概述
NM端启动container会经过一系列event：
- initApplication类型的ApplicationEvent
- init_application_resources类型的LocalizationEvent
ResourceLocalizationService是LocalizationEvent的事件处理器。在处理init_application_resources类型的LocalizationEvent时，记录用户与LocalResourcesTracker的map映射到privateRsrc map集合，记录appId与LocalResourcesTracker的map映射到appRsrc map集合。 - applicationInited类型的ApplicationEvent
- initContainer类型的ContainerEvent
- localize_container_resources类型的LocalizationEvent
ResourceLocalizationService是LocalizationEvent的事件处理器。在处理localize_container_resources类型的LocalizationEvent时，将LocalizationEvent显式转换为ContainerLocalizationRequestEvent，根据event中请求的resource的可见性、用户、appId，从ResourceLocalizationService的privateRsrc map集合和appRsrc map集合中获取相应的LocalResourcesTracker，由LocalResourcesTracker发起request类型的ResourceEvent。 - request类型的ResourceEvent
LocalResourceTrackerImpl是ResourceEvent的事件处理器。在处理request类型的ResourceEvent时，创建LocalizedResource，并记录该request与LocalizedResource的映射关系到map集合中。LocalizedResource同样也是ResourceEvent的事件处理器，ResourceEvent被抛给LocalizedResource处理，然后LocalizedResource发起request_resource_localization类型的LocalizerEvent。
ContainerManagerImpl
startContainers()方法
startContainers对request中的container请求进行遍历，调用startContainerInternal启动一个container，这个container是在nodemanager上准备运行task的。启动成功的放入succeededContainers列表中，失败的则放入failedContainers中，遍历结束构造一个response返回给rm。
/**
* Start a list of containers on this NodeManager.
*/
@Override
public StartContainersResponse startContainers(
StartContainersRequest requests) throws YarnException, IOException {
UserGroupInformation remoteUgi = getRemoteUgi();
NMTokenIdentifier nmTokenIdentifier = selectNMTokenIdentifier(remoteUgi);
authorizeUser(remoteUgi, nmTokenIdentifier);
List<ContainerId> succeededContainers = new ArrayList<ContainerId>();
Map<ContainerId, SerializedException> failedContainers =
new HashMap<ContainerId, SerializedException>();
// Synchronize with NodeStatusUpdaterImpl#registerWithRM
// to avoid race condition during NM-RM resync (due to RM restart) while a
// container is being started, in particular when the container has not yet
// been added to the containers map in NMContext.
synchronized (this.context) {
for (StartContainerRequest request : requests
.getStartContainerRequests()) {
ContainerId containerId = null;
try {
if (request.getContainerToken() == null
|| request.getContainerToken().getIdentifier() == null) {
throw new IOException(INVALID_CONTAINERTOKEN_MSG);
}
ContainerTokenIdentifier containerTokenIdentifier = BuilderUtils
.newContainerTokenIdentifier(request.getContainerToken());
verifyAndGetContainerTokenIdentifier(request.getContainerToken(),
containerTokenIdentifier);
containerId = containerTokenIdentifier.getContainerID();
// Initialize the AMRMProxy service instance only if the container is of
// type AM and if the AMRMProxy service is enabled
if (amrmProxyEnabled && containerTokenIdentifier.getContainerType()
.equals(ContainerType.APPLICATION_MASTER)) {
this.getAMRMProxyService().processApplicationStartRequest(request);
}
performContainerPreStartChecks(nmTokenIdentifier, request,
containerTokenIdentifier);
startContainerInternal(containerTokenIdentifier, request);
succeededContainers.add(containerId);
} catch (YarnException e) {
failedContainers.put(containerId, SerializedException.newInstance(e));
} catch (InvalidToken ie) {
failedContainers
.put(containerId, SerializedException.newInstance(ie));
throw ie;
} catch (IOException e) {
throw RPCUtil.getRemoteException(e);
}
}
return StartContainersResponse
.newInstance(getAuxServiceMetaData(), succeededContainers,
failedContainers);
}
}startContainerInternal()方法
创建container；
创建initApplication类型的ApplicationEvent；
protected void startContainerInternal(
ContainerTokenIdentifier containerTokenIdentifier,
StartContainerRequest request) throws YarnException, IOException {
ContainerId containerId = containerTokenIdentifier.getContainerID();
String containerIdStr = containerId.toString();
String user = containerTokenIdentifier.getApplicationSubmitter();
LOG.info("Start request for " + containerIdStr + " by user " + user);
