From 15744c614bfaed6fac752af4265bea18d7589ff9 Mon Sep 17 00:00:00 2001 From: Vinod Kumar Vavilapalli Date: Wed, 12 Mar 2014 18:26:53 +0000 Subject: [PATCH] YARN-1812. Fixed ResourceManager to synchrously renew tokens after recovery and thus recover app itself synchronously and avoid races with resyncing NodeManagers. Contributed by Jian He. svn merge --ignore-ancestry -c 1576843 ../../trunk/ git-svn-id: https://svn.apache.org/repos/asf/hadoop/common/branches/branch-2@1576844 13f79535-47bb-0310-9956-ffa450edef68 --- hadoop-yarn-project/CHANGES.txt | 4 + .../resourcemanager/ClientRMService.java | 2 +- .../server/resourcemanager/RMAppManager.java | 88 +++++++++++-------- .../resourcemanager/rmapp/RMAppImpl.java | 4 +- .../security/DelegationTokenRenewer.java | 46 +++++----- .../yarn/server/resourcemanager/MockRM.java | 2 +- .../server/resourcemanager/RMHATestBase.java | 3 +- .../resourcemanager/TestAppManager.java | 2 +- .../yarn/server/resourcemanager/TestRM.java | 2 +- .../server/resourcemanager/TestRMRestart.java | 57 ++++++++++++ .../applicationsmanager/TestAMRestart.java | 2 +- .../security/TestDelegationTokenRenewer.java | 16 ++-- 12 files changed, 152 insertions(+), 76 deletions(-) diff --git a/hadoop-yarn-project/CHANGES.txt b/hadoop-yarn-project/CHANGES.txt index 9c2873fcb8b..ba9844d74bf 100644 --- a/hadoop-yarn-project/CHANGES.txt +++ b/hadoop-yarn-project/CHANGES.txt @@ -438,6 +438,10 @@ Release 2.4.0 - UNRELEASED specify host/rack requests without off-switch request. (Wangda Tan via acmurthy) + YARN-1812. Fixed ResourceManager to synchrously renew tokens after recovery + and thus recover app itself synchronously and avoid races with resyncing + NodeManagers. (Jian He via vinodkv) + Release 2.3.1 - UNRELEASED INCOMPATIBLE CHANGES diff --git a/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/ClientRMService.java b/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/ClientRMService.java index 7715addaa99..84a6b7bef57 100644 --- a/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/ClientRMService.java +++ b/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/ClientRMService.java @@ -337,7 +337,7 @@ public class ClientRMService extends AbstractService implements try { // call RMAppManager to submit application directly rmAppManager.submitApplication(submissionContext, - System.currentTimeMillis(), user, false, null); + System.currentTimeMillis(), user); LOG.info("Application with id " + applicationId.getId() + " submitted by user " + user); diff --git a/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/RMAppManager.java b/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/RMAppManager.java index 7855042db78..e4bcd4fb230 100644 --- a/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/RMAppManager.java +++ b/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/RMAppManager.java @@ -263,48 +263,75 @@ public class RMAppManager implements EventHandler, @SuppressWarnings("unchecked") protected void submitApplication( ApplicationSubmissionContext submissionContext, long submitTime, - String user, boolean isRecovered, RMState state) throws YarnException { + String user) throws YarnException { ApplicationId applicationId = submissionContext.getApplicationId(); RMAppImpl application = createAndPopulateNewRMApp(submissionContext, submitTime, user); + ApplicationId appId = submissionContext.getApplicationId(); - if (isRecovered) { - recoverApplication(state, application); - RMAppState rmAppState = - state.getApplicationState().get(applicationId).getState(); - if (isApplicationInFinalState(rmAppState)) { - // We are synchronously moving the application into final state so that - // momentarily client will not see this application in NEW state. Also - // for finished applications we will avoid renewing tokens. - application - .handle(new RMAppEvent(applicationId, RMAppEventType.RECOVER)); - return; - } - } - if (UserGroupInformation.isSecurityEnabled()) { Credentials credentials = null; try { credentials = parseCredentials(submissionContext); + this.rmContext.getDelegationTokenRenewer().addApplicationAsync(appId, + credentials, submissionContext.getCancelTokensWhenComplete()); } catch (Exception e) { - LOG.warn( - "Unable to parse credentials.", e); + LOG.warn("Unable to parse credentials.", e); // Sending APP_REJECTED is fine, since we assume that the // RMApp is in NEW state and thus we haven't yet informed the // scheduler about the existence of the application assert