pcmk_remote(node)
if self.pcmk_started == 0:
self.failed = 1
self.fail_string = "Failed to start pacemaker_remote on node %s" % (node)
return
# convert node to baremetal node now that it has shutdow the cluster stack
pats = [ ]
watch = self.create_watch(pats, 120)
watch.setwatch()
pats.append(self.templates["Pat:RscOpOK"] % (self.remote_node, "start"))
pats.append(self.templates["Pat:DC_IDLE"])
self.add_connection_rsc(node)
self.set_timer("remoteMetalInit")
watch.lookforall()
self.log_timer("remoteMetalInit")
if watch.unmatched:
self.fail_string = "Unmatched patterns: %s" % (repr(watch.unmatched))
self.failed = 1
def migrate_connection(self, node):
if self.failed == 1:
return
pats = [ ]
pats.append(self.templates["Pat:RscOpOK"] % (self.remote_node, "migrate_to"))
pats.append(self.templates["Pat:RscOpOK"] % (self.remote_node, "migrate_from"))
pats.append(self.templates["Pat:DC_IDLE"])
watch = self.create_watch(pats, 120)
watch.setwatch()
(rc, lines) = self.rsh(node, "crm_resource -M -r %s" % (self.remote_node), None)
if rc != 0:
self.fail_string = "failed to move remote node connection resource"
self.logger.log(self.fail_string)
self.failed = 1
return
self.set_timer("remoteMetalMigrate")
watch.lookforall()
self.log_timer("remoteMetalMigrate")
if watch.unmatched:
self.fail_string = "Unmatched patterns: %s" % (repr(watch.unmatched))
self.logger.log(self.fail_string)
self.failed = 1
return
def fail_rsc(self, node):
if self.failed == 1:
return
watchpats = [ ]
watchpats.append(self.templates["Pat:RscRemoteOpOK"] % (self.remote_rsc, "stop", self.remote_node))
watchpats.append(self.templates["Pat:RscRemoteOpOK"] % (self.remote_rsc, "start", self.remote_node))
watchpats.append(self.templates["Pat:DC_IDLE"])
watch = self.create_watch(watchpats, 120)
watch.setwatch()
self.debug("causing dummy rsc to fail.")
rc = self.rsh(node, "rm -f /var/run/resource-agents/Dummy*")
self.set_timer("remoteRscFail")
watch.lookforall()
self.log_timer("remoteRscFail")
if watch.unmatched:
self.fail_string = "Unmatched patterns during rsc fail: %s" % (repr(watch.unmatched))
self.logger.log(self.fail_string)
self.failed = 1
def fail_connection(self, node):
if self.failed == 1:
return
watchpats = [ ]
watchpats.append(self.templates["Pat:FenceOpOK"] % self.remote_node)
watchpats.append(self.templates["Pat:NodeFenced"] % self.remote_node)
watch = self.create_watch(watchpats, 120)
watch.setwatch()
# force stop the pcmk remote daemon. this will result in fencing
self.debug("Force stopped active remote node")
self.stop_pcmk_remote(node)
self.debug("Waiting for remote node to be fenced.")
self.set_timer("remoteMetalFence")
watch.lookforall()
self.log_timer("remoteMetalFence")
if watch.unmatched:
self.fail_string = "Unmatched patterns: %s" % (repr(watch.unmatched))
self.logger.log(self.fail_string)
self.failed = 1
return
self.debug("Waiting for the remote node to come back up")
self.CM.ns.WaitForNodeToComeUp(node, 120);
pats = [ ]
watch = self.create_watch(pats, 200)
watch.setwatch()
pats.append(self.templates["Pat:RscOpOK"] % (self.remote_node, "start"))
if self.remote_rsc_added == 1:
pats.append(self.templates["Pat:RscOpOK"] % (self.remote_rsc, "monitor"))
# start the remote node again watch it integrate back into cluster.
self.start_pcmk_remote(node)
if self.pcmk_started == 0:
self.failed = 1
self.fail_string = "Failed to start pacemaker_remote on node %s" % (node)
self.logger.log(self.fail_string)
return
self.debug("Waiting for remote node to rejoin cluster after being fenced.")
self.set_timer("remoteMetalRestart")
watch.lookforall()
self.log_timer("remoteMetalRestart")
if watch.unmatched:
self.fail_string = "Unmatched patterns: %s" % (repr(watch.unmatched))
self.failed = 1
self.logger.log(self.fail_string)
return
def add_dummy_rsc(self, node):
if self.failed == 1:
return
# verify we can put a resource on the remote node
pats = [ ]
watch = self.create_watch(pats, 120)
watch.setwatch()
pats.append(self.templates["Pat:RscRemoteOpOK"] % (self.remote_rsc, "start", self.remote_node))
pats.append(self.templates["Pat:DC_IDLE"])
# Add a resource that must live on remote-node
self.add_primitive_rsc(node)
# force that rsc to prefer the remote node.
(rc, line) = self.CM.rsh(node, "crm_resource -M -r %s -N %s -f" % (self.remote_rsc, self.remote_node), None)
if rc != 0:
self.fail_string = "Failed to place remote resource on remote node."
self.failed = 1
return
self.set_timer("remoteMetalRsc")
watch.lookforall()
self.log_timer("remoteMetalRsc")
if watch.unmatched:
self.fail_string = "Unmatched patterns: %s" % (repr(watch.unmatched))
self.failed = 1
def test_attributes(self, node):
if self.failed == 1:
return
# This verifies permanent attributes can be set on a remote-node. It also
# verifies the remote-node can edit it's own cib node section remotely.
(rc, line) = self.CM.rsh(node, "crm_attribute -l forever -n testattr -v testval -N %s" % (self.remote_node), None)
if rc != 0:
self.fail_string = "Failed to set remote-node attribute. rc:%s output:%s" % (rc, line)
self.failed = 1
return
(rc, line) = self.CM.rsh(node, "crm_attribute -l forever -n testattr -Q -N %s" % (self.remote_node), None)
if rc != 0:
self.fail_string = "Failed to get remote-node attribute"
self.failed = 1
return
(rc, line) = self.CM.rsh(node, "crm_attribute -l forever -n testattr -D -N %s" % (self.remote_node), None)
if rc != 0:
self.fail_string = "Failed to delete remote-node attribute"
self.failed = 1
return
def cleanup_metal(self, node):
if self.pcmk_started == 0:
return
pats = [ ]
watch = self.create_watch(pats, 120)
watch.setwatch()
if self.remote_rsc_added == 1:
pats.append(self.templates["Pat:RscOpOK"] % (self.remote_rsc, "stop"))
if self.remote_node_added == 1:
pats.append(self.templates["Pat:RscOpOK"] % (self.remote_node, "stop"))
self.set_timer("remoteMetalCleanup")
if self.remote_use_reconnect_interval == "true":
self.debug("Cleaning up re-check interval")
self.rsh(self.get_othernode(node), self.templates["ClearCheckInterval"])
if self.remote_rsc_added == 1:
self.debug("Cleaning up dummy rsc put on remote node")
self.rsh(node, "crm_resource -U -r %s -N %s" % (self.remote_rsc, self.remote_node))
self.del_rsc(node, self.remote_rsc)
if self.remote_node_added == 1:
self.debug("Cleaning up remote node connection resource")
self.rsh(node, "crm_resource -U -r %s" % (self.remote_node))
self.del_rsc(node, self.remote_node)
watch.lookforall()
self.log_timer("remoteMetalCleanup")
if watch.unmatched:
self.fail_string = "Unmatched patterns: %s" % (repr(watch.unmatched))
self.failed = 1
self.stop_pcmk_remote(node)
def setup_env(self, node):
self.remote_node = "remote_%s" % (node)
sync_key = 0
# we are assuming if all nodes have a key, that it is
# the right key... If any node doesn't have a remote
# key, we regenerate it everywhere.
for node in self.Env["nodes"]:
rc = self.rsh(node, "ls /etc/pacemaker/authkey")
if rc != 0:
sync_key = 1
break
if sync_key == 0:
return
# create key locally
os.system("/usr/share/pacemaker/tests/cts/lxc_autogen.sh -k &> /dev/null")
# sync key throughout the cluster
for node in self.Env["nodes"]:
rc = self.rsh(node, "mkdir /etc/pacemaker")
self.rsh.cp("/etc/pacemaker/authkey", "%s:/etc/pacemaker/authkey" % (node))
def is_applicable(self):
if not self.is_applicable_common():
return False
for node in self.Env["nodes"]:
rc = self.rsh(node, "type pacemaker_remoted >/dev/null 2>&1")
if rc != 0:
return False
return True
def __call__(self, node):
'''Perform the 'RemoteBaremetal' test. '''
self.incr("calls")
ret = self.startall(None)
if not ret:
return self.failure("Setup failed, start all nodes failed.")
self.setup_env(node)
self.start_metal(node)
self.add_dummy_rsc(node)
self.test_attributes(node)
self.cleanup_metal(node)
self.debug("Waiting for the cluster to recover")
self.CM.cluster_stable()
if self.failed == 1:
return self.failure(self.fail_string)
return self.success()
def errorstoignore(self):
'''Return list of errors which should be ignored'''
return [ """is running on remote.*which isn't allowed""",
"""Connection terminated""",
"""Failed to send remote""",
]
# Remote driver is called by other tests.
###################################################################
class RemoteBasic(CTSTest):
###################################################################
def __init__(self, cm):
CTSTest.__init__(self,cm)
self.name = "RemoteBasic"
self.start = StartTest(cm)
self.startall = SimulStartLite(cm)
self.driver = RemoteDriver(cm)
self.is_docker_unsafe = 1
def __call__(self, node):
'''Perform the 'RemoteBaremetal' test. '''
self.incr("calls")
ret = self.startall(None)
if not ret:
return self.failure("Setup failed, start all nodes failed.")
self.driver.setup_env(node)
self.driver.start_metal(node)
self.driver.add_dummy_rsc(node)
self.driver.test_attributes(node)
self.driver.cleanup_metal(node)
self.debug("Waiting for the cluster to recover")
self.CM.cluster_stable()
if self.driver.failed == 1:
return self.failure(self.driver.fail_string)
return self.success()
def is_applicable(self):
return self.driver.is_applicable()
def errorstoignore(self):
return self.driver.errorstoignore()
AllTestClasses.append(RemoteBasic)
###################################################################
class RemoteStonithd(CTSTest):
###################################################################
def __init__(self, cm):
CTSTest.__init__(self,cm)
self.name = "RemoteStonithd"
self.start = StartTest(cm)
self.startall = SimulStartLite(cm)
self.driver = RemoteDriver(cm)
self.is_docker_unsafe = 1
def __call__(self, node):
'''Perform the 'RemoteStonithd' test. '''
self.incr("calls")
ret = self.startall(None)
if not ret:
return self.failure("Setup failed, start all nodes failed.")