ContainerLaunchContext launchContext = request.getContainerLaunchContext();
// Sanity check for local resources
for (Map.Entry<String, LocalResource> rsrc : launchContext
.getLocalResources().entrySet()) {
if (rsrc.getValue() == null || rsrc.getValue().getResource() == null) {
throw new YarnException(
"Null resource URL for local resource " + rsrc.getKey() + " : " + rsrc.getValue());
} else if (rsrc.getValue().getType() == null) {
throw new YarnException(
"Null resource type for local resource " + rsrc.getKey() + " : " + rsrc.getValue());
} else if (rsrc.getValue().getVisibility() == null) {
throw new YarnException(
"Null resource visibility for local resource " + rsrc.getKey() + " : " + rsrc.getValue());
}
}
Credentials credentials =
YarnServerSecurityUtils.parseCredentials(launchContext);
long containerStartTime = SystemClock.getInstance().getTime();
//创建container
Container container =
new ContainerImpl(getConfig(), this.dispatcher,
launchContext, credentials, metrics, containerTokenIdentifier,
context, containerStartTime);
ApplicationId applicationID =
containerId.getApplicationAttemptId().getApplicationId();
if (context.getContainers().putIfAbsent(containerId, container) != null) {
NMAuditLogger.logFailure(user, AuditConstants.START_CONTAINER,
"ContainerManagerImpl", "Container already running on this node!",
applicationID, containerId);
throw RPCUtil.getRemoteException("Container " + containerIdStr
+ " already is running on this node!!");
}
this.readLock.lock();
try {
if (!isServiceStopped()) {
if (!context.getApplications().containsKey(applicationID)) {
// Create the application
// populate the flow context from the launch context if the timeline
// service v.2 is enabled
FlowContext flowContext =
getFlowContext(launchContext, applicationID);
Application application =
new ApplicationImpl(dispatcher, user, flowContext,
applicationID, credentials, context);
if (context.getApplications().putIfAbsent(applicationID,
application) == null) {
LOG.info("Creating a new application reference for app "
+ applicationID);
LogAggregationContext logAggregationContext =
containerTokenIdentifier.getLogAggregationContext();
Map<ApplicationAccessType, String> appAcls =
container.getLaunchContext().getApplicationACLs();
context.getNMStateStore().storeApplication(applicationID,
buildAppProto(applicationID, user, credentials, appAcls,
logAggregationContext, flowContext));
//创建initApplication类型的ApplicationEvent
dispatcher.getEventHandler().handle(new ApplicationInitEvent(
applicationID, appAcls, logAggregationContext));
}
} else if (containerTokenIdentifier.getContainerType()
== ContainerType.APPLICATION_MASTER) {
FlowContext flowContext =
getFlowContext(launchContext, applicationID);
if (flowContext != null) {
ApplicationImpl application =
(ApplicationImpl) context.getApplications().get(applicationID);
// update flowContext reference in ApplicationImpl
application.setFlowContext(flowContext);
// Required to update state store for recovery.
context.getNMStateStore().storeApplication(applicationID,
buildAppProto(applicationID, user, credentials,
container.getLaunchContext().getApplicationACLs(),
containerTokenIdentifier.getLogAggregationContext(),
flowContext));
LOG.info(
"Updated application reference with flowContext " + flowContext
+ " for app " + applicationID);
} else {
LOG.info("TimelineService V2.0 is not enabled. Skipping updating "
+ "flowContext for application " + applicationID);
}
}
this.context.getNMStateStore().storeContainer(containerId,
containerTokenIdentifier.getVersion(), containerStartTime, request);
//创建initContainer类型的ApplicationEvent
dispatcher.getEventHandler().handle(
new ApplicationContainerInitEvent(container));
this.context.getContainerTokenSecretManager().startContainerSuccessful(
containerTokenIdentifier);
NMAuditLogger.logSuccess(user, AuditConstants.START_CONTAINER,
"ContainerManageImpl", applicationID, containerId);
// TODO launchedContainer misplaced -> doesn't necessarily mean a container
// launch. A finished Application will not launch containers.
