use of org.apache.flink.api.common.JobID in project flink by apache.
the class TaskExecutorTest method testJobLeaderDetection.
/**
* Tests that a TaskManager detects a job leader for which has reserved slots. Upon detecting
* the job leader, it will offer all reserved slots to the JobManager.
*/
@Test
public void testJobLeaderDetection() throws Exception {
final JobID jobId = new JobID();
final TestingSerialRpcService rpc = new TestingSerialRpcService();
final Configuration configuration = new Configuration();
final TaskManagerConfiguration taskManagerConfiguration = TaskManagerConfiguration.fromConfiguration(configuration);
final ResourceID resourceId = new ResourceID("foobar");
final TaskManagerLocation taskManagerLocation = new TaskManagerLocation(resourceId, InetAddress.getLoopbackAddress(), 1234);
final TestingHighAvailabilityServices haServices = new TestingHighAvailabilityServices();
final TimerService<AllocationID> timerService = mock(TimerService.class);
final TaskSlotTable taskSlotTable = new TaskSlotTable(Arrays.asList(mock(ResourceProfile.class)), timerService);
final JobManagerTable jobManagerTable = new JobManagerTable();
final JobLeaderService jobLeaderService = new JobLeaderService(taskManagerLocation);
final TestingFatalErrorHandler testingFatalErrorHandler = new TestingFatalErrorHandler();
final TestingLeaderRetrievalService resourceManagerLeaderRetrievalService = new TestingLeaderRetrievalService();
final TestingLeaderRetrievalService jobManagerLeaderRetrievalService = new TestingLeaderRetrievalService();
haServices.setResourceManagerLeaderRetriever(resourceManagerLeaderRetrievalService);
haServices.setJobMasterLeaderRetriever(jobId, jobManagerLeaderRetrievalService);
final String resourceManagerAddress = "rm";
final UUID resourceManagerLeaderId = UUID.randomUUID();
final ResourceManagerGateway resourceManagerGateway = mock(ResourceManagerGateway.class);
final InstanceID registrationId = new InstanceID();
when(resourceManagerGateway.registerTaskExecutor(eq(resourceManagerLeaderId), any(String.class), eq(resourceId), any(SlotReport.class), any(Time.class))).thenReturn(FlinkCompletableFuture.<RegistrationResponse>completed(new TaskExecutorRegistrationSuccess(registrationId, 1000L)));
final String jobManagerAddress = "jm";
final UUID jobManagerLeaderId = UUID.randomUUID();
final ResourceID jmResourceId = new ResourceID(jobManagerAddress);
final int blobPort = 42;
final JobMasterGateway jobMasterGateway = mock(JobMasterGateway.class);
when(jobMasterGateway.registerTaskManager(any(String.class), eq(taskManagerLocation), eq(jobManagerLeaderId), any(Time.class))).thenReturn(FlinkCompletableFuture.<RegistrationResponse>completed(new JMTMRegistrationSuccess(jmResourceId, blobPort)));
when(jobMasterGateway.getHostname()).thenReturn(jobManagerAddress);
rpc.registerGateway(resourceManagerAddress, resourceManagerGateway);
rpc.registerGateway(jobManagerAddress, jobMasterGateway);
final AllocationID allocationId = new AllocationID();
final SlotID slotId = new SlotID(resourceId, 0);
final SlotOffer slotOffer = new SlotOffer(allocationId, 0, ResourceProfile.UNKNOWN);
try {
TaskExecutor taskManager = new TaskExecutor(taskManagerConfiguration, taskManagerLocation, rpc, mock(MemoryManager.class), mock(IOManager.class), mock(NetworkEnvironment.class), haServices, mock(HeartbeatServices.class, RETURNS_MOCKS), mock(MetricRegistry.class), mock(TaskManagerMetricGroup.class), mock(BroadcastVariableManager.class), mock(FileCache.class), taskSlotTable, jobManagerTable, jobLeaderService, testingFatalErrorHandler);
taskManager.start();
// tell the task manager about the rm leader
resourceManagerLeaderRetrievalService.notifyListener(resourceManagerAddress, resourceManagerLeaderId);
// request slots from the task manager under the given allocation id
TMSlotRequestReply reply = taskManager.requestSlot(slotId, jobId, allocationId, jobManagerAddress, resourceManagerLeaderId);
// this is hopefully successful :-)
assertTrue(reply instanceof TMSlotRequestRegistered);
// now inform the task manager about the new job leader
jobManagerLeaderRetrievalService.notifyListener(jobManagerAddress, jobManagerLeaderId);
// the job leader should get the allocation id offered
verify(jobMasterGateway).offerSlots(any(ResourceID.class), (Iterable<SlotOffer>) Matchers.argThat(contains(slotOffer)), eq(jobManagerLeaderId), any(Time.class));
// check if a concurrent error occurred
testingFatalErrorHandler.rethrowError();
} finally {
rpc.stopService();
}
}
use of org.apache.flink.api.common.JobID in project flink by apache.
