use of org.apache.flink.runtime.jobmaster.JobMasterRegistrationSuccess in project flink by apache.
the class ResourceManager method registerJobManager.
// ------------------------------------------------------------------------
// RPC methods
// ------------------------------------------------------------------------
@RpcMethod
public Future<RegistrationResponse> registerJobManager(final UUID resourceManagerLeaderId, final UUID jobManagerLeaderId, final String jobManagerAddress, final JobID jobId) {
checkNotNull(resourceManagerLeaderId);
checkNotNull(jobManagerLeaderId);
checkNotNull(jobManagerAddress);
checkNotNull(jobId);
if (isValid(resourceManagerLeaderId)) {
if (!jobLeaderIdService.containsJob(jobId)) {
try {
jobLeaderIdService.addJob(jobId);
} catch (Exception e) {
ResourceManagerException exception = new ResourceManagerException("Could not add the job " + jobId + " to the job id leader service.", e);
onFatalErrorAsync(exception);
log.error("Could not add job {} to job leader id service.", jobId, e);
return FlinkCompletableFuture.completedExceptionally(exception);
}
}
log.info("Registering job manager {}@{} for job {}.", jobManagerLeaderId, jobManagerAddress, jobId);
Future<UUID> jobLeaderIdFuture;
try {
jobLeaderIdFuture = jobLeaderIdService.getLeaderId(jobId);
} catch (Exception e) {
// we cannot check the job leader id so let's fail
// TODO: Maybe it's also ok to skip this check in case that we cannot check the leader id
ResourceManagerException exception = new ResourceManagerException("Cannot obtain the " + "job leader id future to verify the correct job leader.", e);
onFatalErrorAsync(exception);
log.debug("Could not obtain the job leader id future to verify the correct job leader.");
return FlinkCompletableFuture.completedExceptionally(exception);
}
Future<JobMasterGateway> jobMasterGatewayFuture = getRpcService().connect(jobManagerAddress, JobMasterGateway.class);
Future<RegistrationResponse> registrationResponseFuture = jobMasterGatewayFuture.thenCombineAsync(jobLeaderIdFuture, new BiFunction<JobMasterGateway, UUID, RegistrationResponse>() {
@Override
public RegistrationResponse apply(JobMasterGateway jobMasterGateway, UUID jobLeaderId) {
if (isValid(resourceManagerLeaderId)) {
if (jobLeaderId.equals(jobManagerLeaderId)) {
if (jobManagerRegistrations.containsKey(jobId)) {
JobManagerRegistration oldJobManagerRegistration = jobManagerRegistrations.get(jobId);
if (oldJobManagerRegistration.getLeaderID().equals(jobLeaderId)) {
// same registration
log.debug("Job manager {}@{} was already registered.", jobManagerLeaderId, jobManagerAddress);
} else {
// tell old job manager that he is no longer the job leader
disconnectJobManager(oldJobManagerRegistration.getJobID(), new Exception("New job leader for job " + jobId + " found."));
JobManagerRegistration jobManagerRegistration = new JobManagerRegistration(jobId, jobLeaderId, jobMasterGateway);
jobManagerRegistrations.put(jobId, jobManagerRegistration);
}
} else {
// new registration for the job
JobManagerRegistration jobManagerRegistration = new JobManagerRegistration(jobId, jobLeaderId, jobMasterGateway);
jobManagerRegistrations.put(jobId, jobManagerRegistration);
}
log.info("Registered job manager {}@{} for job {}.", jobManagerLeaderId, jobManagerAddress, jobId);
return new JobMasterRegistrationSuccess(resourceManagerConfiguration.getHeartbeatInterval().toMilliseconds(), getLeaderSessionId());
} else {
log.debug("The job manager leader id {} did not match the job " + "leader id {}.", jobManagerLeaderId, jobLeaderId);
return new RegistrationResponse.Decline("Job manager leader id did not match.");
}
} else {
log.debug("The resource manager leader id changed {}. Discarding job " + "manager registration from {}.", getLeaderSessionId(), jobManagerAddress);
return new RegistrationResponse.Decline("Resource manager leader id changed.");
}
}
}, getMainThreadExecutor());
// handle exceptions which might have occurred in one of the futures inputs of combine
return registrationResponseFuture.handleAsync(new BiFunction<RegistrationResponse, Throwable, RegistrationResponse>() {
@Override
public RegistrationResponse apply(RegistrationResponse registrationResponse, Throwable throwable) {
if (throwable != null) {
if (log.isDebugEnabled()) {
log.debug("Registration of job manager {}@{} failed.", jobManagerLeaderId, jobManagerAddress, throwable);
} else {
log.info("Registration of job manager {}@{} failed.", jobManagerLeaderId, jobManagerAddress);
}
return new RegistrationResponse.Decline(throwable.getMessage());
} else {
return registrationResponse;
}
}
}, getRpcService().getExecutor());
} else {
log.debug("Discard register job manager message from {}, because the leader id " + "{} did not match the expected leader id {}.", jobManagerAddress, resourceManagerLeaderId, leaderSessionId);
return FlinkCompletableFuture.<RegistrationResponse>completed(new RegistrationResponse.Decline("Resource manager leader id did not match."));
}
}
use of org.apache.flink.runtime.jobmaster.JobMasterRegistrationSuccess in project flink by apache.