metrics.launchedContainer();
metrics.allocateContainer(containerTokenIdentifier.getResource());
} else {
throw new YarnException(
"Container start failed as the NodeManager is " +
"in the process of shutting down");
}
} finally {
this.readLock.unlock();
}
}
public ApplicationInitEvent(ApplicationId appId,
Map<ApplicationAccessType, String> acls,
LogAggregationContext logAggregationContext) {
super(appId, ApplicationEventType.INIT_APPLICATION);
this.applicationACLs = acls;
this.logAggregationContext = logAggregationContext;
}
public ApplicationContainerInitEvent(Container container) {
super(container.getContainerId().getApplicationAttemptId()
.getApplicationId(), ApplicationEventType.INIT_CONTAINER);
this.container = container;
}AsyncDispatcher注册处理器
public ContainerManagerImpl(Context context, ContainerExecutor exec,
DeletionService deletionContext, NodeStatusUpdater nodeStatusUpdater,
NodeManagerMetrics metrics, LocalDirsHandlerService dirsHandler) {
..................
// ContainerManager level dispatcher.
dispatcher = new AsyncDispatcher("NM ContainerManager dispatcher");
containersLauncher = createContainersLauncher(context, exec);
addService(containersLauncher);
dispatcher.register(ContainerEventType.class,
new ContainerEventDispatcher());
dispatcher.register(ApplicationEventType.class,
createApplicationEventDispatcher());
dispatcher.register(LocalizationEventType.class,
new LocalizationEventHandlerWrapper(rsrcLocalizationSrvc,
nmMetricsPublisher));
dispatcher.register(AuxServicesEventType.class, auxiliaryServices);
dispatcher.register(ContainersMonitorEventType.class, containersMonitor);
dispatcher.register(ContainersLauncherEventType.class, containersLauncher);
dispatcher.register(ContainerSchedulerEventType.class, containerScheduler);
addService(dispatcher);
}
protected EventHandler<ApplicationEvent> createApplicationEventDispatcher() {
return new ApplicationEventDispatcher();
}
@Override
public void serviceInit(Configuration conf) throws Exception {
logHandler =
createLogHandler(conf, this.context, this.deletionService);
addIfService(logHandler);
dispatcher.register(LogHandlerEventType.class, logHandler);
// add the shared cache upload service (it will do nothing if the shared
// cache is disabled)
SharedCacheUploadService sharedCacheUploader =
createSharedCacheUploaderService();
addService(sharedCacheUploader);
dispatcher.register(SharedCacheUploadEventType.class, sharedCacheUploader);
.............
}ApplicationEvent的处理器 —— ApplicationImpl
ApplicationEventDispatcher
class ApplicationEventDispatcher implements EventHandler<ApplicationEvent> {
@Override
public void handle(ApplicationEvent event) {
Application app =
ContainerManagerImpl.this.context.getApplications().get(
event.getApplicationID());
if (app != null) {
app.handle(event);
if (nmMetricsPublisher != null) {
nmMetricsPublisher.publishApplicationEvent(event);
}
} else {
LOG.warn("Event " + event + " sent to absent application "
+ event.getApplicationID());
}
}
}ApplicationImpl使用transition处理ApplicationEvent
@Override
public void handle(ApplicationEvent event) {
this.writeLock.lock();
try {
ApplicationId applicationID = event.getApplicationID();
if (LOG.isDebugEnabled()) {
LOG.debug(
"Processing " + applicationID + " of type " + event.getType());
}
ApplicationState oldState = stateMachine.getCurrentState();
ApplicationState newState = null;
try {
// queue event requesting init of the same app
newState = stateMachine.doTransition(event.getType(), event);
} catch (InvalidStateTransitionException e) {
LOG.warn("Can't handle this event at current state", e);
}
if (newState != null && oldState != newState) {
LOG.info("Application " + applicationID + " transitioned from "
+ oldState + " to " + newState);
}
} finally {
this.writeLock.unlock();
}
}处理initApplication类型ApplicationEvent的transition
ApplicationImpl的内部stateMachineFactory注册处理initApplication类型的transition
// Transitions from NEW state
.addTransition(ApplicationState.NEW, ApplicationState.INITING,
ApplicationEventType.INIT_APPLICATION, new AppInitTransition())AppInitTransition
创建ApplicationStarted类型的LogHandlerEvent，并提交给AsyncDispatcher
/**
* Notify services of new application.