the class TaskExecutorTest method testSubmitTaskBeforeAcceptSlot.
/**
* This tests task executor receive SubmitTask before OfferSlot response.
*/
@Test
public void testSubmitTaskBeforeAcceptSlot() throws Exception {
final JobID jobId = new JobID();
final TestingSerialRpcService rpc = new TestingSerialRpcService();
final Configuration configuration = new Configuration();
final TaskManagerConfiguration taskManagerConfiguration = TaskManagerConfiguration.fromConfiguration(configuration);
final ResourceID resourceId = new ResourceID("foobar");
final TaskManagerLocation taskManagerLocation = new TaskManagerLocation(resourceId, InetAddress.getLoopbackAddress(), 1234);
final TestingHighAvailabilityServices haServices = new TestingHighAvailabilityServices();
final TimerService<AllocationID> timerService = mock(TimerService.class);
final TaskSlotTable taskSlotTable = new TaskSlotTable(Arrays.asList(mock(ResourceProfile.class), mock(ResourceProfile.class)), timerService);
final JobManagerTable jobManagerTable = new JobManagerTable();
final JobLeaderService jobLeaderService = new JobLeaderService(taskManagerLocation);
final TestingFatalErrorHandler testingFatalErrorHandler = new TestingFatalErrorHandler();
final String resourceManagerAddress = "rm";
final UUID resourceManagerLeaderId = UUID.randomUUID();
final String jobManagerAddress = "jm";
final UUID jobManagerLeaderId = UUID.randomUUID();
final LeaderRetrievalService resourceManagerLeaderRetrievalService = new TestingLeaderRetrievalService(resourceManagerAddress, resourceManagerLeaderId);
final LeaderRetrievalService jobManagerLeaderRetrievalService = new TestingLeaderRetrievalService(jobManagerAddress, jobManagerLeaderId);
haServices.setResourceManagerLeaderRetriever(resourceManagerLeaderRetrievalService);
haServices.setJobMasterLeaderRetriever(jobId, jobManagerLeaderRetrievalService);
final ResourceManagerGateway resourceManagerGateway = mock(ResourceManagerGateway.class);
final InstanceID registrationId = new InstanceID();
when(resourceManagerGateway.registerTaskExecutor(eq(resourceManagerLeaderId), any(String.class), eq(resourceId), any(SlotReport.class), any(Time.class))).thenReturn(FlinkCompletableFuture.<RegistrationResponse>completed(new TaskExecutorRegistrationSuccess(registrationId, 1000L)));
final ResourceID jmResourceId = new ResourceID(jobManagerAddress);
final int blobPort = 42;
final AllocationID allocationId1 = new AllocationID();
final AllocationID allocationId2 = new AllocationID();
final SlotOffer offer1 = new SlotOffer(allocationId1, 0, ResourceProfile.UNKNOWN);
final JobMasterGateway jobMasterGateway = mock(JobMasterGateway.class);
when(jobMasterGateway.registerTaskManager(any(String.class), eq(taskManagerLocation), eq(jobManagerLeaderId), any(Time.class))).thenReturn(FlinkCompletableFuture.<RegistrationResponse>completed(new JMTMRegistrationSuccess(jmResourceId, blobPort)));
when(jobMasterGateway.getHostname()).thenReturn(jobManagerAddress);
rpc.registerGateway(resourceManagerAddress, resourceManagerGateway);
rpc.registerGateway(jobManagerAddress, jobMasterGateway);
final LibraryCacheManager libraryCacheManager = mock(LibraryCacheManager.class);
when(libraryCacheManager.getClassLoader(eq(jobId))).thenReturn(getClass().getClassLoader());