the class ResourceManagerJobMasterTest method testRegisterJobMaster.
/**
* Test receive normal registration from job master and receive duplicate registration from job master
*/
@Test
public void testRegisterJobMaster() throws Exception {
String jobMasterAddress = "/jobMasterAddress1";
JobID jobID = mockJobMaster(jobMasterAddress);
TestingLeaderElectionService resourceManagerLeaderElectionService = new TestingLeaderElectionService();
UUID jmLeaderID = UUID.randomUUID();
TestingLeaderRetrievalService jobMasterLeaderRetrievalService = new TestingLeaderRetrievalService(jobMasterAddress, jmLeaderID);
TestingFatalErrorHandler testingFatalErrorHandler = new TestingFatalErrorHandler();
final ResourceManager resourceManager = createAndStartResourceManager(resourceManagerLeaderElectionService, jobID, jobMasterLeaderRetrievalService, testingFatalErrorHandler);
final UUID rmLeaderSessionId = grantResourceManagerLeadership(resourceManagerLeaderElectionService);
// test response successful
Future<RegistrationResponse> successfulFuture = resourceManager.registerJobManager(rmLeaderSessionId, jmLeaderID, jobMasterAddress, jobID);
RegistrationResponse response = successfulFuture.get(5L, TimeUnit.SECONDS);
assertTrue(response instanceof JobMasterRegistrationSuccess);
if (testingFatalErrorHandler.hasExceptionOccurred()) {
testingFatalErrorHandler.rethrowError();
}
}
use of org.apache.flink.runtime.jobmaster.JobMasterRegistrationSuccess in project flink by apache.
the class TaskExecutorITCase method testSlotAllocation.
@Test
public void testSlotAllocation() throws Exception {
TestingFatalErrorHandler testingFatalErrorHandler = new TestingFatalErrorHandler();
TestingHighAvailabilityServices testingHAServices = new TestingHighAvailabilityServices();
final Configuration configuration = new Configuration();
final ScheduledExecutorService scheduledExecutorService = new ScheduledThreadPoolExecutor(1);
final ResourceID taskManagerResourceId = new ResourceID("foobar");
final UUID rmLeaderId = UUID.randomUUID();
final TestingLeaderElectionService rmLeaderElectionService = new TestingLeaderElectionService();
final TestingLeaderRetrievalService rmLeaderRetrievalService = new TestingLeaderRetrievalService();
final String rmAddress = "rm";
final String jmAddress = "jm";
final UUID jmLeaderId = UUID.randomUUID();
final JobID jobId = new JobID();
final ResourceProfile resourceProfile = new ResourceProfile(1.0, 1);
testingHAServices.setResourceManagerLeaderElectionService(rmLeaderElectionService);
testingHAServices.setResourceManagerLeaderRetriever(rmLeaderRetrievalService);
testingHAServices.setJobMasterLeaderRetriever(jobId, new TestingLeaderRetrievalService(jmAddress, jmLeaderId));
TestingSerialRpcService rpcService = new TestingSerialRpcService();
ResourceManagerConfiguration resourceManagerConfiguration = new ResourceManagerConfiguration(Time.milliseconds(500L), Time.milliseconds(500L), Time.minutes(5L));
SlotManagerFactory slotManagerFactory = new DefaultSlotManager.Factory();
JobLeaderIdService jobLeaderIdService = new JobLeaderIdService(testingHAServices, rpcService.getScheduledExecutor(), resourceManagerConfiguration.getJobTimeout());
MetricRegistry metricRegistry = mock(MetricRegistry.class);
