vmProfile.setMemoryOvercommitRatio(Float.parseFloat(cluster_detail_ram.getValue()));
StartAnswer startAnswer = null;
try {
if (!changeState(vm, Event.OperationRetry, destHostId, work, Step.Prepare)) {
throw new ConcurrentOperationException("Unable to update the state of the Virtual Machine");
}
} catch (NoTransitionException e1) {
throw new ConcurrentOperationException(e1.getMessage());
}
try {
if (s_logger.isDebugEnabled()) {
s_logger.debug("VM is being created in podId: " + vm.getPodIdToDeployIn());
}
_networkMgr.prepare(vmProfile, dest, ctx);
if (vm.getHypervisorType() != HypervisorType.BareMetal) {
volumeMgr.prepare(vmProfile, dest);
}
//since StorageMgr succeeded in volume creation, reuse Volume for further tries until current cluster has capacity
if (!reuseVolume) {
reuseVolume = true;
}
Commands cmds = null;
vmGuru.finalizeVirtualMachineProfile(vmProfile, dest, ctx);
VirtualMachineTO vmTO = hvGuru.implement(vmProfile);
handlePath(vmTO.getDisks(), vm.getHypervisorType());
cmds = new Commands(Command.OnError.Stop);
cmds.addCommand(new StartCommand(vmTO, dest.getHost(), getExecuteInSequence(vm.getHypervisorType())));
vmGuru.finalizeDeployment(cmds, vmProfile, dest, ctx);
work = _workDao.findById(work.getId());
if (work == null || work.getStep() != Step.Prepare) {
throw new ConcurrentOperationException("Work steps have been changed: " + work);
}
_workDao.updateStep(work, Step.Starting);
_agentMgr.send(destHostId, cmds);
_workDao.updateStep(work, Step.Started);
startAnswer = cmds.getAnswer(StartAnswer.class);
if (startAnswer != null && startAnswer.getResult()) {
handlePath(vmTO.getDisks(), startAnswer.getIqnToPath());
String host_guid = startAnswer.getHost_guid();
if (host_guid != null) {
HostVO finalHost = _resourceMgr.findHostByGuid(host_guid);
if (finalHost == null) {
throw new CloudRuntimeException("Host Guid " + host_guid + " doesn't exist in DB, something wrong here");
}
destHostId = finalHost.getId();
}
if (vmGuru.finalizeStart(vmProfile, destHostId, cmds, ctx)) {
syncDiskChainChange(startAnswer);
if (!changeState(vm, Event.OperationSucceeded, destHostId, work, Step.Done)) {
throw new ConcurrentOperationException("Unable to transition to a new state.");
}
// Update GPU device capacity
GPUDeviceTO gpuDevice = startAnswer.getVirtualMachine().getGpuDevice();
if (gpuDevice != null) {
_resourceMgr.updateGPUDetails(destHostId, gpuDevice.getGroupDetails());
}
startedVm = vm;
if (s_logger.isDebugEnabled()) {
s_logger.debug("Start completed for VM " + vm);
}
return;
} else {
if (s_logger.isDebugEnabled()) {
s_logger.info("The guru did not like the answers so stopping " + vm);
}
StopCommand cmd = new StopCommand(vm, getExecuteInSequence(vm.getHypervisorType()), false);
StopAnswer answer = (StopAnswer) _agentMgr.easySend(destHostId, cmd);
if ( answer != null ) {
if (vm.getType() == VirtualMachine.Type.User) {
String platform = answer.getPlatform();
if (platform != null) {
Map<String,String> vmmetadata = new HashMap<String,String>();
vmmetadata.put(vm.getInstanceName(), platform);
syncVMMetaData(vmmetadata);
}
}
}
if (answer == null || !answer.getResult()) {
s_logger.warn("Unable to stop " + vm + " due to " + (answer != null ? answer.getDetails() : "no answers"));
_haMgr.scheduleStop(vm, destHostId, WorkType.ForceStop);
throw new ExecutionException("Unable to stop " + vm + " so we are unable to retry the start operation");
}
throw new ExecutionException("Unable to start " + vm + " due to error in finalizeStart, not retrying");
}
}
s_logger.info("Unable to start VM on " + dest.getHost() + " due to " + (startAnswer == null ? " no start answer" : startAnswer.getDetails()));
} catch (OperationTimedoutException e) {
s_logger.debug("Unable to send the start command to host " + dest.getHost());
if (e.isActive()) {
_haMgr.scheduleStop(vm, destHostId, WorkType.CheckStop);
}
canRetry = false;
throw new AgentUnavailableException("Unable to start " + vm.getHostName(), destHostId, e);
} catch (ResourceUnavailableException e) {
s_logger.info("Unable to contact resource.", e);
if (!avoids.add(e)) {
if (e.getScope() == Volume.class || e.getScope() == Nic.class) {
throw e;
} else {
s_logger.warn("unexpected ResourceUnavailableException : " + e.getScope().getName(), e);
throw e;
}
}
} catch (InsufficientCapacityException e) {
s_logger.info("Insufficient capacity ", e);
if (!avoids.add(e)) {
if (e.getScope() == Volume.class || e.getScope() == Nic.class) {
throw e;
} else {
s_logger.warn("unexpected InsufficientCapacityException : " + e.getScope().getName(), e);
}
}
} catch (Exception e) {
s_logger.error("Failed to start instance " + vm, e);
throw new AgentUnavailableException("Unable to start instance due to " + e.getMessage(), destHostId, e);
} finally {
if (startedVm == null && canRetry) {
Step prevStep = work.getStep();
_workDao.updateStep(work, Step.Release);
// If previous step was started/ing && we got a valid answer
if ((prevStep == Step.Started || prevStep == Step.Starting) && (startAnswer != null && startAnswer.getResult())) { //TODO check the response of cleanup and record it in DB for retry
cleanup(vmGuru, vmProfile, work, Event.OperationFailed, false);
} else {
//if step is not starting/started, send cleanup command with force=true
cleanup(vmGuru, vmProfile, work, Event.OperationFailed, true);
}
}
}
}
} finally {
if (startedVm == null) {
if (canRetry) {
try {
changeState(vm, Event.OperationFailed, null, work, Step.Done);
} catch (NoTransitionException e) {
throw new ConcurrentOperationException(e.getMessage());
}
}
}
if (planToDeploy != null) {