self.driver.setup_env(node)
self.driver.start_metal(node)
self.driver.add_dummy_rsc(node)
self.driver.fail_connection(node)
self.driver.cleanup_metal(node)
self.debug("Waiting for the cluster to recover")
self.CM.cluster_stable()
if self.driver.failed == 1:
return self.failure(self.driver.fail_string)
return self.success()
def is_applicable(self):
if not self.driver.is_applicable():
return False
if self.Env.has_key("DoFencing"):
return self.Env["DoFencing"]
return True
def errorstoignore(self):
ignore_pats = [
r"Unexpected disconnect on remote-node",
r"crmd.*: error.*: Operation remote_.*_monitor",
r"pengine.*: Recover remote_.*\s*\(.*\)",
r"Calculated Transition .* /var/lib/pacemaker/pengine/pe-error",
r"error.*: Resource .*ocf::.* is active on 2 nodes attempting recovery",
]
ignore_pats.extend(self.driver.errorstoignore())
return ignore_pats
AllTestClasses.append(RemoteStonithd)
###################################################################
class RemoteMigrate(CTSTest):
###################################################################
def __init__(self, cm):
CTSTest.__init__(self,cm)
self.name = "RemoteMigrate"
self.start = StartTest(cm)
self.startall = SimulStartLite(cm)
self.driver = RemoteDriver(cm)
self.is_docker_unsafe = 1
def __call__(self, node):
'''Perform the 'RemoteMigrate' test. '''
self.incr("calls")
ret = self.startall(None)
if not ret:
return self.failure("Setup failed, start all nodes failed.")
self.driver.setup_env(node)
self.driver.start_metal(node)
self.driver.add_dummy_rsc(node)
self.driver.migrate_connection(node)
self.driver.cleanup_metal(node)
self.debug("Waiting for the cluster to recover")
self.CM.cluster_stable()
if self.driver.failed == 1:
return self.failure(self.driver.fail_string)
return self.success()
def is_applicable(self):
return self.driver.is_applicable()
def errorstoignore(self):
return self.driver.errorstoignore()
AllTestClasses.append(RemoteMigrate)
###################################################################
class RemoteRscFailure(CTSTest):
###################################################################
def __init__(self, cm):
# fail a rsc on a remote node, verify recovery.
CTSTest.__init__(self,cm)
self.name = "RemoteRscFailure"
self.start = StartTest(cm)
self.startall = SimulStartLite(cm)
self.driver = RemoteDriver(cm)
self.is_docker_unsafe = 1
def __call__(self, node):
'''Perform the 'RemoteRscFailure' test. '''
self.incr("calls")
ret = self.startall(None)
if not ret:
return self.failure("Setup failed, start all nodes failed.")
self.driver.setup_env(node)
self.driver.start_metal(node)
self.driver.add_dummy_rsc(node)
# This is an important step. We are migrating the connection
# before failing the resource. This verifies that the migration
# has properly maintained control over the remote-node.
self.driver.migrate_connection(node)
self.driver.fail_rsc(node)
self.driver.cleanup_metal(node)
self.debug("Waiting for the cluster to recover")
self.CM.cluster_stable()
if self.driver.failed == 1:
return self.failure(self.driver.fail_string)
return self.success()
def is_applicable(self):
return self.driver.is_applicable()
def errorstoignore(self):
ignore_pats = [
r"pengine.*: Recover remote-rsc\s*\(.*\)",
]
ignore_pats.extend(self.driver.errorstoignore())
return ignore_pats
AllTestClasses.append(RemoteRscFailure)
# vim:ts=4:sw=4:et:
diff --git a/cts/OCFIPraTest.py b/cts/OCFIPraTest.py
index 9900a620f2..03d964b059 100755
--- a/cts/OCFIPraTest.py
+++ b/cts/OCFIPraTest.py
@@ -1,180 +1,180 @@
#!/usr/bin/python
'''OCF IPaddr/IPaddr2 Resource Agent Test'''
__copyright__ = '''
Author: Huang Zhen <zhenhltc@cn.ibm.com>
Copyright (C) 2004 International Business Machines
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., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
import string, sys, struct, os, random, time, syslog
from cts.CTSvars import *
def usage():
- print "usage: " + sys.argv[0] \
+ print("usage: " + sys.argv[0] \
+ " [-2]"\
+ " [--ipbase|-i first-test-ip]"\
+ " [--ipnum|-n test-ip-num]"\
+ " [--help|-h]"\
+ " [--perform|-p op]"\
- + " [number-of-iterations]"
+ + " [number-of-iterations]")
sys.exit(1)
def perform_op(ra, ip, op):
os.environ["OCF_RA_VERSION_MAJOR"] = "1"
os.environ["OCF_RA_VERSION_MINOR"] = "0"
os.environ["OCF_ROOT"] = CTSvars.OCF_ROOT_DIR
os.environ["OCF_RESOURCE_INSTANCE"] = ip
os.environ["OCF_RESOURCE_TYPE"] = ra
os.environ["OCF_RESKEY_ip"] = ip
os.environ["HA_LOGFILE"] = "/dev/null"
os.environ["HA_LOGFACILITY"] = "local7"
path = CTSvars.OCF_ROOT_DIR + "/resource.d/heartbeat/" + ra
return os.spawnvpe(os.P_WAIT, path, [ra, op], os.environ)
def audit(ra, iplist, ipstatus, summary):
passed = 1
for ip in iplist:
ret = perform_op(ra, ip, "monitor")
if ret != ipstatus[ip]:
passed = 0
log("audit: status of %s should be %d but it is %d\t [failure]" %
(ip,ipstatus[ip],ret))
ipstatus[ip] = ret
summary["audit"]["called"] += 1;
if passed :
summary["audit"]["success"] += 1
else :
summary["audit"]["failure"] += 1
def log(towrite):
t = time.strftime("%Y/%m/%d_%H:%M:%S\t", time.localtime(time.time()))
logstr = t + " "+str(towrite)
syslog.syslog(logstr)
- print logstr
+ print(logstr)
if __name__ == '__main__':
ra = "IPaddr"
ipbase = "127.0.0.10"
ipnum = 1
itnum = 50
perform = None
summary = {
"start":{"called":0,"success":0,"failure":0},
"stop" :{"called":0,"success":0,"failure":0},
"audit":{"called":0,"success":0,"failure":0}
}
syslog.openlog(sys.argv[0], 0, syslog.LOG_LOCAL7)
# Process arguments...
skipthis = None
args = sys.argv[1:]
for i in range(0, len(args)) :
if skipthis :
skipthis = None
continue
elif args[i] == "-2" :
ra = "IPaddr2"
elif args[i] == "--ip" or args[i] == "-i" :
skipthis = 1
ipbase = args[i+1]
elif args[i] == "--ipnum" or args[i] == "-n" :
skipthis = 1
ipnum = int(args[i+1])
elif args[i] == "--perform" or args[i] == "-p" :
skipthis = 1
perform = args[i+1]
elif args[i] == "--help" or args[i] == "-h" :
usage()
else:
itnum = int(args[i])
log("Begin OCF IPaddr/IPaddr2 Test")
# Generate the test ips
iplist = []
ipstatus = {}
fields = string.split(ipbase, '.')
for i in range(0, ipnum) :
ip = string.join(fields, '.')
iplist.append(ip)
ipstatus[ip] = perform_op(ra,ip,"monitor")
fields[3] = str(int(fields[3])+1)
log("Test ip:" + str(iplist))
# If use ask perform an operation
if perform != None:
log("Perform opeartion %s"%perform)
for ip in iplist:
perform_op(ra, ip, perform)
log("Done")
sys.exit()
log("RA Type:" + ra)
log("Test Count:" + str(itnum))
# Prepare Random
f = open("/dev/urandom", "r")
seed = struct.unpack("BBB", f.read(3))
f.close()
#seed=(123,321,231)
rand = random.Random()
rand.seed(seed[0])
log("Test Random Seed:" + str(seed))
#
# Begin Tests
log(">>>>>>>>>>>>>>>>>>>>>>>>")
for i in range(0, itnum):
ip = rand.choice(iplist)
if ipstatus[ip] == 0:
op = "stop"
elif ipstatus[ip] == 7:
op = "start"
else :
op = rand.choice(["start","stop"])
ret = perform_op(ra, ip, op)
# update status
if op == "start" and ret == 0:
ipstatus[ip] = 0
elif op == "stop" and ret == 0:
ipstatus[ip] = 7
else :
ipstatus[ip] = 1
result = ""
if ret == 0:
result = "success"
else :
result = "failure"
summary[op]["called"] += 1
summary[op][result] += 1
log( "%d:%s %s \t[%s]"%(i, op, ip, result))
audit(ra, iplist, ipstatus, summary)
log("<<<<<<<<<<<<<<<<<<<<<<<<")
log("start:\t" + str(summary["start"]))
log("stop: \t" + str(summary["stop"]))
log("audit:\t" + str(summary["audit"]))
diff --git a/cts/environment.py b/cts/environment.py
index 61d421187b..a028fcd0ee 100644
--- a/cts/environment.py
+++ b/cts/environment.py
@@ -1,682 +1,682 @@
'''
Classes related to producing and searching logs
'''
__copyright__='''
Copyright (C) 2014 Andrew Beekhof <andrew@beekhof.net>
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., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
import sys, time, os, socket, random
from cts.remote import *
class Environment:
def __init__(self, args):
self.data = {}
self.Nodes = []
self["DeadTime"] = 300
self["StartTime"] = 300
self["StableTime"] = 30
self["tests"] = []
self["IPagent"] = "IPaddr2"
self["DoStandby"] = 1
self["DoFencing"] = 1
self["XmitLoss"] = "0.0"
self["RecvLoss"] = "0.0"
self["ClobberCIB"] = 0
self["CIBfilename"] = None
self["CIBResource"] = 0
self["DoBSC"] = 0
self["use_logd"] = 0
self["oprofile"] = []
self["warn-inactive"] = 0
self["ListTests"] = 0
self["benchmark"] = 0
self["LogWatcher"] = "any"
self["SyslogFacility"] = "daemon"
self["LogFileName"] = "/var/log/messages"
self["Schema"] = "pacemaker-2.0"
self["Stack"] = "corosync"
self["stonith-type"] = "external/ssh"
self["stonith-params"] = "hostlist=all,livedangerously=yes"
self["loop-minutes"] = 60
self["valgrind-prefix"] = None
self["valgrind-procs"] = "attrd cib crmd lrmd pengine stonith-ng"
self["valgrind-opts"] = """--leak-check=full --show-reachable=yes --trace-children=no --num-callers=25 --gen-suppressions=all --suppressions="""+CTSvars.CTS_home+"""/cts.supp"""
self["experimental-tests"] = 0
self["container-tests"] = 0
self["valgrind-tests"] = 0
self["unsafe-tests"] = 1
self["loop-tests"] = 1
self["scenario"] = "random"
self["stats"] = 0
self["docker"] = 0
self.RandomGen = random.Random()
self.logger = LogFactory()
self.SeedRandom()
self.rsh = RemoteFactory().getInstance()
self.target = "localhost"
self.parse_args(args)
self.discover()
self.validate()
def SeedRandom(self, seed=None):
if not seed:
seed = int(time.time())
self["RandSeed"] = seed
self.RandomGen.seed(str(seed))
def dump(self):
keys = []
for key in self.data.keys():
keys.append(key)
keys.sort()
for key in keys:
self.logger.debug("Environment["+key+"]:\t"+str(self[key]))
def keys(self):
return self.data.keys()
def has_key(self, key):
if key == "nodes":
return True
return self.data.has_key(key)
def __getitem__(self, key):
if key == "nodes":
return self.Nodes
elif key == "Name":
return self.get_stack_short()
elif self.data.has_key(key):
return self.data[key]
else:
return None
def __setitem__(self, key, value):
if key == "Stack":
self.set_stack(value)
elif key == "node-limit":
self.data[key] = value
self.filter_nodes()
elif key == "nodes":
self.Nodes = []
for node in value:
