Examples with LogAggregationContext org.apache.hadoop.yarn.api.records.LogAggregationContext used on opensource projects

Example 11 with LogAggregationContext

use of org.apache.hadoop.yarn.api.records.LogAggregationContext in project hadoop by apache.

the class TestLogAggregationService method testNoLogsUploadedOnAppFinish.

/* Test to verify fix for YARN-3793 */
@Test
public void testNoLogsUploadedOnAppFinish() throws Exception {
    this.delSrvc = new DeletionService(createContainerExecutor());
    delSrvc = spy(delSrvc);
    this.delSrvc.init(conf);
    this.conf.set(YarnConfiguration.NM_LOG_DIRS, localLogDir.getAbsolutePath());
    this.conf.set(YarnConfiguration.NM_REMOTE_APP_LOG_DIR, this.remoteRootLogDir.getAbsolutePath());
    LogAggregationService logAggregationService = new LogAggregationService(dispatcher, this.context, this.delSrvc, super.dirsHandler);
    logAggregationService.init(this.conf);
    logAggregationService.start();
    ApplicationId app = BuilderUtils.newApplicationId(1234, 1);
    File appLogDir = new File(localLogDir, app.toString());
    appLogDir.mkdir();
    LogAggregationContext context = LogAggregationContext.newInstance("HOST*", "sys*");
    logAggregationService.handle(new LogHandlerAppStartedEvent(app, this.user, null, this.acls, context));
    ApplicationAttemptId appAttemptId = BuilderUtils.newApplicationAttemptId(app, 1);
    ContainerId cont = createContainer(appAttemptId, 1, ContainerType.APPLICATION_MASTER);
    writeContainerLogs(appLogDir, cont, new String[] { "stdout", "stderr", "syslog" });
    logAggregationService.handle(new LogHandlerContainerFinishedEvent(cont, 0));
    logAggregationService.handle(new LogHandlerAppFinishedEvent(app));
    logAggregationService.stop();
    delSrvc.stop();
    // Aggregated logs should not be deleted if not uploaded.
    verify(delSrvc, times(0)).delete(user, null);
}

Example 12 with LogAggregationContext

use of org.apache.hadoop.yarn.api.records.LogAggregationContext in project hadoop by apache.

the class TestContainerManagerRecovery method testContainerCleanupOnShutdown.

@Test
public void testContainerCleanupOnShutdown() throws Exception {
    ApplicationId appId = ApplicationId.newInstance(0, 1);
    ApplicationAttemptId attemptId = ApplicationAttemptId.newInstance(appId, 1);
    ContainerId cid = ContainerId.newContainerId(attemptId, 1);
    Map<String, LocalResource> localResources = Collections.emptyMap();
    Map<String, String> containerEnv = Collections.emptyMap();
    List<String> containerCmds = Collections.emptyList();
    Map<String, ByteBuffer> serviceData = Collections.emptyMap();
    Credentials containerCreds = new Credentials();
    DataOutputBuffer dob = new DataOutputBuffer();
    containerCreds.writeTokenStorageToStream(dob);
    ByteBuffer containerTokens = ByteBuffer.wrap(dob.getData(), 0, dob.getLength());
    Map<ApplicationAccessType, String> acls = Collections.emptyMap();
    ContainerLaunchContext clc = ContainerLaunchContext.newInstance(localResources, containerEnv, containerCmds, serviceData, containerTokens, acls);
    // create the logAggregationContext
    LogAggregationContext logAggregationContext = LogAggregationContext.newInstance("includePattern", "excludePattern");
    // verify containers are stopped on shutdown without recovery
    conf.setBoolean(YarnConfiguration.NM_RECOVERY_ENABLED, false);
    conf.setBoolean(YarnConfiguration.NM_RECOVERY_SUPERVISED, false);
    Context context = createContext(conf, new NMNullStateStoreService());
    ContainerManagerImpl cm = spy(createContainerManager(context));
    cm.init(conf);
    cm.start();
    StartContainersResponse startResponse = startContainer(context, cm, cid, clc, logAggregationContext);
    assertEquals(1, startResponse.getSuccessfullyStartedContainers().size());
    cm.stop();
    verify(cm).handle(isA(CMgrCompletedAppsEvent.class));
    // verify containers are stopped on shutdown with unsupervised recovery
    conf.setBoolean(YarnConfiguration.NM_RECOVERY_ENABLED, true);
    conf.setBoolean(YarnConfiguration.NM_RECOVERY_SUPERVISED, false);
    NMMemoryStateStoreService memStore = new NMMemoryStateStoreService();
    memStore.init(conf);
    memStore.start();
    context = createContext(conf, memStore);
    cm = spy(createContainerManager(context));
    cm.init(conf);
    cm.start();
    startResponse = startContainer(context, cm, cid, clc, logAggregationContext);
    assertEquals(1, startResponse.getSuccessfullyStartedContainers().size());
    cm.stop();
    memStore.close();
    verify(cm).handle(isA(CMgrCompletedAppsEvent.class));
    // verify containers are not stopped on shutdown with supervised recovery
    conf.setBoolean(YarnConfiguration.NM_RECOVERY_ENABLED, true);
    conf.setBoolean(YarnConfiguration.NM_RECOVERY_SUPERVISED, true);
    memStore = new NMMemoryStateStoreService();
    memStore.init(conf);
    memStore.start();
    context = createContext(conf, memStore);
    cm = spy(createContainerManager(context));
    cm.init(conf);
    cm.start();
    startResponse = startContainer(context, cm, cid, clc, logAggregationContext);
    assertEquals(1, startResponse.getSuccessfullyStartedContainers().size());
    cm.stop();
    memStore.close();
    verify(cm, never()).handle(isA(CMgrCompletedAppsEvent.class));
}