final JobManagerConnection jobManagerConnection = new JobManagerConnection(jobId, jmResourceId, jobMasterGateway, jobManagerLeaderId, mock(TaskManagerActions.class), mock(CheckpointResponder.class), libraryCacheManager, mock(ResultPartitionConsumableNotifier.class), mock(PartitionProducerStateChecker.class));
jobManagerTable.put(jobId, jobManagerConnection);
try {
final TaskExecutor taskManager = new TaskExecutor(taskManagerConfiguration, taskManagerLocation, rpc, mock(MemoryManager.class), mock(IOManager.class), mock(NetworkEnvironment.class), haServices, mock(HeartbeatServices.class, RETURNS_MOCKS), mock(MetricRegistry.class), mock(TaskManagerMetricGroup.class), mock(BroadcastVariableManager.class), mock(FileCache.class), taskSlotTable, jobManagerTable, jobLeaderService, testingFatalErrorHandler);
taskManager.start();
taskSlotTable.allocateSlot(0, jobId, allocationId1, Time.milliseconds(10000L));
taskSlotTable.allocateSlot(1, jobId, allocationId2, Time.milliseconds(10000L));
final JobVertexID jobVertexId = new JobVertexID();
JobInformation jobInformation = new JobInformation(jobId, name.getMethodName(), new SerializedValue<>(new ExecutionConfig()), new Configuration(), Collections.<BlobKey>emptyList(), Collections.<URL>emptyList());
TaskInformation taskInformation = new TaskInformation(jobVertexId, "test task", 1, 1, TestInvokable.class.getName(), new Configuration());
SerializedValue<JobInformation> serializedJobInformation = new SerializedValue<>(jobInformation);
SerializedValue<TaskInformation> serializedJobVertexInformation = new SerializedValue<>(taskInformation);
final TaskDeploymentDescriptor tdd = new TaskDeploymentDescriptor(serializedJobInformation, serializedJobVertexInformation, new ExecutionAttemptID(), allocationId1, 0, 0, 0, null, Collections.<ResultPartitionDeploymentDescriptor>emptyList(), Collections.<InputGateDeploymentDescriptor>emptyList());
CompletableFuture<Iterable<SlotOffer>> offerResultFuture = new FlinkCompletableFuture<>();
// submit task first and then return acceptance response
when(jobMasterGateway.offerSlots(any(ResourceID.class), any(Iterable.class), eq(jobManagerLeaderId), any(Time.class))).thenReturn(offerResultFuture);
// we have to add the job after the TaskExecutor, because otherwise the service has not
// been properly started. This will also offer the slots to the job master
jobLeaderService.addJob(jobId, jobManagerAddress);
verify(jobMasterGateway).offerSlots(any(ResourceID.class), any(Iterable.class), eq(jobManagerLeaderId), any(Time.class));
// submit the task without having acknowledge the offered slots
taskManager.submitTask(tdd, jobManagerLeaderId);
// acknowledge the offered slots
offerResultFuture.complete(Collections.singleton(offer1));
verify(resourceManagerGateway).notifySlotAvailable(eq(resourceManagerLeaderId), eq(registrationId), eq(new SlotID(resourceId, 1)));
assertTrue(taskSlotTable.existsActiveSlot(jobId, allocationId1));
assertFalse(taskSlotTable.existsActiveSlot(jobId, allocationId2));
assertTrue(taskSlotTable.isSlotFree(1));
// check if a concurrent error occurred
testingFatalErrorHandler.rethrowError();
} finally {
rpc.stopService();
}
}
use of org.apache.flink.api.common.JobID in project flink by apache.