# I don't think I need the IP address, etc. but this validates
# the node name against /etc/hosts and/or DNS, so it's a
# GoodThing(tm).
try:
n = node.strip()
if self.data["docker"] == 0:
socket.gethostbyname_ex(n)
self.Nodes.append(n)
except:
self.logger.log(node+" not found in DNS... aborting")
raise
self.filter_nodes()
else:
self.data[key] = value
def RandomNode(self):
'''Choose a random node from the cluster'''
return self.RandomGen.choice(self["nodes"])
def set_stack(self, name):
# Normalize stack names
if name == "heartbeat" or name == "lha":
self.data["Stack"] = "heartbeat"
elif name == "openais" or name == "ais" or name == "whitetank":
self.data["Stack"] = "corosync (plugin v0)"
elif name == "corosync" or name == "cs" or name == "mcp":
self.data["Stack"] = "corosync 2.x"
elif name == "cman":
self.data["Stack"] = "corosync (cman)"
elif name == "v1":
self.data["Stack"] = "corosync (plugin v1)"
elif name == "v0":
self.data["Stack"] = "corosync (plugin v0)"
else:
- print "Unknown stack: "+name
+ print("Unknown stack: "+name)
sys.exit(1)
def get_stack_short(self):
# Create the Cluster Manager object
if not self.data.has_key("Stack"):
return "unknown"
elif self.data["Stack"] == "heartbeat":
return "crm-lha"
elif self.data["Stack"] == "corosync 2.x":
if self["docker"]:
return "crm-mcp-docker"
else:
return "crm-mcp"
elif self.data["Stack"] == "corosync (cman)":
return "crm-cman"
elif self.data["Stack"] == "corosync (plugin v1)":
return "crm-plugin-v1"
elif self.data["Stack"] == "corosync (plugin v0)":
return "crm-plugin-v0"
else:
LogFactory().log("Unknown stack: "+self.data["stack"])
sys.exit(1)
def detect_syslog(self):
# Detect syslog variant
if not self.has_key("syslogd"):
if self["have_systemd"]:
# Systemd
self["syslogd"] = self.rsh(self.target, "systemctl list-units | grep syslog.*\.service.*active.*running | sed 's:.service.*::'", stdout=1).strip()
else:
# SYS-V
self["syslogd"] = self.rsh(self.target, "chkconfig --list | grep syslog.*on | awk '{print $1}' | head -n 1", stdout=1).strip()
if not self.has_key("syslogd") or not self["syslogd"]:
# default
self["syslogd"] = "rsyslog"
def detect_at_boot(self):
# Detect if the cluster starts at boot
if not self.has_key("at-boot"):
atboot = 0
if self["have_systemd"]:
# Systemd
atboot = atboot or not self.rsh(self.target, "systemctl is-enabled heartbeat.service")
atboot = atboot or not self.rsh(self.target, "systemctl is-enabled corosync.service")
atboot = atboot or not self.rsh(self.target, "systemctl is-enabled pacemaker.service")
else:
# SYS-V
atboot = atboot or not self.rsh(self.target, "chkconfig --list | grep -e corosync.*on -e heartbeat.*on -e pacemaker.*on")
self["at-boot"] = atboot
def detect_ip_offset(self):
# Try to determin an offset for IPaddr resources
if self["CIBResource"] and not self.has_key("IPBase"):
network=self.rsh(self.target, "ip addr | grep inet | grep -v -e link -e inet6 -e '/32' -e ' lo' | awk '{print $2}'", stdout=1).strip()
self["IPBase"] = self.rsh(self.target, "nmap -sn -n %s | grep 'scan report' | awk '{print $NF}' | sed 's:(::' | sed 's:)::' | sort -V | tail -n 1" % network, stdout=1).strip()
if not self["IPBase"]:
self["IPBase"] = " fe80::1234:56:7890:1000"
self.logger.log("Could not determine an offset for IPaddr resources. Perhaps nmap is not installed on the nodes.")
self.logger.log("Defaulting to '%s', use --test-ip-base to override" % self["IPBase"])
elif int(self["IPBase"].split('.')[3]) >= 240:
self.logger.log("Could not determine an offset for IPaddr resources. Upper bound is too high: %s %s"
% (self["IPBase"], self["IPBase"].split('.')[3]))
self["IPBase"] = " fe80::1234:56:7890:1000"
self.logger.log("Defaulting to '%s', use --test-ip-base to override" % self["IPBase"])
def filter_nodes(self):
if self["node-limit"] > 0:
if len(self["nodes"]) > self["node-limit"]:
self.logger.log("Limiting the number of nodes configured=%d (max=%d)"
%(len(self["nodes"]), self["node-limit"]))
while len(self["nodes"]) > self["node-limit"]:
self["nodes"].pop(len(self["nodes"])-1)
def validate(self):
if len(self["nodes"]) < 1:
- print "No nodes specified!"
+ print("No nodes specified!")
sys.exit(1)
def discover(self):
self.target = random.Random().choice(self["nodes"])
master = socket.gethostname()
# Use the IP where possible to avoid name lookup failures
for ip in socket.gethostbyname_ex(master)[2]:
if ip != "127.0.0.1":
master = ip
break;
self["cts-master"] = master
if not self.has_key("have_systemd"):
self["have_systemd"] = not self.rsh(self.target, "systemctl list-units")
self.detect_syslog()
self.detect_at_boot()
self.detect_ip_offset()
self.validate()
def parse_args(self, args):
skipthis=None
if not args:
args=sys.argv[1:]
for i in range(0, len(args)):
if skipthis:
skipthis=None
continue
elif args[i] == "-l" or args[i] == "--limit-nodes":
skipthis=1
self["node-limit"] = int(args[i+1])
elif args[i] == "-r" or args[i] == "--populate-resources":
self["CIBResource"] = 1
self["ClobberCIB"] = 1
elif args[i] == "--outputfile":
skipthis=1
self["OutputFile"] = args[i+1]
LogFactory().add_file(self["OutputFile"])
elif args[i] == "-L" or args[i] == "--logfile":
skipthis=1
self["LogWatcher"] = "remote"
self["LogAuditDisabled"] = 1
self["LogFileName"] = args[i+1]
elif args[i] == "--ip" or args[i] == "--test-ip-base":
skipthis=1
self["IPBase"] = args[i+1]
self["CIBResource"] = 1
self["ClobberCIB"] = 1
elif args[i] == "--oprofile":
skipthis=1
self["oprofile"] = args[i+1].split(' ')
elif args[i] == "--trunc":
self["TruncateLog"]=1
elif args[i] == "--list-tests" or args[i] == "--list" :
self["ListTests"]=1
elif args[i] == "--benchmark":
self["benchmark"]=1
elif args[i] == "--bsc":
self["DoBSC"] = 1
self["scenario"] = "basic-sanity"
elif args[i] == "--qarsh":
RemoteFactory().enable_qarsh()
elif args[i] == "--docker":
self["docker"] = 1
RemoteFactory().enable_docker()
elif args[i] == "--stonith" or args[i] == "--fencing":
skipthis=1
if args[i+1] == "1" or args[i+1] == "yes":
self["DoFencing"]=1
elif args[i+1] == "0" or args[i+1] == "no":
self["DoFencing"]=0
elif args[i+1] == "phd":
self["DoStonith"]=1
self["stonith-type"] = "fence_phd_kvm"
self["stonith-params"] = "pcmk_arg_map=domain:uname,delay=0"
elif args[i+1] == "rhcs" or args[i+1] == "xvm" or args[i+1] == "virt":
self["DoStonith"]=1
self["stonith-type"] = "fence_xvm"
self["stonith-params"] = "pcmk_arg_map=domain:uname,delay=0"
elif args[i+1] == "docker":
self["DoStonith"]=1
self["stonith-type"] = "fence_docker_cts"
elif args[i+1] == "scsi":
self["DoStonith"]=1
self["stonith-type"] = "fence_scsi"
self["stonith-params"] = "delay=0"
elif args[i+1] == "ssh" or args[i+1] == "lha":
self["DoStonith"]=1
self["stonith-type"] = "external/ssh"
self["stonith-params"] = "hostlist=all,livedangerously=yes"
elif args[i+1] == "north":
self["DoStonith"]=1
self["stonith-type"] = "fence_apc"
self["stonith-params"] = "ipaddr=north-apc,login=apc,passwd=apc,pcmk_host_map=north-01:2;north-02:3;north-03:4;north-04:5;north-05:6;north-06:7;north-07:9;north-08:10;north-09:11;north-10:12;north-11:13;north-12:14;north-13:15;north-14:18;north-15:17;north-16:19;"
elif args[i+1] == "south":
self["DoStonith"]=1
self["stonith-type"] = "fence_apc"
self["stonith-params"] = "ipaddr=south-apc,login=apc,passwd=apc,pcmk_host_map=south-01:2;south-02:3;south-03:4;south-04:5;south-05:6;south-06:7;south-07:9;south-08:10;south-09:11;south-10:12;south-11:13;south-12:14;south-13:15;south-14:18;south-15:17;south-16:19;"
elif args[i+1] == "east":
self["DoStonith"]=1
self["stonith-type"] = "fence_apc"
self["stonith-params"] = "ipaddr=east-apc,login=apc,passwd=apc,pcmk_host_map=east-01:2;east-02:3;east-03:4;east-04:5;east-05:6;east-06:7;east-07:9;east-08:10;east-09:11;east-10:12;east-11:13;east-12:14;east-13:15;east-14:18;east-15:17;east-16:19;"
elif args[i+1] == "west":
self["DoStonith"]=1
self["stonith-type"] = "fence_apc"
self["stonith-params"] = "ipaddr=west-apc,login=apc,passwd=apc,pcmk_host_map=west-01:2;west-02:3;west-03:4;west-04:5;west-05:6;west-06:7;west-07:9;west-08:10;west-09:11;west-10:12;west-11:13;west-12:14;west-13:15;west-14:18;west-15:17;west-16:19;"
elif args[i+1] == "openstack":
self["DoStonith"]=1
self["stonith-type"] = "fence_openstack"
- print "Obtaining OpenStack credentials from the current environment"
+ print("Obtaining OpenStack credentials from the current environment")