Example 13 with LogAggregationContext

use of org.apache.hadoop.yarn.api.records.LogAggregationContext in project hadoop by apache.

the class TestContainerManagerRecovery method testApplicationRecovery.

@Test
public void testApplicationRecovery() throws Exception {
    conf.setBoolean(YarnConfiguration.NM_RECOVERY_ENABLED, true);
    conf.setBoolean(YarnConfiguration.NM_RECOVERY_SUPERVISED, true);
    conf.setBoolean(YarnConfiguration.YARN_ACL_ENABLE, true);
    conf.set(YarnConfiguration.YARN_ADMIN_ACL, "yarn_admin_user");
    NMStateStoreService stateStore = new NMMemoryStateStoreService();
    stateStore.init(conf);
    stateStore.start();
    Context context = createContext(conf, stateStore);
    ContainerManagerImpl cm = createContainerManager(context);
    cm.init(conf);
    cm.start();
    // add an application by starting a container
    String appUser = "app_user1";
    String modUser = "modify_user1";
    String viewUser = "view_user1";
    String enemyUser = "enemy_user";
    ApplicationId appId = ApplicationId.newInstance(0, 1);
    ApplicationAttemptId attemptId = ApplicationAttemptId.newInstance(appId, 1);
    ContainerId cid = ContainerId.newContainerId(attemptId, 1);
    Map<String, LocalResource> localResources = Collections.emptyMap();
    Map<String, String> containerEnv = Collections.emptyMap();
    List<String> containerCmds = Collections.emptyList();
    Map<String, ByteBuffer> serviceData = Collections.emptyMap();
    Credentials containerCreds = new Credentials();
    DataOutputBuffer dob = new DataOutputBuffer();
    containerCreds.writeTokenStorageToStream(dob);
    ByteBuffer containerTokens = ByteBuffer.wrap(dob.getData(), 0, dob.getLength());
    Map<ApplicationAccessType, String> acls = new HashMap<ApplicationAccessType, String>();
    acls.put(ApplicationAccessType.MODIFY_APP, modUser);
    acls.put(ApplicationAccessType.VIEW_APP, viewUser);
    ContainerLaunchContext clc = ContainerLaunchContext.newInstance(localResources, containerEnv, containerCmds, serviceData, containerTokens, acls);
    // create the logAggregationContext
    LogAggregationContext logAggregationContext = LogAggregationContext.newInstance("includePattern", "excludePattern", "includePatternInRollingAggregation", "excludePatternInRollingAggregation");
    StartContainersResponse startResponse = startContainer(context, cm, cid, clc, logAggregationContext);
    assertTrue(startResponse.getFailedRequests().isEmpty());
    assertEquals(1, context.getApplications().size());
    Application app = context.getApplications().get(appId);
    assertNotNull(app);
    waitForAppState(app, ApplicationState.INITING);
    assertTrue(context.getApplicationACLsManager().checkAccess(UserGroupInformation.createRemoteUser(modUser), ApplicationAccessType.MODIFY_APP, appUser, appId));
    assertFalse(context.getApplicationACLsManager().checkAccess(UserGroupInformation.createRemoteUser(viewUser), ApplicationAccessType.MODIFY_APP, appUser, appId));
    assertTrue(context.getApplicationACLsManager().checkAccess(UserGroupInformation.createRemoteUser(viewUser), ApplicationAccessType.VIEW_APP, appUser, appId));
    assertFalse(context.getApplicationACLsManager().checkAccess(UserGroupInformation.createRemoteUser(enemyUser), ApplicationAccessType.VIEW_APP, appUser, appId));
    // reset container manager and verify app recovered with proper acls
    cm.stop();
    context = createContext(conf, stateStore);
    cm = createContainerManager(context);
    cm.init(conf);
    cm.start();
    assertEquals(1, context.getApplications().size());