*
* In particular, this initializes the {@link LogAggregationService}
*/
@SuppressWarnings("unchecked")
static class AppInitTransition implements
SingleArcTransition<ApplicationImpl, ApplicationEvent> {
@Override
public void transition(ApplicationImpl app, ApplicationEvent event) {
ApplicationInitEvent initEvent = (ApplicationInitEvent)event;
app.applicationACLs = initEvent.getApplicationACLs();
app.aclsManager.addApplication(app.getAppId(), app.applicationACLs);
// Inform the logAggregator
app.logAggregationContext = initEvent.getLogAggregationContext();
app.dispatcher.getEventHandler().handle(
new LogHandlerAppStartedEvent(app.appId, app.user,
app.credentials, app.applicationACLs,
app.logAggregationContext, app.applicationLogInitedTimestamp));
}
}
public LogHandlerAppStartedEvent(ApplicationId appId, String user,
Credentials credentials, Map<ApplicationAccessType, String> appAcls,
LogAggregationContext logAggregationContext, long appLogInitedTime) {
//创建ApplicationStarted类型的LogHandlerEvent
super(LogHandlerEventType.APPLICATION_STARTED);
this.applicationId = appId;
this.user = user;
this.credentials = credentials;
this.appAcls = appAcls;
this.logAggregationContext = logAggregationContext;
this.recoveredAppLogInitedTime = appLogInitedTime;
}AsyncDispatcher注册处理LogHandlerEvent的处理器
dispatcher.register(LogHandlerEventType.class, logHandler);
protected LogHandler createLogHandler(Configuration conf, Context context,
DeletionService deletionService) {
if (conf.getBoolean(YarnConfiguration.LOG_AGGREGATION_ENABLED,
YarnConfiguration.DEFAULT_LOG_AGGREGATION_ENABLED)) {
return new LogAggregationService(this.dispatcher, context,
deletionService, dirsHandler);
} else {
return new NonAggregatingLogHandler(this.dispatcher, deletionService,
dirsHandler,
context.getNMStateStore());
}
}NonAggregatingLogHandler处理ApplicationStarted类型的LogHandlerEvent
@Override
public void handle(LogHandlerEvent event) {
switch (event.getType()) {
case APPLICATION_STARTED:
LogHandlerAppStartedEvent appStartedEvent =
(LogHandlerAppStartedEvent) event;
this.appOwners.put(appStartedEvent.getApplicationId(),
appStartedEvent.getUser());
this.dispatcher.getEventHandler().handle(
new ApplicationEvent(appStartedEvent.getApplicationId(),
ApplicationEventType.APPLICATION_LOG_HANDLING_INITED));
break;
..................
}
}处理application_log_handling_inited类型ApplicationEvent的transition
ApplicationImpl的内部stateMachineFactory注册处理application_log_handling_inited类型的transition
.addTransition(ApplicationState.INITING, ApplicationState.INITING,
ApplicationEventType.APPLICATION_LOG_HANDLING_INITED,
new AppLogInitDoneTransition())AppLogInitDoneTransition
创建init_application_resources类型的LocalizationEvent，提交给AsyncDispatcher。
/**
* Handles the APPLICATION_LOG_HANDLING_INITED event that occurs after
* {@link LogAggregationService} has created the directories for the app
* and started the aggregation thread for the app.
*
* In particular, this requests that the {@link ResourceLocalizationService}
* localize the application-scoped resources.