the class TaskAsyncCallTest method createTask.
private static Task createTask() throws Exception {
LibraryCacheManager libCache = mock(LibraryCacheManager.class);
when(libCache.getClassLoader(any(JobID.class))).thenReturn(ClassLoader.getSystemClassLoader());
ResultPartitionManager partitionManager = mock(ResultPartitionManager.class);
ResultPartitionConsumableNotifier consumableNotifier = mock(ResultPartitionConsumableNotifier.class);
PartitionProducerStateChecker partitionProducerStateChecker = mock(PartitionProducerStateChecker.class);
Executor executor = mock(Executor.class);
NetworkEnvironment networkEnvironment = mock(NetworkEnvironment.class);
when(networkEnvironment.getResultPartitionManager()).thenReturn(partitionManager);
when(networkEnvironment.getDefaultIOMode()).thenReturn(IOManager.IOMode.SYNC);
when(networkEnvironment.createKvStateTaskRegistry(any(JobID.class), any(JobVertexID.class))).thenReturn(mock(TaskKvStateRegistry.class));
JobInformation jobInformation = new JobInformation(new JobID(), "Job Name", new SerializedValue<>(new ExecutionConfig()), new Configuration(), Collections.<BlobKey>emptyList(), Collections.<URL>emptyList());
TaskInformation taskInformation = new TaskInformation(new JobVertexID(), "Test Task", 1, 1, CheckpointsInOrderInvokable.class.getName(), new Configuration());
return new Task(jobInformation, taskInformation, new ExecutionAttemptID(), new AllocationID(), 0, 0, Collections.<ResultPartitionDeploymentDescriptor>emptyList(), Collections.<InputGateDeploymentDescriptor>emptyList(), 0, new TaskStateHandles(), mock(MemoryManager.class), mock(IOManager.class), networkEnvironment, mock(BroadcastVariableManager.class), mock(TaskManagerActions.class), mock(InputSplitProvider.class), mock(CheckpointResponder.class), libCache, mock(FileCache.class), new TestingTaskManagerRuntimeInfo(), mock(TaskMetricGroup.class), consumableNotifier, partitionProducerStateChecker, executor);
}
use of org.apache.flink.api.common.JobID in project flink by apache.
the class SlotProtocolTest method testSlotAvailableRequest.
/**
* Tests whether
* 1) a SlotRequest is routed to the SlotManager
* 2) a SlotRequest is confirmed
* 3) a SlotRequest leads to an allocation of a registered slot
* 4) a SlotRequest is routed to the TaskExecutor
*/
@Test
public void testSlotAvailableRequest() throws Exception {
final String rmAddress = "/rm1";
final String jmAddress = "/jm1";
final String tmAddress = "/tm1";
final JobID jobID = new JobID();
testRpcService.registerGateway(jmAddress, mock(JobMasterGateway.class));
final TestingHighAvailabilityServices testingHaServices = new TestingHighAvailabilityServices();
final UUID rmLeaderID = UUID.randomUUID();
final UUID jmLeaderID = UUID.randomUUID();
TestingLeaderElectionService rmLeaderElectionService = configureHA(testingHaServices, jobID, rmAddress, rmLeaderID, jmAddress, jmLeaderID);
TaskExecutorGateway taskExecutorGateway = mock(TaskExecutorGateway.class);
Mockito.when(taskExecutorGateway.requestSlot(any(SlotID.class), any(JobID.class), any(AllocationID.class), any(String.class), any(UUID.class), any(Time.class))).thenReturn(new FlinkCompletableFuture<TMSlotRequestReply>());
testRpcService.registerGateway(tmAddress, taskExecutorGateway);