HeartbeatServices heartbeatServices = mock(HeartbeatServices.class, RETURNS_MOCKS);
final TaskManagerConfiguration taskManagerConfiguration = TaskManagerConfiguration.fromConfiguration(configuration);
final TaskManagerLocation taskManagerLocation = new TaskManagerLocation(taskManagerResourceId, InetAddress.getLocalHost(), 1234);
final MemoryManager memoryManager = mock(MemoryManager.class);
final IOManager ioManager = mock(IOManager.class);
final NetworkEnvironment networkEnvironment = mock(NetworkEnvironment.class);
final TaskManagerMetricGroup taskManagerMetricGroup = mock(TaskManagerMetricGroup.class);
final BroadcastVariableManager broadcastVariableManager = mock(BroadcastVariableManager.class);
final FileCache fileCache = mock(FileCache.class);
final TaskSlotTable taskSlotTable = new TaskSlotTable(Arrays.asList(resourceProfile), new TimerService<AllocationID>(scheduledExecutorService, 100L));
final JobManagerTable jobManagerTable = new JobManagerTable();
final JobLeaderService jobLeaderService = new JobLeaderService(taskManagerLocation);
ResourceManager<ResourceID> resourceManager = new StandaloneResourceManager(rpcService, resourceManagerConfiguration, testingHAServices, slotManagerFactory, metricRegistry, jobLeaderIdService, testingFatalErrorHandler);
TaskExecutor taskExecutor = new TaskExecutor(taskManagerConfiguration, taskManagerLocation, rpcService, memoryManager, ioManager, networkEnvironment, testingHAServices, heartbeatServices, metricRegistry, taskManagerMetricGroup, broadcastVariableManager, fileCache, taskSlotTable, jobManagerTable, jobLeaderService, testingFatalErrorHandler);
JobMasterGateway jmGateway = mock(JobMasterGateway.class);
when(jmGateway.registerTaskManager(any(String.class), any(TaskManagerLocation.class), eq(jmLeaderId), any(Time.class))).thenReturn(FlinkCompletableFuture.<RegistrationResponse>completed(new JMTMRegistrationSuccess(taskManagerResourceId, 1234)));
when(jmGateway.getHostname()).thenReturn(jmAddress);
rpcService.registerGateway(rmAddress, resourceManager.getSelf());
rpcService.registerGateway(jmAddress, jmGateway);
final AllocationID allocationId = new AllocationID();
final SlotRequest slotRequest = new SlotRequest(jobId, allocationId, resourceProfile);
final SlotOffer slotOffer = new SlotOffer(allocationId, 0, resourceProfile);
try {
resourceManager.start();
taskExecutor.start();
// notify the RM that it is the leader
rmLeaderElectionService.isLeader(rmLeaderId);
// notify the TM about the new RM leader
rmLeaderRetrievalService.notifyListener(rmAddress, rmLeaderId);
Future<RegistrationResponse> registrationResponseFuture = resourceManager.registerJobManager(rmLeaderId, jmLeaderId, jmAddress, jobId);
RegistrationResponse registrationResponse = registrationResponseFuture.get();
assertTrue(registrationResponse instanceof JobMasterRegistrationSuccess);
resourceManager.requestSlot(jmLeaderId, rmLeaderId, slotRequest);
verify(jmGateway).offerSlots(eq(taskManagerResourceId), (Iterable<SlotOffer>) argThat(Matchers.contains(slotOffer)), eq(jmLeaderId), any(Time.class));
} finally {
if (testingFatalErrorHandler.hasExceptionOccurred()) {
testingFatalErrorHandler.rethrowError();
}
}
}
Aggregations