    app = context.getApplications().get(appId);
    assertNotNull(app);
    // check whether LogAggregationContext is recovered correctly
    LogAggregationContext recovered = ((ApplicationImpl) app).getLogAggregationContext();
    assertNotNull(recovered);
    assertEquals(logAggregationContext.getIncludePattern(), recovered.getIncludePattern());
    assertEquals(logAggregationContext.getExcludePattern(), recovered.getExcludePattern());
    assertEquals(logAggregationContext.getRolledLogsIncludePattern(), recovered.getRolledLogsIncludePattern());
    assertEquals(logAggregationContext.getRolledLogsExcludePattern(), recovered.getRolledLogsExcludePattern());
    waitForAppState(app, ApplicationState.INITING);
    assertTrue(context.getApplicationACLsManager().checkAccess(UserGroupInformation.createRemoteUser(modUser), ApplicationAccessType.MODIFY_APP, appUser, appId));
    assertFalse(context.getApplicationACLsManager().checkAccess(UserGroupInformation.createRemoteUser(viewUser), ApplicationAccessType.MODIFY_APP, appUser, appId));
    assertTrue(context.getApplicationACLsManager().checkAccess(UserGroupInformation.createRemoteUser(viewUser), ApplicationAccessType.VIEW_APP, appUser, appId));
    assertFalse(context.getApplicationACLsManager().checkAccess(UserGroupInformation.createRemoteUser(enemyUser), ApplicationAccessType.VIEW_APP, appUser, appId));
    // simulate application completion
    List<ApplicationId> finishedApps = new ArrayList<ApplicationId>();
    finishedApps.add(appId);
    app.handle(new ApplicationFinishEvent(appId, "Application killed by ResourceManager"));
    waitForAppState(app, ApplicationState.APPLICATION_RESOURCES_CLEANINGUP);
    // restart and verify app is marked for finishing
    cm.stop();
    context = createContext(conf, stateStore);
    cm = createContainerManager(context);
    cm.init(conf);
    cm.start();
    assertEquals(1, context.getApplications().size());
    app = context.getApplications().get(appId);
    assertNotNull(app);
    // no longer saving FINISH_APP event in NM stateStore,
    // simulate by resending FINISH_APP event
    app.handle(new ApplicationFinishEvent(appId, "Application killed by ResourceManager"));
    waitForAppState(app, ApplicationState.APPLICATION_RESOURCES_CLEANINGUP);
    assertTrue(context.getApplicationACLsManager().checkAccess(UserGroupInformation.createRemoteUser(modUser), ApplicationAccessType.MODIFY_APP, appUser, appId));
    assertFalse(context.getApplicationACLsManager().checkAccess(UserGroupInformation.createRemoteUser(viewUser), ApplicationAccessType.MODIFY_APP, appUser, appId));
    assertTrue(context.getApplicationACLsManager().checkAccess(UserGroupInformation.createRemoteUser(viewUser), ApplicationAccessType.VIEW_APP, appUser, appId));
    assertFalse(context.getApplicationACLsManager().checkAccess(UserGroupInformation.createRemoteUser(enemyUser), ApplicationAccessType.VIEW_APP, appUser, appId));
    // simulate log aggregation completion
    app.handle(new ApplicationEvent(app.getAppId(), ApplicationEventType.APPLICATION_RESOURCES_CLEANEDUP));
    assertEquals(app.getApplicationState(), ApplicationState.FINISHED);
    app.handle(new ApplicationEvent(app.getAppId(), ApplicationEventType.APPLICATION_LOG_HANDLING_FINISHED));
    // restart and verify app is no longer present after recovery
    cm.stop();
    context = createContext(conf, stateStore);
    cm = createContainerManager(context);
    cm.init(conf);
    cm.start();
    assertTrue(context.getApplications().isEmpty());
    cm.stop();
}