*/
@SuppressWarnings("unchecked")
static class AppLogInitDoneTransition implements
SingleArcTransition<ApplicationImpl, ApplicationEvent> {
@Override
public void transition(ApplicationImpl app, ApplicationEvent event) {
app.dispatcher.getEventHandler().handle(
new ApplicationLocalizationEvent(
LocalizationEventType.INIT_APPLICATION_RESOURCES, app));
app.setAppLogInitedTimestamp(event.getTimestamp());
try {
app.appStateStore.storeApplication(app.appId, buildAppProto(app));
} catch (Exception ex) {
LOG.warn("failed to update application state in state store", ex);
}
}
}处理LocalizationEvent的处理器 —— ResourceLocalizationService
private final ResourceLocalizationService rsrcLocalizationSrvc;
dispatcher.register(LocalizationEventType.class,
new LocalizationEventHandlerWrapper(rsrcLocalizationSrvc,
nmMetricsPublisher));
LocalizationEventHandlerWrapper只是ResourceLocalizationService的代理
@Override
public void handle(LocalizationEvent event) {
origLocalizationEventHandler.handle(event);
if (timelinePublisher != null) {
timelinePublisher.publishLocalizationEvent(event);
}
}ResourceLocalizationService
创建applicationInited类型的ApplicationEvent，并提交给AsyncDispatcher
public void handle(LocalizationEvent event) {
// TODO: create log dir as $logdir/$user/$appId
switch (event.getType()) {
case INIT_APPLICATION_RESOURCES:
handleInitApplicationResources(
((ApplicationLocalizationEvent)event).getApplication());
break;
case LOCALIZE_CONTAINER_RESOURCES:
handleInitContainerResources((ContainerLocalizationRequestEvent) event);
break;
case CONTAINER_RESOURCES_LOCALIZED:
handleContainerResourcesLocalized((ContainerLocalizationEvent) event);
break;
.................
}
}
private void handleInitApplicationResources(Application app) {
// 0) Create application tracking structs
String userName = app.getUser();
privateRsrc.putIfAbsent(userName, new LocalResourcesTrackerImpl(userName,
null, dispatcher, true, super.getConfig(), stateStore, dirsHandler));
String appIdStr = app.getAppId().toString();
appRsrc.putIfAbsent(appIdStr, new LocalResourcesTrackerImpl(app.getUser(),
app.getAppId(), dispatcher, false, super.getConfig(), stateStore,
dirsHandler));
// 1) Signal container init
//
// This is handled by the ApplicationImpl state machine and allows
// containers to proceed with launching.
dispatcher.getEventHandler().handle(new ApplicationInitedEvent(
app.getAppId()));
}
public ApplicationInitedEvent(ApplicationId appID) {
super(appID, ApplicationEventType.APPLICATION_INITED);
}处理applicationInited类型ApplicationEvent的transition
ApplicationImpl的内部stateMachineFactory注册处理applicationInited类型的transition
.addTransition(ApplicationState.INITING, ApplicationState.RUNNING,
ApplicationEventType.APPLICATION_INITED,
new AppInitDoneTransition())AppInitDoneTransition
static class AppInitDoneTransition implements
SingleArcTransition<ApplicationImpl, ApplicationEvent> {
@Override
public void transition(ApplicationImpl app, ApplicationEvent event) {
// Start all the containers waiting for ApplicationInit
for (Container container : app.containers.values()) {
app.dispatcher.getEventHandler().handle(new ContainerInitEvent(
container.getContainerId()));
}
}
}
public ContainerInitEvent(ContainerId c) {
super(c, ContainerEventType.INIT_CONTAINER);
}处理ContainerEvent的处理器和处理initContainer类型ContainerEvent的transition
ContianerManagerImpl中的AsyncDispatcher注册处理ContainerEvent类型的处理器 —— ContainerEventDispatcher
dispatcher.register(ContainerEventType.class,
new ContainerEventDispatcher());
class ContainerEventDispatcher implements EventHandler<ContainerEvent> {
@Override
public void handle(ContainerEvent event) {
Map<ContainerId,Container> containers =
ContainerManagerImpl.this.context.getContainers();
Container c = containers.get(event.getContainerID());
if (c != null) {
c.handle(event);
if (nmMetricsPublisher != null) {
nmMetricsPublisher.publishContainerEvent(event);
}
} else {
LOG.warn("Event " + event + " sent to absent container " +
event.getContainerID());
}
}
}ContainerImpl处理ContainerEvent
public void handle(ContainerEvent event) {
try {
this.writeLock.lock();
ContainerId containerID = event.getContainerID();
if (LOG.isDebugEnabled()) {
LOG.debug("Processing " + containerID + " of type " + event.getType());
}
ContainerState oldState = stateMachine.getCurrentState();
ContainerState newState = null;