self["stonith-params"] = "region=%s,tenant=%s,auth=%s,user=%s,password=%s" % (
os.environ['OS_REGION_NAME'],
os.environ['OS_TENANT_NAME'],
os.environ['OS_AUTH_URL'],
os.environ['OS_USERNAME'],
os.environ['OS_PASSWORD']
)
elif args[i+1] == "rhevm":
self["DoStonith"]=1
self["stonith-type"] = "fence_rhevm"
- print "Obtaining RHEV-M credentials from the current environment"
+ print("Obtaining RHEV-M credentials from the current environment")
self["stonith-params"] = "login=%s,passwd=%s,ipaddr=%s,ipport=%s,ssl=1,shell_timeout=10" % (
os.environ['RHEVM_USERNAME'],
os.environ['RHEVM_PASSWORD'],
os.environ['RHEVM_SERVER'],
os.environ['RHEVM_PORT'],
)
else:
self.usage(args[i+1])
elif args[i] == "--stonith-type":
self["stonith-type"] = args[i+1]
skipthis=1
elif args[i] == "--stonith-args":
self["stonith-params"] = args[i+1]
skipthis=1
elif args[i] == "--standby":
skipthis=1
if args[i+1] == "1" or args[i+1] == "yes":
self["DoStandby"] = 1
elif args[i+1] == "0" or args[i+1] == "no":
self["DoStandby"] = 0
else:
self.usage(args[i+1])
elif args[i] == "--clobber-cib" or args[i] == "-c":
self["ClobberCIB"] = 1
elif args[i] == "--cib-filename":
skipthis=1
self["CIBfilename"] = args[i+1]
elif args[i] == "--xmit-loss":
try:
float(args[i+1])
except ValueError:
- print ("--xmit-loss parameter should be float")
+ print("--xmit-loss parameter should be float")
self.usage(args[i+1])
skipthis=1
self["XmitLoss"] = args[i+1]
elif args[i] == "--recv-loss":
try:
float(args[i+1])
except ValueError:
- print ("--recv-loss parameter should be float")
+ print("--recv-loss parameter should be float")
self.usage(args[i+1])
skipthis=1
self["RecvLoss"] = args[i+1]
elif args[i] == "--choose":
skipthis=1
self["tests"].append(args[i+1])
self["scenario"] = "sequence"
elif args[i] == "--nodes":
skipthis=1
self["nodes"] = args[i+1].split(' ')
elif args[i] == "-g" or args[i] == "--group" or args[i] == "--dsh-group":
skipthis=1
self["OutputFile"] = "%s/cluster-%s.log" % (os.environ['HOME'], args[i+1])
LogFactory().add_file(self["OutputFile"], "CTS")
dsh_file = "%s/.dsh/group/%s" % (os.environ['HOME'], args[i+1])
# Hacks to make my life easier
if args[i+1] == "r6":
self["Stack"] = "cman"
self["DoStonith"]=1
self["stonith-type"] = "fence_xvm"
self["stonith-params"] = "delay=0"
self["IPBase"] = " fe80::1234:56:7890:4000"
elif args[i+1] == "virt1":
self["Stack"] = "corosync"
self["DoStonith"]=1
self["stonith-type"] = "fence_xvm"
self["stonith-params"] = "delay=0"
self["IPBase"] = " fe80::1234:56:7890:1000"
elif args[i+1] == "east16" or args[i+1] == "nsew":
self["Stack"] = "corosync"
self["DoStonith"]=1
self["stonith-type"] = "fence_apc"
self["stonith-params"] = "ipaddr=east-apc,login=apc,passwd=apc,pcmk_host_map=east-01:2;east-02:3;east-03:4;east-04:5;east-05:6;east-06:7;east-07:9;east-08:10;east-09:11;east-10:12;east-11:13;east-12:14;east-13:15;east-14:18;east-15:17;east-16:19;"
self["IPBase"] = " fe80::1234:56:7890:2000"
if args[i+1] == "east16":
# Requires newer python than available via nsew
self["IPagent"] = "Dummy"
elif args[i+1] == "corosync8":
self["Stack"] = "corosync"
self["DoStonith"]=1
self["stonith-type"] = "fence_rhevm"
- print "Obtaining RHEV-M credentials from the current environment"
+ print("Obtaining RHEV-M credentials from the current environment")
self["stonith-params"] = "login=%s,passwd=%s,ipaddr=%s,ipport=%s,ssl=1,shell_timeout=10" % (
os.environ['RHEVM_USERNAME'],
os.environ['RHEVM_PASSWORD'],
os.environ['RHEVM_SERVER'],
os.environ['RHEVM_PORT'],
)
self["IPBase"] = " fe80::1234:56:7890:3000"
if os.path.isfile(dsh_file):
self["nodes"] = []
f = open(dsh_file, 'r')
for line in f:
l = line.strip().rstrip()
if not l.startswith('#'):
self["nodes"].append(l)
f.close()
else:
print("Unknown DSH group: %s" % args[i+1])
elif args[i] == "--syslog-facility" or args[i] == "--facility":
skipthis=1
self["SyslogFacility"] = args[i+1]
elif args[i] == "--seed":
skipthis=1
self.SeedRandom(args[i+1])
elif args[i] == "--warn-inactive":
self["warn-inactive"] = 1
elif args[i] == "--schema":
skipthis=1
self["Schema"] = args[i+1]
elif args[i] == "--ais":
self["Stack"] = "openais"
elif args[i] == "--at-boot" or args[i] == "--cluster-starts-at-boot":
skipthis=1
if args[i+1] == "1" or args[i+1] == "yes":
self["at-boot"] = 1
elif args[i+1] == "0" or args[i+1] == "no":
self["at-boot"] = 0
else:
self.usage(args[i+1])
elif args[i] == "--heartbeat" or args[i] == "--lha":
self["Stack"] = "heartbeat"
elif args[i] == "--hae":
self["Stack"] = "openais"
self["Schema"] = "hae"
elif args[i] == "--stack":
if args[i+1] == "fedora" or args[i+1] == "fedora-17" or args[i+1] == "fedora-18":
self["Stack"] = "corosync"
elif args[i+1] == "rhel-6":
self["Stack"] = "cman"
elif args[i+1] == "rhel-7":
self["Stack"] = "corosync"
else:
self["Stack"] = args[i+1]
skipthis=1
elif args[i] == "--once":
self["scenario"] = "all-once"
elif args[i] == "--boot":
self["scenario"] = "boot"
elif args[i] == "--valgrind-tests":
self["valgrind-tests"] = 1
elif args[i] == "--valgrind-procs":
self["valgrind-procs"] = args[i+1]
skipthis = 1
elif args[i] == "--no-loop-tests":
self["loop-tests"] = 0
elif args[i] == "--loop-minutes":
skipthis=1
try:
self["loop-minutes"]=int(args[i+1])
except ValueError:
self.usage(args[i])
elif args[i] == "--no-unsafe-tests":
self["unsafe-tests"] = 0
elif args[i] == "--experimental-tests":
self["experimental-tests"] = 1
elif args[i] == "--container-tests":
self["container-tests"] = 1
elif args[i] == "--set":
skipthis=1
(name, value) = args[i+1].split('=')
self[name] = value
- print "Setting %s = %s" % (name, value)
+ print("Setting %s = %s" % (name, value))
elif args[i] == "--help":
self.usage(args[i], 0)
elif args[i] == "--":
break
else:
try:
NumIter=int(args[i])
self["iterations"] = NumIter
except ValueError:
self.usage(args[i])
def usage(self, arg, status=1):
if status:
- print "Illegal argument %s" % arg
- print "usage: " + sys.argv[0] +" [options] number-of-iterations"
- print "\nCommon options: "
- print "\t [--nodes 'node list'] list of cluster nodes separated by whitespace"
- print "\t [--group | -g 'name'] use the nodes listed in the named DSH group (~/.dsh/groups/$name)"
- print "\t [--limit-nodes max] only use the first 'max' cluster nodes supplied with --nodes"
- print "\t [--stack (v0|v1|cman|corosync|heartbeat|openais)] which cluster stack is installed"
- print "\t [--list-tests] list the valid tests"
- print "\t [--benchmark] add the timing information"
- print "\t "
- print "Options that CTS will usually auto-detect correctly: "
- print "\t [--logfile path] where should the test software look for logs from cluster nodes"
- print "\t [--syslog-facility name] which syslog facility should the test software log to"
- print "\t [--at-boot (1|0)] does the cluster software start at boot time"
- print "\t [--test-ip-base ip] offset for generated IP address resources"
- print "\t "
- print "Options for release testing: "
- print "\t [--populate-resources | -r] generate a sample configuration"
- print "\t [--choose name] run only the named test"
- print "\t [--stonith (1 | 0 | yes | no | rhcs | ssh)]"
- print "\t [--once] run all valid tests once"
- print "\t "
- print "Additional (less common) options: "
- print "\t [--clobber-cib | -c ] erase any existing configuration"
- print "\t [--outputfile path] optional location for the test software to write logs to"
- print "\t [--trunc] truncate logfile before starting"
- print "\t [--xmit-loss lost-rate(0.0-1.0)]"
- print "\t [--recv-loss lost-rate(0.0-1.0)]"
- print "\t [--standby (1 | 0 | yes | no)]"
- print "\t [--fencing (1 | 0 | yes | no | rhcs | lha | openstack )]"
- print "\t [--stonith-type type]"
- print "\t [--stonith-args name=value]"
- print "\t [--bsc]"
- print "\t [--no-loop-tests] dont run looping/time-based tests"
- print "\t [--no-unsafe-tests] dont run tests that are unsafe for use with ocfs2/drbd"
- print "\t [--valgrind-tests] include tests using valgrind"
- print "\t [--experimental-tests] include experimental tests"
- print "\t [--container-tests] include pacemaker_remote tests that run in lxc container resources"
- print "\t [--oprofile 'node list'] list of cluster nodes to run oprofile on]"
- print "\t [--qarsh] use the QARSH backdoor to access nodes instead of SSH"
- print "\t [--docker] Indicates nodes are docker nodes."