Example 14 with LogAggregationContext

use of org.apache.hadoop.yarn.api.records.LogAggregationContext in project hadoop by apache.

the class ContainerManagerImpl method recoverApplication.

private void recoverApplication(ContainerManagerApplicationProto p) throws IOException {
    ApplicationId appId = new ApplicationIdPBImpl(p.getId());
    Credentials creds = new Credentials();
    creds.readTokenStorageStream(new DataInputStream(p.getCredentials().newInput()));
    List<ApplicationACLMapProto> aclProtoList = p.getAclsList();
    Map<ApplicationAccessType, String> acls = new HashMap<ApplicationAccessType, String>(aclProtoList.size());
    for (ApplicationACLMapProto aclProto : aclProtoList) {
        acls.put(ProtoUtils.convertFromProtoFormat(aclProto.getAccessType()), aclProto.getAcl());
    }
    LogAggregationContext logAggregationContext = null;
    if (p.getLogAggregationContext() != null) {
        logAggregationContext = new LogAggregationContextPBImpl(p.getLogAggregationContext());
    }
    LOG.info("Recovering application " + appId);
    //TODO: Recover flow and flow run ID
    ApplicationImpl app = new ApplicationImpl(dispatcher, p.getUser(), appId, creds, context, p.getAppLogAggregationInitedTime());
    context.getApplications().put(appId, app);
    app.handle(new ApplicationInitEvent(appId, acls, logAggregationContext));
}

Example 15 with LogAggregationContext

use of org.apache.hadoop.yarn.api.records.LogAggregationContext in project hadoop by apache.

the class TestAppLogAggregatorImpl method createAppLogAggregator.

private static AppLogAggregatorInTest createAppLogAggregator(ApplicationId applicationId, String rootLogDir, YarnConfiguration config, long recoveredLogInitedTimeMillis, DeletionService deletionServiceWithFilesToExpect) throws IOException {
    final Dispatcher dispatcher = createNullDispatcher();
    final NodeId nodeId = NodeId.newInstance("localhost", 0);
    final String userId = "AppLogAggregatorTest";
    final UserGroupInformation ugi = UserGroupInformation.createRemoteUser(userId);
    final LocalDirsHandlerService dirsService = createLocalDirsHandlerService(config, rootLogDir);
    final DeletionService deletionService = deletionServiceWithFilesToExpect;
    final LogAggregationContext logAggregationContext = null;
    final Map<ApplicationAccessType, String> appAcls = new HashMap<>();
    final Context context = createContext(config);
    final FileContext fakeLfs = mock(FileContext.class);
    final Path remoteLogDirForApp = new Path(REMOTE_LOG_FILE.getAbsolutePath());
    return new AppLogAggregatorInTest(dispatcher, deletionService, config, applicationId, ugi, nodeId, dirsService, remoteLogDirForApp, appAcls, logAggregationContext, context, fakeLfs, recoveredLogInitedTimeMillis);
}