try {
newState =
stateMachine.doTransition(event.getType(), event);
} catch (InvalidStateTransitionException e) {
LOG.warn("Can't handle this event at current state: Current: ["
+ oldState + "], eventType: [" + event.getType() + "]," +
" container: [" + containerID + "]", e);
}
if (newState != null && oldState != newState) {
LOG.info("Container " + containerID + " transitioned from "
+ oldState
+ " to " + newState);
}
} finally {
this.writeLock.unlock();
}
} 处理initContainer类型ContainerEvent的transition
ContainerImpl的内部stateMachineFactory注册处理initContainer类型ContainerEvent的transition
// From NEW State
.addTransition(ContainerState.NEW,
EnumSet.of(ContainerState.LOCALIZING,
ContainerState.SCHEDULED,
ContainerState.LOCALIZATION_FAILED,
ContainerState.DONE),
ContainerEventType.INIT_CONTAINER, new RequestResourcesTransition())RequestResourcesTransition 处理initContainer类型ContainerEvent
当接收到initContainer信息时，进行状态迁移。
如果有资源可以分配，发送ContainerLocalizationRequest请求给ResourceLocalizationManager，并进入Localizing状态。
如果没有资源可以分配，发送launchCoantainer event并直接进入scheduled状态。
/**
* State transition when a NEW container receives the INIT_CONTAINER
* message.
*
* If there are resources to localize, sends a
* ContainerLocalizationRequest (LOCALIZE_CONTAINER_RESOURCES)
* to the ResourceLocalizationManager and enters LOCALIZING state.
*
* If there are no resources to localize, sends LAUNCH_CONTAINER event
* and enters SCHEDULED state directly.
*
* If there are any invalid resources specified, enters LOCALIZATION_FAILED
* directly.
*/
@SuppressWarnings("unchecked") // dispatcher not typed
static class RequestResourcesTransition implements
MultipleArcTransition<ContainerImpl,ContainerEvent,ContainerState> {
@Override
public ContainerState transition(ContainerImpl container,
ContainerEvent event) {
if (container.recoveredStatus == RecoveredContainerStatus.COMPLETED) {
container.sendFinishedEvents();
return ContainerState.DONE;
} else if (container.recoveredStatus == RecoveredContainerStatus.QUEUED) {
return ContainerState.SCHEDULED;
} else if (container.recoveredAsKilled &&
container.recoveredStatus == RecoveredContainerStatus.REQUESTED) {
// container was killed but never launched
container.metrics.killedContainer();
NMAuditLogger.logSuccess(container.user,
AuditConstants.FINISH_KILLED_CONTAINER, "ContainerImpl",
container.containerId.getApplicationAttemptId().getApplicationId(),
container.containerId);
container.metrics.releaseContainer(
container.containerTokenIdentifier.getResource());
container.sendFinishedEvents();
return ContainerState.DONE;
}
final ContainerLaunchContext ctxt = container.launchContext;
container.metrics.initingContainer();
container.dispatcher.getEventHandler().handle(new AuxServicesEvent
(AuxServicesEventType.CONTAINER_INIT, container));
// Inform the AuxServices about the opaque serviceData
Map<String,ByteBuffer> csd = ctxt.getServiceData();
if (csd != null) {
// This can happen more than once per Application as each container may
// have distinct service data
for (Map.Entry<String,ByteBuffer> service : csd.entrySet()) {
container.dispatcher.getEventHandler().handle(
new AuxServicesEvent(AuxServicesEventType.APPLICATION_INIT,
container.user, container.containerId
.getApplicationAttemptId().getApplicationId(),
service.getKey().toString(), service.getValue()));
}
}
container.containerLocalizationStartTime = clock.getTime();
// Send requests for public, private resources
Map<String,LocalResource> cntrRsrc = ctxt.getLocalResources();
if (!cntrRsrc.isEmpty()) {
try {
Map<LocalResourceVisibility, Collection<LocalResourceRequest>> req =
container.resourceSet.addResources(ctxt.getLocalResources());
container.dispatcher.getEventHandler().handle(
new ContainerLocalizationRequestEvent(container, req));
} catch (URISyntaxException e) {
// malformed resource; abort container launch
LOG.warn("Failed to parse resource-request", e);
container.cleanup();
container.metrics.endInitingContainer();
return ContainerState.LOCALIZATION_FAILED;
}
return ContainerState.LOCALIZING;
} else {
//发送launchCoantainer event到AsyncDispatcher
container.sendScheduleEvent();
container.metrics.endInitingContainer();
return ContainerState.SCHEDULED;
}
}
}localizeContainerResource类型的ContainerLocalizationEvent
/**
* Event requesting the localization of the rsrc.