- print "\t [--seed random_seed]"
- print "\t [--set option=value]"
- print "\t "
- print "\t Example: "
- print "\t python sys.argv[0] -g virt1 --stack cs -r --stonith ssh --schema pacemaker-1.0 500"
+ print("Illegal argument %s" % arg)
+ print("usage: " + sys.argv[0] +" [options] number-of-iterations")
+ print("\nCommon options: ")
+ print("\t [--nodes 'node list'] list of cluster nodes separated by whitespace")
+ print("\t [--group | -g 'name'] use the nodes listed in the named DSH group (~/.dsh/groups/$name)")
+ print("\t [--limit-nodes max] only use the first 'max' cluster nodes supplied with --nodes")
+ print("\t [--stack (v0|v1|cman|corosync|heartbeat|openais)] which cluster stack is installed")
+ print("\t [--list-tests] list the valid tests")
+ print("\t [--benchmark] add the timing information")
+ print("\t ")
+ print("Options that CTS will usually auto-detect correctly: ")
+ print("\t [--logfile path] where should the test software look for logs from cluster nodes")
+ print("\t [--syslog-facility name] which syslog facility should the test software log to")
+ print("\t [--at-boot (1|0)] does the cluster software start at boot time")
+ print("\t [--test-ip-base ip] offset for generated IP address resources")
+ print("\t ")
+ print("Options for release testing: ")
+ print("\t [--populate-resources | -r] generate a sample configuration")
+ print("\t [--choose name] run only the named test")
+ print("\t [--stonith (1 | 0 | yes | no | rhcs | ssh)]")
+ print("\t [--once] run all valid tests once")
+ print("\t ")
+ print("Additional (less common) options: ")
+ print("\t [--clobber-cib | -c ] erase any existing configuration")
+ print("\t [--outputfile path] optional location for the test software to write logs to")
+ print("\t [--trunc] truncate logfile before starting")
+ print("\t [--xmit-loss lost-rate(0.0-1.0)]")
+ print("\t [--recv-loss lost-rate(0.0-1.0)]")
+ print("\t [--standby (1 | 0 | yes | no)]")
+ print("\t [--fencing (1 | 0 | yes | no | rhcs | lha | openstack )]")
+ print("\t [--stonith-type type]")
+ print("\t [--stonith-args name=value]")
+ print("\t [--bsc]")
+ print("\t [--no-loop-tests] dont run looping/time-based tests")
+ print("\t [--no-unsafe-tests] dont run tests that are unsafe for use with ocfs2/drbd")
+ print("\t [--valgrind-tests] include tests using valgrind")
+ print("\t [--experimental-tests] include experimental tests")
+ print("\t [--container-tests] include pacemaker_remote tests that run in lxc container resources")
+ print("\t [--oprofile 'node list'] list of cluster nodes to run oprofile on]")
+ print("\t [--qarsh] use the QARSH backdoor to access nodes instead of SSH")
+ print("\t [--docker] Indicates nodes are docker nodes.")
+ print("\t [--seed random_seed]")
+ print("\t [--set option=value]")
+ print("\t ")
+ print("\t Example: ")
+ print("\t python sys.argv[0] -g virt1 --stack cs -r --stonith ssh --schema pacemaker-1.0 500")
sys.exit(status)
class EnvFactory:
instance = None
def __init__(self):
pass
def getInstance(self, args=None):
if not EnvFactory.instance:
EnvFactory.instance = Environment(args)
return EnvFactory.instance
diff --git a/cts/patterns.py b/cts/patterns.py
index 493b69060a..8a46d57774 100644
--- a/cts/patterns.py
+++ b/cts/patterns.py
@@ -1,538 +1,538 @@
import sys, os
from cts.CTSvars import *
patternvariants = {}
class BasePatterns:
def __init__(self, name):
self.name = name
patternvariants[name] = self
self.ignore = [
"avoid confusing Valgrind",
]
self.BadNews = []
self.components = {}
self.commands = {
"StatusCmd" : "crmadmin -t 60000 -S %s 2>/dev/null",
"CibQuery" : "cibadmin -Ql",
"CibAddXml" : "cibadmin --modify -c --xml-text %s",
"CibDelXpath" : "cibadmin --delete --xpath %s",
# 300,000 == 5 minutes
"RscRunning" : CTSvars.CRM_DAEMON_DIR + "/lrmd_test -R -r %s",
"CIBfile" : "%s:"+CTSvars.CRM_CONFIG_DIR+"/cib.xml",
"TmpDir" : "/tmp",
"BreakCommCmd" : "iptables -A INPUT -s %s -j DROP >/dev/null 2>&1",
"FixCommCmd" : "iptables -D INPUT -s %s -j DROP >/dev/null 2>&1",
# tc qdisc add dev lo root handle 1: cbq avpkt 1000 bandwidth 1000mbit
# tc class add dev lo parent 1: classid 1:1 cbq rate "$RATE"kbps allot 17000 prio 5 bounded isolated
# tc filter add dev lo parent 1: protocol ip prio 16 u32 match ip dst 127.0.0.1 match ip sport $PORT 0xFFFF flowid 1:1
# tc qdisc add dev lo parent 1: netem delay "$LATENCY"msec "$(($LATENCY/4))"msec 10% 2> /dev/null > /dev/null
"ReduceCommCmd" : "",
"RestoreCommCmd" : "tc qdisc del dev lo root",
"UUIDQueryCmd" : "crmadmin -N",
"SetCheckInterval" : "cibadmin --modify -c --xml-text '<cluster_property_set id=\"cib-bootstrap-options\"><nvpair id=\"cts-recheck-interval-setting\" name=\"cluster-recheck-interval\" value=\"%s\"/></cluster_property_set>'",
"ClearCheckInterval" : "cibadmin --delete --xpath \"//nvpair[@name='cluster-recheck-interval']\"",
"MaintenanceModeOn" : "cibadmin --modify -c --xml-text '<cluster_property_set id=\"cib-bootstrap-options\"><nvpair id=\"cts-maintenance-mode-setting\" name=\"maintenance-mode\" value=\"true\"/></cluster_property_set>'",
"MaintenanceModeOff" : "cibadmin --delete --xpath \"//nvpair[@name='maintenance-mode']\"",
"StandbyCmd" : "crm_attribute -VQ -U %s -n standby -l forever -v %s 2>/dev/null",
"StandbyQueryCmd" : "crm_attribute -QG -U %s -n standby -l forever -d off 2>/dev/null",
}
self.search = {
"Pat:DC_IDLE" : "crmd.*State transition.*-> S_IDLE",
# This wont work if we have multiple partitions
"Pat:Local_started" : "%s\W.*The local CRM is operational",
"Pat:Slave_started" : "%s\W.*State transition.*-> S_NOT_DC",
"Pat:Master_started": "%s\W.*State transition.*-> S_IDLE",
"Pat:We_stopped" : "heartbeat.*%s.*Heartbeat shutdown complete",
"Pat:Logd_stopped" : "%s\W.*logd:.*Exiting write process",
"Pat:They_stopped" : "%s\W.*LOST:.* %s ",
"Pat:They_dead" : "node %s.*: is dead",
"Pat:TransitionComplete" : "Transition status: Complete: complete",
"Pat:Fencing_start" : "Initiating remote operation .* for %s",
"Pat:Fencing_ok" : r"stonith.*:\s*Operation .* of %s by .* for .*@.*: OK",
"Pat:Fencing_recover" : r"pengine.*: Recover %s",
"Pat:RscOpOK" : r"crmd.*:\s*Operation %s_%s.*:\s*ok \(.*confirmed=\S+\)",
"Pat:RscRemoteOpOK" : r"crmd.*:\s*Operation %s_%s.*:\s*ok \(node=%s,.*,\s*confirmed=true\)",
"Pat:NodeFenced" : r"crmd.*:\s*Peer\s+%s\s+was\s+terminated\s+\(.*\)\s+by\s+.*\s+for\s+.*:\s+OK",
"Pat:FenceOpOK" : "Operation .* for host '%s' with device .* returned: 0",
}
def get_component(self, key):
if self.components.has_key(key):
return self.components[key]
- print "Unknown component '%s' for %s" % (key, self.name)
+ print("Unknown component '%s' for %s" % (key, self.name))
return []
def get_patterns(self, key):
if key == "BadNews":
return self.BadNews
elif key == "BadNewsIgnore":
return self.ignore
elif key == "Commands":
return self.commands
elif key == "Search":
return self.search
elif key == "Components":
return self.components
def __getitem__(self, key):
if key == "Name":
return self.name
elif self.commands.has_key(key):
return self.commands[key]
elif self.search.has_key(key):
return self.search[key]
else:
- print "Unknown template '%s' for %s" % (key, self.name)
+ print("Unknown template '%s' for %s" % (key, self.name))
return None
class crm_lha(BasePatterns):
def __init__(self, name):
BasePatterns.__init__(self, name)
self.commands.update({
"StartCmd" : "service heartbeat start > /dev/null 2>&1",
"StopCmd" : "service heartbeat stop > /dev/null 2>&1",
"EpochCmd" : "crm_node -H -e",
"QuorumCmd" : "crm_node -H -q",
"PartitionCmd" : "crm_node -H -p",
})