ResourceManagerConfiguration resourceManagerConfiguration = new ResourceManagerConfiguration(Time.seconds(5L), Time.seconds(5L), Time.minutes(5L));
JobLeaderIdService jobLeaderIdService = new JobLeaderIdService(testingHaServices, testRpcService.getScheduledExecutor(), resourceManagerConfiguration.getJobTimeout());
TestingSlotManagerFactory slotManagerFactory = new TestingSlotManagerFactory();
ResourceManager<ResourceID> resourceManager = Mockito.spy(new StandaloneResourceManager(testRpcService, resourceManagerConfiguration, testingHaServices, slotManagerFactory, mock(MetricRegistry.class), jobLeaderIdService, mock(FatalErrorHandler.class)));
resourceManager.start();
rmLeaderElectionService.isLeader(rmLeaderID);
Thread.sleep(1000);
Future<RegistrationResponse> registrationFuture = resourceManager.registerJobManager(rmLeaderID, jmLeaderID, jmAddress, jobID);
try {
registrationFuture.get(5L, TimeUnit.SECONDS);
} catch (Exception e) {
Assert.fail("JobManager registration Future didn't become ready.");
}
final SlotManager slotManager = slotManagerFactory.slotManager;
final ResourceID resourceID = ResourceID.generate();
final AllocationID allocationID = new AllocationID();
final ResourceProfile resourceProfile = new ResourceProfile(1.0, 100);
final SlotID slotID = new SlotID(resourceID, 0);
final SlotStatus slotStatus = new SlotStatus(slotID, resourceProfile);
final SlotReport slotReport = new SlotReport(Collections.singletonList(slotStatus));
// register slot at SlotManager
slotManager.registerTaskExecutor(resourceID, new TaskExecutorRegistration(taskExecutorGateway), slotReport);
SlotRequest slotRequest = new SlotRequest(jobID, allocationID, resourceProfile);
RMSlotRequestReply slotRequestReply = resourceManager.requestSlot(jmLeaderID, rmLeaderID, slotRequest);
// 1) a SlotRequest is routed to the SlotManager
verify(slotManager).requestSlot(slotRequest);
// 2) a SlotRequest is confirmed
Assert.assertEquals(slotRequestReply.getAllocationID(), allocationID);
// 3) a SlotRequest leads to an allocation of a registered slot
Assert.assertTrue(slotManager.isAllocated(slotID));
Assert.assertTrue(slotManager.isAllocated(allocationID));
// 4) a SlotRequest is routed to the TaskExecutor
verify(taskExecutorGateway, timeout(5000)).requestSlot(eq(slotID), eq(jobID), eq(allocationID), any(String.class), any(UUID.class), any(Time.class));
}
use of org.apache.flink.api.common.JobID in project flink by apache.
the class FsSavepointStreamFactoryTest method testSavepointStreamDirectoryLayout.
/**
* Tests that the factory creates all files in the given directory without
* creating any sub directories.
*/
@Test
public void testSavepointStreamDirectoryLayout() throws Exception {
File testRoot = folder.newFolder();
JobID jobId = new JobID();
FsSavepointStreamFactory savepointStreamFactory = new FsSavepointStreamFactory(new Path(testRoot.getAbsolutePath()), jobId, 0);
File[] listed = testRoot.listFiles();
assertNotNull(listed);
assertEquals(0, listed.length);
FsCheckpointStateOutputStream stream = savepointStreamFactory.createCheckpointStateOutputStream(1273, 19231);
stream.write(1);
FileStateHandle handle = (FileStateHandle) stream.closeAndGetHandle();
listed = testRoot.listFiles();
assertNotNull(listed);
assertEquals(1, listed.length);
assertEquals(handle.getFilePath().getPath(), listed[0].getPath());
}
Aggregations