* @param c Container
* @param rsrc LocalResourceRequests map
*/
public ContainerLocalizationRequestEvent(Container c,
Map<LocalResourceVisibility, Collection<LocalResourceRequest>> rsrc) {
super(LocalizationEventType.LOCALIZE_CONTAINER_RESOURCES, c);
this.rsrc = rsrc;
}处理LocalizationEvent的处理器 —— ResourceLocalizationService
处理localize_container_resources类型的LocalizationEvent
public void handle(LocalizationEvent event) {
// TODO: create log dir as $logdir/$user/$appId
switch (event.getType()) {
case INIT_APPLICATION_RESOURCES:
handleInitApplicationResources(
((ApplicationLocalizationEvent)event).getApplication());
break;
case LOCALIZE_CONTAINER_RESOURCES:
handleInitContainerResources((ContainerLocalizationRequestEvent) event);
break;
case CONTAINER_RESOURCES_LOCALIZED:
handleContainerResourcesLocalized((ContainerLocalizationEvent) event);
break;
.................
}
}
/**
* For each of the requested resources for a container, determines the
* appropriate {@link LocalResourcesTracker} and forwards a
* {@link LocalResourceRequest} to that tracker.
*/
private void handleInitContainerResources(
ContainerLocalizationRequestEvent rsrcReqs) {
Container c = rsrcReqs.getContainer();
//前置校验container的当前状态，如果状态不为localizing、running、reinitializing这三者之一，则不往下处理
EnumSet<ContainerState> set =
EnumSet.of(ContainerState.LOCALIZING,
ContainerState.RUNNING, ContainerState.REINITIALIZING);
if (!set.contains(c.getContainerState())) {
LOG.warn(c.getContainerId() + " is at " + c.getContainerState()
+ " state, do not localize resources.");
return;
}
// create a loading cache for the file statuses
LoadingCache<Path,Future<FileStatus>> statCache =
CacheBuilder.newBuilder().build(FSDownload.createStatusCacheLoader(getConfig()));
LocalizerContext ctxt = new LocalizerContext(
c.getUser(), c.getContainerId(), c.getCredentials(), statCache);
Map<LocalResourceVisibility, Collection<LocalResourceRequest>> rsrcs =
rsrcReqs.getRequestedResources();
for (Map.Entry<LocalResourceVisibility, Collection<LocalResourceRequest>> e :
rsrcs.entrySet()) {
//ResourceLocalizationService记录了每个用户，或者每个应用对应的LocalResourcesTracker。
//用户映射的LocalResourcesTracker用于跟踪private resource；
//appId映射的LocalResourcesTracker用于跟踪application resource；
//LocalResourcesTracker publicRsrc成员变量用于跟踪public resource；
//根据LocalResource的可见性（public、private或application）、用户名、appId，获取相应的LocalResourcesTracker
LocalResourcesTracker tracker =
getLocalResourcesTracker(e.getKey(), c.getUser(),
c.getContainerId().getApplicationAttemptId()
.getApplicationId());
for (LocalResourceRequest req : e.getValue()) {
//由LocalResourcesTracker发起ResourceRequestEvent事件处理请求
tracker.handle(new ResourceRequestEvent(req, e.getKey(), ctxt));
if (LOG.isDebugEnabled()) {
LOG.debug("Localizing " + req.getPath() +
" for container " + c.getContainerId());
}
}
}
}
public ResourceRequestEvent(LocalResourceRequest resource,
LocalResourceVisibility vis, LocalizerContext context) {
super(resource, ResourceEventType.REQUEST);
this.vis = vis;
this.context = context;
}处理ResourceEvent的处理器 —— LocalResourcesTrackerImpl
/*
* Synchronizing this method for avoiding races due to multiple ResourceEvent's
* coming to LocalResourcesTracker from Public/Private localizer and
* Resource Localization Service.