self.search.update({
# Patterns to look for in the log files for various occasions...
"Pat:ChildKilled" : "%s\W.*heartbeat.*%s.*killed by signal 9",
"Pat:ChildRespawn" : "%s\W.*heartbeat.*Respawning client.*%s",
"Pat:ChildExit" : "(ERROR|error): Client .* exited with return code",
})
self.BadNews = [
r"error:",
r"crit:",
r"ERROR:",
r"CRIT:",
r"Shutting down...NOW",
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"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",
r"Search terminated:",
r"No need to invoke the TE",
r"global_timer_callback:",
r"Faking parameter digest creation",
r"Parameters to .* action changed:",
r"Parameters to .* changed",
]
self.ignore = self.ignore + [
r"(ERROR|error):.*\s+assert\s+at\s+crm_glib_handler:"
"(ERROR|error): Message hist queue is filling up",
"stonithd.*CRIT: external_hostlist:.*'vmware gethosts' returned an empty hostlist",
"stonithd.*(ERROR|error): Could not list nodes for stonith RA external/vmware.",
"pengine.*Preventing .* from re-starting",
]
class crm_cs_v0(BasePatterns):
def __init__(self, name):
BasePatterns.__init__(self, name)
self.commands.update({
"EpochCmd" : "crm_node -e --openais",
"QuorumCmd" : "crm_node -q --openais",
"PartitionCmd" : "crm_node -p --openais",
"StartCmd" : "service corosync start",
"StopCmd" : "service corosync stop",
})
self.search.update({
# The next pattern is too early
# "Pat:We_stopped" : "%s.*Service engine unloaded: Pacemaker Cluster Manager",
# The next pattern would be preferred, but it doesn't always come out
# "Pat:We_stopped" : "%s.*Corosync Cluster Engine exiting with status",
"Pat:We_stopped" : "%s\W.*Service engine unloaded: corosync cluster quorum service",
"Pat:They_stopped" : "%s\W.*crmd.*Node %s\[.*state is now lost",
"Pat:They_dead" : "corosync:.*Node %s is now: lost",
"Pat:ChildExit" : "Child process .* exited",
"Pat:ChildKilled" : "%s\W.*corosync.*Child process %s terminated with signal 9",
"Pat:ChildRespawn" : "%s\W.*corosync.*Respawning failed child process: %s",
"Pat:InfraUp" : "%s\W.*corosync.*Initializing transport",
"Pat:PacemakerUp" : "%s\W.*pacemakerd.*Starting Pacemaker",
})
self.ignore = self.ignore + [
r"crm_mon:",
r"crmadmin:",
r"update_trace_data",
r"async_notify:.*strange, client not found",
r"Parse error: Ignoring unknown option .*nodename",
r"error.*: Operation 'reboot' .* with device 'FencingFail' returned:",
r"Child process .* terminated with signal 9",
r"getinfo response error: 1$",
"sbd.* error: inquisitor_child: DEBUG MODE IS ACTIVE",
r"sbd.* pcmk:\s*error:.*Connection to cib_ro failed",
r"sbd.* pcmk:\s*error:.*Connection to cib_ro.* closed .I/O condition=17",
]
self.BadNews = [
r"error:",
r"crit:",
r"ERROR:",
r"CRIT:",
r"Shutting down...NOW",
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|warn).*Ignoring HA message.*vote.*not in our membership list",
r"pengine.*Attempting recovery of resource",
r"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",
r"Search terminated:",
r":global_timer_callback",
r"Faking parameter digest creation",
r"Parameters to .* action changed:",
r"Parameters to .* changed",
r"The .* process .* terminated with signal",
r"Child process .* terminated with signal",
r"pengine:.*Recover .*\(.* -\> .*\)",
r"rsyslogd.* imuxsock lost .* messages from pid .* due to rate-limiting",
r"Peer is not part of our cluster",
r"We appear to be in an election loop",
r"Unknown node -> we will not deliver message",
r"(Blackbox dump requested|Problem detected)",
r"pacemakerd.*Could not connect to Cluster Configuration Database API",
r"Receiving messages from a node we think is dead",
r"share the same cluster nodeid",
r"share the same name",
#r"crm_ipc_send:.*Request .* failed",
#r"crm_ipc_send:.*Sending to .* is disabled until pending reply is received",
# Not inherently bad, but worth tracking
#r"No need to invoke the TE",
#r"ping.*: DEBUG: Updated connected = 0",
#r"Digest mis-match:",
r"crmd:.*Transition failed: terminated",
r"Local CIB .* differs from .*:",
r"warn.*:\s*Continuing but .* will NOT be used",
r"warn.*:\s*Cluster configuration file .* is corrupt",
#r"Executing .* fencing operation",
#r"fence_pcmk.* Call to fence",
#r"fence_pcmk",
r"cman killed by node",
r"Election storm",
r"stalled the FSA with pending inputs",
]
self.components["common-ignore"] = [
"Pending action:",
"error: crm_log_message_adv:",
"resources were active at shutdown",
"pending LRM operations at shutdown",
"Lost connection to the CIB service",
"Connection to the CIB terminated...",
"Sending message to CIB service FAILED",
"apply_xml_diff:.*Diff application failed!",
r"crmd.*:\s*Action A_RECOVER .* not supported",
"unconfirmed_actions:.*Waiting on .* unconfirmed actions",
"cib_native_msgready:.*Message pending on command channel",
r"crmd.*:\s*Performing A_EXIT_1 - forcefully exiting the CRMd",
"verify_stopped:.*Resource .* was active at shutdown. You may ignore this error if it is unmanaged.",
"error: attrd_connection_destroy:.*Lost connection to attrd",
r".*:\s*Executing .* fencing operation \(.*\) on ",
r"(Blackbox dump requested|Problem detected)",
# "error: native_create_actions: Resource .*stonith::.* is active on 2 nodes attempting recovery",
# "error: process_pe_message: Transition .* ERRORs found during PE processing",
]
self.components["corosync-ignore"] = [
r"error:.*Connection to the CPG API failed: Library error",
r"The .* process .* exited",
r"pacemakerd.*error:.*Child process .* exited",
r"cib.*error:.*Corosync connection lost",
r"stonith-ng.*error:.*Corosync connection terminated",
r"The cib process .* exited: Invalid argument",
r"The attrd process .* exited: Transport endpoint is not connected",
r"The crmd process .* exited: Link has been severed",
r"error:.*Child process cib .* exited: Invalid argument",
r"error:.*Child process attrd .* exited: Transport endpoint is not connected",
r"error:.*Child process crmd .* exited: Link has been severed",
r"lrmd.*error:.*Connection to stonith-ng failed",
r"lrmd.*error:.*Connection to stonith-ng.* closed",
r"lrmd.*error:.*LRMD lost STONITH connection",
r"crmd.*State transition .* S_RECOVERY",
r"crmd.*error:.*FSA: Input I_ERROR",
r"crmd.*error:.*FSA: Input I_TERMINATE",
r"crmd.*error:.*Connection to cman failed",
r"crmd.*error:.*Could not recover from internal error",
r"error:.*Connection to cib_shm failed",
r"error:.*Connection to cib_shm.* closed",
r"error:.*STONITH connection failed",
r"error: Connection to stonith-ng failed",
r"crit: Fencing daemon connection failed",
r"error: Connection to stonith-ng.* closed",
]
self.components["corosync"] = [
r"pacemakerd.*error:.*Connection destroyed",
r"attrd.*:\s*crit:.*Lost connection to Corosync service",
r"stonith.*:\s*Corosync connection terminated",
r"cib.*:\s*Corosync connection lost!\s+Exiting.",
r"crmd.*:\s*connection terminated",
r"pengine.*Scheduling Node .* for STONITH",
r"crmd.*:\s*Peer %s was terminated \(.*\) by .* for .*:\s*OK",
]
self.components["cib-ignore"] = [
"lrmd.*Connection to stonith-ng failed",
"lrmd.*Connection to stonith-ng.* closed",
"lrmd.*LRMD lost STONITH connection",
"lrmd.*STONITH connection failed, finalizing .* pending operations",
]
self.components["cib"] = [
"State transition .* S_RECOVERY",
"Respawning .* crmd",
"Respawning .* attrd",
"Connection to cib_.* failed",
"Connection to cib_.* closed",
"Connection to the CIB terminated...",
"(Child process|The) crmd .* exited: Generic Pacemaker error",
"(Child process|The) attrd .* exited: (Connection reset by peer|Transport endpoint is not connected)",
"Lost connection to CIB service",
"crmd.*Input I_TERMINATE from do_recover",
"crmd.*I_ERROR.*crmd_cib_connection_destroy",
"crmd.*Could not recover from internal error",
]
self.components["lrmd"] = [
"State transition .* S_RECOVERY",
"LRM Connection failed",
"Respawning .* crmd",
"Connection to lrmd failed",
"Connection to lrmd.* closed",
"crmd.*I_ERROR.*lrm_connection_destroy",
"(Child process|The) crmd .* exited: Generic Pacemaker error",
"crmd.*Input I_TERMINATE from do_recover",
"crmd.*Could not recover from internal error",
]
self.components["lrmd-ignore"] = []
self.components["crmd"] = [
# "WARN: determine_online_status: Node .* is unclean",
# "Scheduling Node .* for STONITH",
# "Executing .* fencing operation",
# Only if the node wasn't the DC: "State transition S_IDLE",
"State transition .* -> S_IDLE",
]
self.components["crmd-ignore"] = []
self.components["attrd"] = []
self.components["attrd-ignore"] = []
self.components["pengine"] = [
"State transition .* S_RECOVERY",
"Respawning .* crmd",
"(The|Child process) crmd .* exited: Generic Pacemaker error",
"Connection to pengine failed",
"Connection to pengine.* closed",
"Connection to the Policy Engine failed",
"crmd.*I_ERROR.*save_cib_contents",
"crmd.*Input I_TERMINATE from do_recover",
"crmd.*Could not recover from internal error",
]
self.components["pengine-ignore"] = []
self.components["stonith"] = [
"Connection to stonith-ng failed",
"LRMD lost STONITH connection",
"Connection to stonith-ng.* closed",
"Fencing daemon connection failed",
r"crmd.*:\s*warn.*:\s*Callback already present",
]
self.components["stonith-ignore"] = [
r"pengine.*: Recover Fencing",
r"Updating failcount for Fencing",
r"error:.*Connection to stonith-ng failed",
r"error:.*Connection to stonith-ng.*closed \(I/O condition=17\)",
r"crit:.*Fencing daemon connection failed",
r"error:.*Sign-in failed: triggered a retry",
"STONITH connection failed, finalizing .* pending operations.",
r"crmd.*:\s*Operation Fencing.* Error",
]
self.components["stonith-ignore"].extend(self.components["common-ignore"])
class crm_mcp(crm_cs_v0):
'''
The crm version 4 cluster manager class.
