svn commit: r1349038 - in /incubator/tashi/trunk/src/tashi: clustermanager/clustermanagerservice.py nodemanager/nodemanagerservice.py nodemanager/vmcontrol/qemu.py

[st...@apache.org]

Author: stroucki
Date: Mon Jun 11 21:52:10 2012
New Revision: 1349038

URL: http://svn.apache.org/viewvc?rev=1349038&view=rev
Log:
clustermanagerservice: in vmupdate, handle MigrateTrans->Exited transition explicitly
nodemanagerservice: name my threads
qemu: use self.controlledVMs in matchHostPids thread
qemu: let NM know of an exit regardless whether it was migrating out

Modified:
    incubator/tashi/trunk/src/tashi/clustermanager/clustermanagerservice.py
    incubator/tashi/trunk/src/tashi/nodemanager/nodemanagerservice.py
    incubator/tashi/trunk/src/tashi/nodemanager/vmcontrol/qemu.py

Modified: incubator/tashi/trunk/src/tashi/clustermanager/clustermanagerservice.py
URL: http://svn.apache.org/viewvc/incubator/tashi/trunk/src/tashi/clustermanager/clustermanagerservice.py?rev=1349038&r1=1349037&r2=1349038&view=diff
==============================================================================
--- incubator/tashi/trunk/src/tashi/clustermanager/clustermanagerservice.py (original)
+++ incubator/tashi/trunk/src/tashi/clustermanager/clustermanagerservice.py Mon Jun 11 21:52:10 2012
@@ -642,18 +642,24 @@ class ClusterManagerService(object):
 		if (instance.state == InstanceState.Exited):
 			# determine why a VM has exited
 			hostname = self.data.getHost(oldInstance.hostId).name
+
 			if (oldInstance.state not in [InstanceState.ShuttingDown, InstanceState.Destroying, InstanceState.Suspending]):
 				self.log.warning('Unexpected exit on %s of instance %s (vmId %d)' % (hostname, oldInstance.name, oldInstance.vmId))
+
 			if (oldInstance.state == InstanceState.Suspending):
 				self.__stateTransition(oldInstance, InstanceState.Suspending, InstanceState.Suspended)
 				oldInstance.hostId = None
 				oldInstance.vmId = None
 				self.data.releaseInstance(oldInstance)
+
+			if (oldInstance.state == InstanceState.MigrateTrans):
+				# Just await update from target host
+				self.data.releaseInstance(oldInstance)
+
 			else:
 				del self.instanceLastContactTime[oldInstance.id]
-				self.log.worning("removing %s" % oldInstance)
 				self.data.removeInstance(oldInstance)
-				self.log.warning("done remove %s" % oldInstance)
+
 		else:
 			if (instance.state):
 				# XXXstroucki does this matter?

Modified: incubator/tashi/trunk/src/tashi/nodemanager/nodemanagerservice.py
URL: http://svn.apache.org/viewvc/incubator/tashi/trunk/src/tashi/nodemanager/nodemanagerservice.py?rev=1349038&r1=1349037&r2=1349038&view=diff
==============================================================================
--- incubator/tashi/trunk/src/tashi/nodemanager/nodemanagerservice.py (original)
+++ incubator/tashi/trunk/src/tashi/nodemanager/nodemanagerservice.py Mon Jun 11 21:52:10 2012
@@ -78,8 +78,8 @@ class NodeManagerService(object):
 		self.id = self.cm.registerNodeManager(self.host, self.instances.values())
 
 		# start service threads
-		threading.Thread(target=self.__registerWithClusterManager).start()
-		threading.Thread(target=self.__statsThread).start()
+		threading.Thread(name="registerWithClusterManager", target=self.__registerWithClusterManager).start()
+		threading.Thread(name="statsThread", target=self.__statsThread).start()
 
 	def __initAccounting(self):
 		self.accountBuffer = []

Modified: incubator/tashi/trunk/src/tashi/nodemanager/vmcontrol/qemu.py
URL: http://svn.apache.org/viewvc/incubator/tashi/trunk/src/tashi/nodemanager/vmcontrol/qemu.py?rev=1349038&r1=1349037&r2=1349038&view=diff
==============================================================================
--- incubator/tashi/trunk/src/tashi/nodemanager/vmcontrol/qemu.py (original)
+++ incubator/tashi/trunk/src/tashi/nodemanager/vmcontrol/qemu.py Mon Jun 11 21:52:10 2012
@@ -162,14 +162,14 @@ class Qemu(VmControlInterface):
 		"""Will return a dict of instances by vmId to the caller"""
 		return dict((x, self.controlledVMs[x].instance) for x in self.controlledVMs.keys())
 
-	def __matchHostPids(self, controlledVMs):
+	def __matchHostPids(self):
 		"""This is run in a separate polling thread and it must do things that are thread safe"""
 
-		vmIds = controlledVMs.keys()
+		vmIds = self.controlledVMs.keys()
 		pids = self.__getHostPids()
 
 		for vmId in vmIds:
-			child = controlledVMs[vmId]
+			child = self.controlledVMs[vmId]
 			instance = child.instance
 			name = instance.name
 
@@ -180,9 +180,9 @@ class Qemu(VmControlInterface):
 				# remove info file
 				os.unlink(self.INFO_DIR + "/%d"%(vmId))
 
-				# XXXstroucki why not use self.controlledVMs
-				# argument, so why modify this fn's formal?
-				del controlledVMs[vmId]
+				# XXXstroucki python should handle
+				# locking here (?)
+				del self.controlledVMs[vmId]
 
 				# remove any stats (appropriate?)
 				try:
@@ -230,8 +230,11 @@ class Qemu(VmControlInterface):
 
 				# let the NM know
 				try:
-					if (not child.migratingOut):
-						self.nm.vmStateChange(vmId, None, InstanceState.Exited)
+					# XXXstroucki: we don't want to treat
+					# the source VM of a migration exiting
+					# as an actual
+					# exit, but the NM should probably know.
+					self.nm.vmStateChange(vmId, None, InstanceState.Exited)
 				except Exception:
 					log.exception("vmStateChange failed for VM %s" % (name))
 			else:
@@ -290,7 +293,7 @@ class Qemu(VmControlInterface):
 		while True:
 			try:
 				time.sleep(self.POLL_DELAY)
-				self.__matchHostPids(self.controlledVMs)
+				self.__matchHostPids()
 			except:
 				log.exception("Exception in poolVMsLoop")