*/
@Override
public synchronized void handle(ResourceEvent event) {
LocalResourceRequest req = event.getLocalResourceRequest();
LocalizedResource rsrc = localrsrc.get(req);
switch (event.getType()) {
case LOCALIZED:
if (useLocalCacheDirectoryManager) {
inProgressLocalResourcesMap.remove(req);
}
break;
case REQUEST:
if (rsrc != null && (!isResourcePresent(rsrc))) {
LOG.info("Resource " + rsrc.getLocalPath()
+ " is missing, localizing it again");
removeResource(req);
rsrc = null;
}
if (null == rsrc) {
rsrc = new LocalizedResource(req, dispatcher);
localrsrc.put(req, rsrc);
}
break;
}
if (rsrc == null) {
LOG.warn("Received " + event.getType() + " event for request " + req
+ " but localized resource is missing");
return;
}
rsrc.handle(event);
// Remove the resource if its downloading and its reference count has
// become 0 after RELEASE. This maybe because a container was killed while
// localizing and no other container is referring to the resource.
// NOTE: This should NOT be done for public resources since the
// download is not associated with a container-specific localizer.
if (event.getType() == ResourceEventType.RELEASE) {
if (rsrc.getState() == ResourceState.DOWNLOADING &&
rsrc.getRefCount() <= 0 &&
rsrc.getRequest().getVisibility() != LocalResourceVisibility.PUBLIC) {
removeResource(req);
}
}
if (event.getType() == ResourceEventType.LOCALIZED) {
if (rsrc.getLocalPath() != null) {
try {
stateStore.finishResourceLocalization(user, appId,
buildLocalizedResourceProto(rsrc));
} catch (IOException ioe) {
LOG.error("Error storing resource state for " + rsrc, ioe);
}
} else {
LOG.warn("Resource " + rsrc + " localized without a location");
}
}
}处理ResourceEvent的处理器 —— LocalizedResource
public void handle(ResourceEvent event) {
try {
this.writeLock.lock();
Path resourcePath = event.getLocalResourceRequest().getPath();
if (LOG.isDebugEnabled()) {
LOG.debug("Processing " + resourcePath + " of type " + event.getType());
}
ResourceState oldState = this.stateMachine.getCurrentState();
ResourceState newState = null;
try {
newState = this.stateMachine.doTransition(event.getType(), event);
} catch (InvalidStateTransitionException e) {
LOG.warn("Can't handle this event at current state", e);
}
if (newState != null && oldState != newState) {
if (LOG.isDebugEnabled()) {
LOG.debug("Resource " + resourcePath + (localPath != null ?
"(->" + localPath + ")": "") + " size : " + getSize()
+ " transitioned from " + oldState + " to " + newState);
}
}
} finally {
this.writeLock.unlock();
}
}处理Request类型ResourceEvent的transition
LocalizedResource的内部StateMachineFactory添加处理Request类型ResourceEvent的transition —— FetchResourceTransition
.addTransition(ResourceState.INIT, ResourceState.DOWNLOADING,
ResourceEventType.REQUEST, new FetchResourceTransition())FetchResourceTransition
/**
* Transition from INIT to DOWNLOADING.
* Sends a {@link LocalizerResourceRequestEvent} to the
* {@link ResourceLocalizationService}.
*/
@SuppressWarnings("unchecked") // dispatcher not typed
private static class FetchResourceTransition extends ResourceTransition {
@Override
public void transition(LocalizedResource rsrc, ResourceEvent event) {
ResourceRequestEvent req = (ResourceRequestEvent) event;
LocalizerContext ctxt = req.getContext();
ContainerId container = ctxt.getContainerId();
rsrc.ref.add(container);
rsrc.dispatcher.getEventHandler().handle(
new LocalizerResourceRequestEvent(rsrc, req.getVisibility(), ctxt,
req.getLocalResourceRequest().getPattern()));
}
}
public LocalizerResourceRequestEvent(LocalizedResource resource,
LocalResourceVisibility vis, LocalizerContext context, String pattern) {
super(LocalizerEventType.REQUEST_RESOURCE_LOCALIZATION,
context.getContainerId().toString());
this.vis = vis;
this.context = context;
this.resource = resource;
this.pattern = pattern;
}
7987
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