It implements the things we need to talk to and manipulate
crm clusters running on top of native corosync (no plugins)
'''
def __init__(self, name):
crm_cs_v0.__init__(self, name)
self.commands.update({
"StartCmd" : "service corosync start && service pacemaker start",
"StopCmd" : "service pacemaker stop; [ ! -e /usr/sbin/pacemaker_remoted ] || service pacemaker_remote stop; service corosync stop",
"EpochCmd" : "crm_node -e",
"QuorumCmd" : "crm_node -q",
"PartitionCmd" : "crm_node -p",
})
self.search.update({
# Close enough... "Corosync Cluster Engine exiting normally" isn't printed
# reliably and there's little interest in doing anything about it
"Pat:We_stopped" : "%s\W.*Unloading all Corosync service engines",
"Pat:They_stopped" : "%s\W.*crmd.*Node %s\[.*state is now lost",
"Pat:They_dead" : "crmd.*Node %s\[.*state is now lost",
"Pat:ChildExit" : "The .* process exited",
"Pat:ChildKilled" : "%s\W.*pacemakerd.*The %s process .* terminated with signal 9",
"Pat:ChildRespawn" : "%s\W.*pacemakerd.*Respawning failed child process: %s",
})
# if self.Env["have_systemd"]:
# self.update({
# # When systemd is in use, we can look for this instead
# "Pat:We_stopped" : "%s.*Stopped Corosync Cluster Engine",
# })
class crm_mcp_docker(crm_mcp):
'''
The crm version 4 cluster manager class.
It implements the things we need to talk to and manipulate
crm clusters running on top of native corosync (no plugins)
'''
def __init__(self, name):
crm_mcp.__init__(self, name)
self.commands.update({
"StartCmd" : "pcmk_start",
"StopCmd" : "pcmk_stop",
})
class crm_cman(crm_cs_v0):
'''
The crm version 3 cluster manager class.
It implements the things we need to talk to and manipulate
crm clusters running on top of openais
'''
def __init__(self, name):
crm_cs_v0.__init__(self, name)
self.commands.update({
"StartCmd" : "service pacemaker start",
"StopCmd" : "service pacemaker stop; [ ! -e /usr/sbin/pacemaker_remoted ] || service pacemaker_remote stop",
"EpochCmd" : "crm_node -e --cman",
"QuorumCmd" : "crm_node -q --cman",
"PartitionCmd" : "crm_node -p --cman",
"Pat:We_stopped" : "%s.*Unloading all Corosync service engines",
"Pat:They_stopped" : "%s\W.*crmd.*Node %s\[.*state is now lost",
"Pat:They_dead" : "crmd.*Node %s\[.*state is now lost",
"Pat:ChildKilled" : "%s\W.*pacemakerd.*The %s process .* terminated with signal 9",
"Pat:ChildRespawn" : "%s\W.*pacemakerd.*Respawning failed child process: %s",
})
class PatternSelector:
def __init__(self, name=None):
self.name = name
self.base = BasePatterns("crm-base")
if not name:
crm_cs_v0("crm-plugin-v0")
crm_cman("crm-cman")
crm_mcp("crm-mcp")
crm_lha("crm-lha")
elif name == "crm-lha":
crm_lha(name)
elif name == "crm-plugin-v0":
crm_cs_v0(name)
elif name == "crm-cman":
crm_cman(name)
elif name == "crm-mcp":
crm_mcp(name)
elif name == "crm-mcp-docker":
crm_mcp_docker(name)
def get_variant(self, variant):
if patternvariants.has_key(variant):
return patternvariants[variant]
- print "defaulting to crm-base for %s" % variant
+ print("defaulting to crm-base for %s" % variant)
return self.base
def get_patterns(self, variant, kind):
return self.get_variant(variant).get_patterns(kind)
def get_template(self, variant, key):
v = self.get_variant(variant)
return v[key]
def get_component(self, variant, kind):
return self.get_variant(variant).get_component(kind)
def __getitem__(self, key):
return self.get_template(self.name, key)
# python cts/CTSpatt.py -k crm-mcp -t StartCmd
if __name__ == '__main__':
pdir=os.path.dirname(sys.path[0])
sys.path.insert(0, pdir) # So that things work from the source directory
kind=None
template=None
skipthis=None
args=sys.argv[1:]
for i in range(0, len(args)):
if skipthis:
skipthis=None
continue
elif args[i] == "-k" or args[i] == "--kind":
skipthis=1
kind = args[i+1]
elif args[i] == "-t" or args[i] == "--template":
skipthis=1
template = args[i+1]
else:
- print "Illegal argument " + args[i]
+ print("Illegal argument " + args[i])
- print PatternSelector(kind)[template]
+ print(PatternSelector(kind)[template])
diff --git a/cts/remote.py b/cts/remote.py
index b32b028e3f..040b48a027 100644
--- a/cts/remote.py
+++ b/cts/remote.py
@@ -1,271 +1,271 @@
'''
Classes related to running command remotely
'''
__copyright__='''
Copyright (C) 2014 Andrew Beekhof <andrew@beekhof.net>
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., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
import string, sys, re, os
from subprocess import Popen,PIPE
from threading import Thread
pdir=os.path.dirname(sys.path[0])
sys.path.insert(0, pdir) # So that things work from the source directory
from cts.CTSvars import *
from cts.logging import *
trace_rsh=None
trace_lw=None
class AsyncWaitProc(Thread):
def __init__(self, proc, node, command, completionDelegate=None):
self.proc = proc
self.node = node
self.command = command
self.logger = LogFactory()
self.delegate = completionDelegate;
Thread.__init__(self)
def run(self):
outLines = None
errLines = None
self.logger.debug("cmd: async: target=%s, pid=%d: %s" % (self.node, self.proc.pid, self.command))
self.proc.wait()
self.logger.debug("cmd: pid %d returned %d" % (self.proc.pid, self.proc.returncode))
if self.proc.stderr:
errLines = self.proc.stderr.readlines()
self.proc.stderr.close()
for line in errLines:
self.logger.debug("cmd: stderr[%d]: %s" % (self.proc.pid, line))
if self.proc.stdout:
outLines = self.proc.stdout.readlines()
self.proc.stdout.close()
# for line in outLines:
# self.logger.debug("cmd: stdout[%d]: %s" % (self.proc.pid, line))
if self.delegate:
self.delegate.async_complete(self.proc.pid, self.proc.returncode, outLines, errLines)
class AsyncRemoteCmd(Thread):
def __init__(self, node, command, completionDelegate=None):
self.proc = None
self.node = node
self.command = command
self.logger = LogFactory()
self.delegate = completionDelegate;
Thread.__init__(self)
def run(self):
outLines = None
errLines = None
self.proc = Popen(self.command, stdout = PIPE, stderr = PIPE, close_fds = True, shell = True)
self.logger.debug("cmd: async: target=%s, pid=%d: %s" % (self.node, self.proc.pid, self.command))
self.proc.wait()
self.logger.debug("cmd: pid %d returned %d to %s" % (self.proc.pid, self.proc.returncode, repr(self.delegate)))
if self.proc.stderr:
errLines = self.proc.stderr.readlines()
self.proc.stderr.close()
for line in errLines:
self.logger.debug("cmd: stderr[%d]: %s" % (self.proc.pid, line))
if self.proc.stdout:
outLines = self.proc.stdout.readlines()
self.proc.stdout.close()
# for line in outLines:
# self.logger.log("cmd: stdout[%d]: %s" % (self.proc.pid, line))
if self.delegate:
self.delegate.async_complete(self.proc.pid, self.proc.returncode, outLines, errLines)
class RemotePrimitives:
def __init__(self, Command=None, CpCommand=None):
if CpCommand:
self.CpCommand = CpCommand
else:
# -B: batch mode, -q: no stats (quiet)
self.CpCommand = "scp -B -q"
if Command:
self.Command = Command
else:
# -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 "
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, rsh, silent=False):
self.async = []
self.rsh = rsh
self.silent = silent
self.logger = LogFactory()
if trace_rsh:
self.silent = False
self.OurNode=string.lower(os.uname()[1])
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)
+ #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.rsh.Command + " " + sysname + " '" + self._fixcmd(command) + "'"
#print ("About to run %s\n" % ret)
return ret
def log(self, args):
if not self.silent:
self.logger.log(args)
def debug(self, args):
if not self.silent:
self.logger.debug(args)
def call_async(self, node, command, completionDelegate=None):
- #if completionDelegate: print "Waiting for %d on %s: %s" % (proc.pid, node, command)
+ #if completionDelegate: print("Waiting for %d on %s: %s" % (proc.pid, node, command))
aproc = AsyncRemoteCmd(node, self._cmd([node, command]), completionDelegate=completionDelegate)
aproc.start()
return aproc
def __call__(self, node, command, stdout=0, synchronous=1, silent=False, blocking=True, completionDelegate=None):
'''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.
'''
if trace_rsh:
silent = False
rc = 0
result = None
proc = Popen(self._cmd([node, command]),
stdout = PIPE, stderr = PIPE, close_fds = True, shell = True)
- #if completionDelegate: print "Waiting for %d on %s: %s" % (proc.pid, node, command)
+ #if completionDelegate: print("Waiting for %d on %s: %s" % (proc.pid, node, command))
if not synchronous and proc.pid > 0 and not self.silent:
aproc = AsyncWaitProc(proc, node, command, completionDelegate=completionDelegate)
aproc.start()
return 0
#if not blocking:
# import fcntl
# 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 completionDelegate:
completionDelegate.async_complete(proc.pid, proc.returncode, result, errors)
if not silent:
for err in errors:
if stdout == 3:
result.append("error: "+err)
else:
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.rsh.CpCommand + " \'" + source + "\'" + " \'" + target + "\'"
rc = os.system(cpstring)
if trace_rsh:
silent = False
if not silent: self.debug("cmd: rc=%d: %s" % (rc, cpstring))
return rc
class RemoteFactory:
# Class variables
rsh = RemotePrimitives()
instance = None
def getInstance(self):
if not RemoteFactory.instance:
RemoteFactory.instance = RemoteExec(RemoteFactory.rsh, False)
return RemoteFactory.instance
def new(self, silent=False):
return RemoteExec(RemoteFactory.rsh, silent)
def enable_docker(self):
- print "Using DOCKER backend for connections to cluster nodes"
+ print("Using DOCKER backend for connections to cluster nodes")
RemoteFactory.rsh.Command = "/usr/libexec/phd/docker/phd_docker_remote_cmd "
RemoteFactory.rsh.CpCommand = "/usr/libexec/phd/docker/phd_docker_cp"
def enable_qarsh(self):
# http://nstraz.wordpress.com/2008/12/03/introducing-qarsh/
- print "Using QARSH for connections to cluster nodes"
+ print("Using QARSH for connections to cluster nodes")
RemoteFactory.rsh.Command = "qarsh -t 300 -l root"
RemoteFactory.rsh.CpCommand = "qacp -q"
diff --git a/cts/watcher.py b/cts/watcher.py
index 1182c8bb00..3d77d8e6f3 100644
--- a/cts/watcher.py
+++ b/cts/watcher.py
@@ -1,550 +1,550 @@
'''
Classes related to searching logs
'''
__copyright__='''
Copyright (C) 2014 Andrew Beekhof <andrew@beekhof.net>
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., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
import time, re, os, threading
from cts.remote import *
from cts.logging import *
has_log_watcher = {}
log_watcher_file = "cts_log_watcher.py"
log_watcher_bin = CTSvars.CRM_DAEMON_DIR + "/" + log_watcher_file
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 = 0
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]
elif args[i] == '-t' or args[i] == '--tag':
skipthis=1
if not os.access(filename, os.R_OK):
- print prefix + 'Last read: %d, limit=%d, count=%d - unreadable' % (0, limit, 0)
+ print(prefix + 'Last read: %d, limit=%d, count=%d - unreadable' % (0, limit, 0))
sys.exit(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))
+ 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 = 0
while True:
if logfile.tell() >= newsize: break
elif limit and count >= limit: break
line = logfile.readline()
if not line: break
- print line.strip()
+ print(line.strip())
count += 1
-print prefix + 'Last read: %d, limit=%d, count=%d' % (logfile.tell(), limit, count)
+print(prefix + 'Last read: %d, limit=%d, count=%d' % (logfile.tell(), limit, count))
logfile.close()
"""
class SearchObj:
def __init__(self, filename, host=None, name=None):
self.limit = None
self.cache = []
self.logger = LogFactory()
self.host = host
self.name = name
self.filename = filename
self.rsh = RemoteFactory().getInstance()
self.offset = "EOF"
if host == None:
host = "localhost"
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)
self.logger.log(message)
def debug(self, args):
message = "lw: %s: %s" % (self, args)
self.logger.debug(message)
def harvest(self, delegate=None):
async = self.harvest_async(delegate)
async.join()
def harvest_async(self, delegate=None):
self.log("Not implemented")
raise
def end(self):
self.debug("Unsetting the limit")
# Unset the limit
self.limit = None
class FileObj(SearchObj):
def __init__(self, filename, host=None, name=None):
global has_log_watcher
SearchObj.__init__(self, filename, host, name)
if host is not None:
if not has_log_watcher.has_key(host):
global log_watcher
global log_watcher_bin
self.debug("Installing %s on %s" % (log_watcher_file, host))
os.system("cat << END >> %s\n%s\nEND" %(log_watcher_file, log_watcher))
os.system("chmod 755 %s" %(log_watcher_file))
self.rsh.cp(log_watcher_file, "root@%s:%s" % (host, log_watcher_bin))
has_log_watcher[host] = 1
os.system("rm -f %s" %(log_watcher_file))
self.harvest()
def async_complete(self, pid, returncode, outLines, errLines):
for line in outLines:
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(outLines), self.offset))
self.debug("Got %d lines, new offset: %s %s" % (len(outLines), self.offset, repr(self.delegate)))
elif re.search("^CTSwatcher:.*truncated", line):
self.log(line)
elif re.search("^CTSwatcher:", line):
self.debug("Got control line: "+ line)
else:
self.cache.append(line)
if self.delegate:
self.delegate.async_complete(pid, returncode, self.cache, errLines)
def harvest_async(self, delegate=None):
self.delegate = delegate
self.cache = []
if self.limit != None and self.offset > self.limit:
if self.delegate:
self.delegate.async_complete(-1, -1, [], [])
return None
global log_watcher_bin
return self.rsh.call_async(self.host,
"python %s -t %s -p CTSwatcher: -l 200 -f %s -o %s" % (log_watcher_bin, self.name, self.filename, self.offset),
completionDelegate=self)
def setend(self):
if self.limit:
return
global log_watcher_bin
(rc, lines) = self.rsh(self.host,
"python %s -t %s -p CTSwatcher: -l 2 -f %s -o %s" % (log_watcher_bin, self.name, self.filename, "EOF"),
None, silent=True)
for line in lines:
match = re.search("^CTSwatcher:Last read: (\d+)", line)
if match:
last_offset = self.offset
self.limit = int(match.group(1))
#if last_offset == "EOF": self.debug("Got %d lines, new offset: %s" % (len(lines), self.offset))
self.debug("Set limit to: %d" % self.limit)
return
class JournalObj(SearchObj):
def __init__(self, host=None, name=None):
SearchObj.__init__(self, name, host, name)
self.harvest()
def async_complete(self, pid, returncode, outLines, errLines):
#self.log( "%d returned on %s" % (pid, self.host))
foundCursor = False
for line in outLines:
match = re.search("^-- cursor: ([^.]+)", line)
if match:
foundCursor = True
last_offset = self.offset
self.offset = match.group(1).strip()
self.debug("Got %d lines, new cursor: %s" % (len(outLines), self.offset))
else:
self.cache.append(line)
if self.limit and not foundCursor:
self.hitLimit = True
self.debug("Got %d lines but no cursor: %s" % (len(outLines), self.offset))
# Get the current cursor
(rc, outLines) = self.rsh(self.host, "journalctl -q -n 0 --show-cursor", stdout=None, silent=True, synchronous=True)
for line in outLines:
match = re.search("^-- cursor: ([^.]+)", line)
if match:
last_offset = self.offset
self.offset = match.group(1).strip()
self.debug("Got %d lines, new cursor: %s" % (len(outLines), self.offset))
else:
self.log("Not a new cursor: %s" % line)
self.cache.append(line)
if self.delegate:
self.delegate.async_complete(pid, returncode, self.cache, errLines)
def harvest_async(self, delegate=None):
self.delegate = delegate
self.cache = []
# Use --lines to prevent journalctl from overflowing the Popen input buffer
if self.limit and self.hitLimit:
return None
elif self.limit:
command = "journalctl -q --after-cursor='%s' --until '%s' --lines=200 --show-cursor" % (self.offset, self.limit)
else:
command = "journalctl -q --after-cursor='%s' --lines=200 --show-cursor" % (self.offset)
if self.offset == "EOF":
command = "journalctl -q -n 0 --show-cursor"
return self.rsh.call_async(self.host, command, completionDelegate=self)
def setend(self):
if self.limit:
return
self.hitLimit = False
(rc, lines) = self.rsh(self.host, "date +'%Y-%m-%d %H:%M:%S'", stdout=None, silent=True)
if (rc == 0) and (len(lines) == 1):
self.limit = lines[0].strip()
self.debug("Set limit to: %s" % self.limit)
else:
self.debug("Unable to set limit for %s because date returned %d lines with status %d" % (self.host,
len(lines), rc))
return
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, log, regexes, name="Anon", timeout=10, debug_level=None, silent=False, hosts=None, kind=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."
'''
self.logger = LogFactory()
self.name = name
self.regexes = regexes
self.debug_level = debug_level
self.whichmatch = -1
self.unmatched = None
self.cache_lock = threading.Lock()
self.file_list = []
self.line_cache = []
# Validate our arguments. Better sooner than later ;-)
for regex in regexes:
assert re.compile(regex)
if kind:
self.kind = kind
else:
raise
self.kind = self.Env["LogWatcher"]
if log:
self.filename = log
else:
raise
self.filename = self.Env["LogFileName"]
if hosts:
self.hosts = hosts
else:
raise
self.hosts = self.Env["nodes"]
if trace_lw:
self.debug_level = 3
silent = False
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)
self.logger.debug(message)
def setwatch(self):
'''Mark the place to start watching the log from.
'''
if self.kind == "remote":
for node in self.hosts:
self.file_list.append(FileObj(self.filename, node, self.name))
elif self.kind == "journal":
for node in self.hosts:
self.file_list.append(JournalObj(node, self.name))
else:
self.file_list.append(FileObj(self.filename))
- # print "%s now has %d files" % (self.name, len(self.file_list))
+ # print("%s now has %d files" % (self.name, len(self.file_list)))
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 async_complete(self, pid, returncode, outLines, errLines):
# TODO: Probably need a lock for updating self.line_cache
self.logger.debug("%s: Got %d lines from %d (total %d)" % (self.name, len(outLines), pid, len(self.line_cache)))
if len(outLines):
self.cache_lock.acquire()
self.line_cache.extend(outLines)
self.cache_lock.release()
def __get_lines(self, timeout):
count=0
if not len(self.file_list):
raise ValueError("No sources to read from")
pending = []
- #print "%s waiting for %d operations" % (self.name, self.pending)
+ #print("%s waiting for %d operations" % (self.name, self.pending))
for f in self.file_list:
t = f.harvest_async(self)
if t:
pending.append(t)
for t in pending:
t.join(60.0)
if t.isAlive():
self.logger.log("%s: Aborting after 20s waiting for %s logging commands" % (self.name, repr(t)))
return
- #print "Got %d lines" % len(self.line_cache)
+ #print("Got %d lines" % len(self.line_cache))
def end(self):
for f in self.file_list:
f.end()
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
if trace_lw:
silent = False
lines=0
needlines=True
begin=time.time()
end=begin+timeout+1
if self.debug_level > 2: self.debug("starting single search: timeout=%d, begin=%d, end=%d" % (timeout, begin, end))
if not self.regexes:
self.debug("Nothing to look for")
return None
if timeout == 0:
for f in self.file_list:
f.setend()
while True:
if len(self.line_cache):
lines += 1
self.cache_lock.acquire()
line = self.line_cache[0]
self.line_cache.remove(line)
self.cache_lock.release()
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 > 3: self.debug("Comparing line to: "+ regex)
#import string
#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 and end < time.time():
if self.debug_level > 1: self.debug("hit timeout: %d" % timeout)
timeout = 0
for f in self.file_list:
f.setend()
else:
self.__get_lines(timeout)
if len(self.line_cache) == 0 and end < time.time():
self.debug("Single search terminated: start=%d, end=%d, now=%d, lines=%d" % (begin, end, time.time(), lines))
return None
else:
self.debug("Waiting: start=%d, end=%d, now=%d, lines=%d" % (begin, end, time.time(), len(self.line_cache)))
time.sleep(1)
self.debug("How did we get here")
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 trace_lw:
silent = False
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
self.end()
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