application.getState() == RMAppState.NEW; - this.rmContext.getDispatcher().getEventHandler().handle( - new RMAppRejectedEvent(applicationId, e.getMessage())); + this.rmContext.getDispatcher().getEventHandler() + .handle(new RMAppRejectedEvent(applicationId, e.getMessage())); throw RPCUtil.getRemoteException(e); } - this.rmContext.getDelegationTokenRenewer().addApplication( - applicationId, credentials, - submissionContext.getCancelTokensWhenComplete(), isRecovered); } else { + // Dispatcher is not yet started at this time, so these START events + // enqueued should be guaranteed to be first processed when dispatcher + // gets started. this.rmContext.getDispatcher().getEventHandler() - .handle(new RMAppEvent(applicationId, - isRecovered ? RMAppEventType.RECOVER : RMAppEventType.START)); + .handle(new RMAppEvent(applicationId, RMAppEventType.START)); + } + } + + @SuppressWarnings("unchecked") + protected void + recoverApplication(ApplicationState appState, RMState rmState) + throws Exception { + ApplicationSubmissionContext appContext = + appState.getApplicationSubmissionContext(); + ApplicationId appId = appState.getAppId(); + + // create and recover app. + RMAppImpl application = + createAndPopulateNewRMApp(appContext, appState.getSubmitTime(), + appState.getUser()); + application.recover(rmState); + if (isApplicationInFinalState(appState.getState())) { + // We are synchronously moving the application into final state so that + // momentarily client will not see this application in NEW state. Also + // for finished applications we will avoid renewing tokens. + application.handle(new RMAppEvent(appId, RMAppEventType.RECOVER)); + return; + } + + if (UserGroupInformation.isSecurityEnabled()) { + Credentials credentials = null; + try { + credentials = parseCredentials(appContext); + // synchronously renew delegation token on recovery. + rmContext.getDelegationTokenRenewer().addApplicationSync(appId, + credentials, appContext.getCancelTokensWhenComplete()); + application.handle(new RMAppEvent(appId, RMAppEventType.RECOVER)); + } catch (Exception e) { + LOG.warn("Unable to parse and renew delegation tokens.", e); + this.rmContext.getDispatcher().getEventHandler() + .handle(new RMAppRejectedEvent(appId, e.getMessage())); + throw e; + } + } else { + application.handle(new RMAppEvent(appId, RMAppEventType.RECOVER)); } } @@ -363,16 +390,6 @@ public class RMAppManager implements EventHandler, } } - private void recoverApplication(RMState state, RMAppImpl application) - throws YarnException { - try { - application.recover(state); - } catch (Exception e) { - LOG.error("Error recovering application", e); - throw new YarnException(e); - } - } - private boolean isApplicationInFinalState(RMAppState rmAppState) { if (rmAppState == RMAppState.FINISHED || rmAppState == RMAppState.FAILED || rmAppState == RMAppState.KILLED) { @@ -403,8 +420,7 @@ public class RMAppManager implements EventHandler, Map appStates = state.getApplicationState(); LOG.info("Recovering " + appStates.size() + " applications"); for (ApplicationState appState : appStates.values()) { - submitApplication(appState.getApplicationSubmissionContext(), - appState.getSubmitTime(), appState.getUser(), true, state); + recoverApplication(appState, state); } } diff --git a/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/rmapp/RMAppImpl.java b/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/rmapp/RMAppImpl.java index 0625708d182..8814100ac1b 100644 --- a/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/rmapp/RMAppImpl.java +++ b/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/rmapp/RMAppImpl.java @@ -731,7 +731,9 @@ public class RMAppImpl implements RMApp, Recoverable { * Therefore we should wait for it to finish. */ for (RMAppAttempt attempt : app.getAppAttempts().values()) { - app.dispatcher.getEventHandler().handle( + // synchronously recover attempt to ensure any incoming external events + // to be processed after the attempt processes the recover event. + attempt.handle( new RMAppAttemptEvent(attempt.getAppAttemptId(), RMAppAttemptEventType.RECOVER)); } diff --git a/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/security/DelegationTokenRenewer.java b/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/security/DelegationTokenRenewer.java index ce9f7ae0625..82464cfeaf7 100644 --- a/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/security/DelegationTokenRenewer.java +++ b/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/security/DelegationTokenRenewer.java @@ -114,6 +114,7 @@ public class DelegationTokenRenewer extends AbstractService { YarnConfiguration.DEFAULT_RM_NM_EXPIRY_INTERVAL_MS); renewerService = createNewThreadPoolService(conf); pendingEventQueue = new LinkedBlockingQueue(); + renewalTimer = new Timer(true); super.serviceInit(conf); } @@ -136,7 +137,6 @@ public class DelegationTokenRenewer extends AbstractService { @Override protected void serviceStart() throws Exception { dtCancelThread.start(); - renewalTimer = new Timer(true); if (tokenKeepAliveEnabled) { delayedRemovalThread = new Thread(new DelayedTokenRemovalRunnable(getConfig()), @@ -151,12 +151,12 @@ public class DelegationTokenRenewer extends AbstractService { isServiceStarted = true; serviceStateLock.writeLock().unlock(); while(!pendingEventQueue.isEmpty()) { - processDelegationTokenRewewerEvent(pendingEventQueue.take()); + processDelegationTokenRenewerEvent(pendingEventQueue.take()); } super.serviceStart(); } - private void processDelegationTokenRewewerEvent( + private void processDelegationTokenRenewerEvent( DelegationTokenRenewerEvent evt) { serviceStateLock.readLock().lock(); try { @@ -325,19 +325,26 @@ public class DelegationTokenRenewer extends AbstractService { } /** - * Add application tokens for renewal. + * Asynchronously add application tokens for renewal. * @param applicationId added application * @param ts tokens * @param shouldCancelAtEnd true if tokens should be canceled when the app is * done else false. * @throws IOException */ - public void addApplication( - ApplicationId applicationId, Credentials ts, boolean shouldCancelAtEnd, - boolean isApplicationRecovered) { - processDelegationTokenRewewerEvent(new DelegationTokenRenewerAppSubmitEvent( - applicationId, ts, - shouldCancelAtEnd, isApplicationRecovered)); + public void addApplicationAsync(ApplicationId applicationId, Credentials ts, + boolean shouldCancelAtEnd) { + processDelegationTokenRenewerEvent(new DelegationTokenRenewerAppSubmitEvent( + applicationId, ts, shouldCancelAtEnd)); + } + + /** + * Synchronously renew delegation tokens. + */ + public void addApplicationSync(ApplicationId applicationId, Credentials ts, + boolean shouldCancelAtEnd) throws IOException{ + handleAppSubmitEvent(new DelegationTokenRenewerAppSubmitEvent( + applicationId, ts, shouldCancelAtEnd)); } private void handleAppSubmitEvent(DelegationTokenRenewerAppSubmitEvent evt) @@ -493,7 +500,7 @@ public class DelegationTokenRenewer extends AbstractService { * @param applicationId completed application */ public void applicationFinished(ApplicationId applicationId) { - processDelegationTokenRewewerEvent(new DelegationTokenRenewerEvent( + processDelegationTokenRenewerEvent(new DelegationTokenRenewerEvent( applicationId, DelegationTokenRenewerEventType.FINISH_APPLICATION)); } @@ -638,9 +645,7 @@ public class DelegationTokenRenewer extends AbstractService { // Setup tokens for renewal DelegationTokenRenewer.this.handleAppSubmitEvent(event); rmContext.getDispatcher().getEventHandler() - .handle(new RMAppEvent(event.getApplicationId(), - event.isApplicationRecovered() ? RMAppEventType.RECOVER - : RMAppEventType.START)); + .handle(new RMAppEvent(event.getApplicationId(), RMAppEventType.START)); } catch (Throwable t) { LOG.warn( "Unable to add the application to the delegation token renewer.", @@ -654,20 +659,17 @@ public class DelegationTokenRenewer extends AbstractService { } } - class DelegationTokenRenewerAppSubmitEvent extends + private static class DelegationTokenRenewerAppSubmitEvent extends DelegationTokenRenewerEvent { private Credentials credentials; private boolean shouldCancelAtEnd; - private boolean isAppRecovered; public DelegationTokenRenewerAppSubmitEvent(ApplicationId appId, - Credentials credentails, boolean shouldCancelAtEnd, - boolean isApplicationRecovered) { + Credentials credentails, boolean shouldCancelAtEnd) { super(appId, DelegationTokenRenewerEventType.VERIFY_AND_START_APPLICATION); this.credentials = credentails; this.shouldCancelAtEnd = shouldCancelAtEnd; - this.isAppRecovered = isApplicationRecovered; } public Credentials getCredentials() { @@ -677,10 +679,6 @@ public class DelegationTokenRenewer extends AbstractService { public boolean shouldCancelAtEnd() { return shouldCancelAtEnd; } - - public boolean isApplicationRecovered() { - return isAppRecovered; - } } enum DelegationTokenRenewerEventType { @@ -688,7 +686,7 @@ public class DelegationTokenRenewer extends AbstractService { FINISH_APPLICATION } - class DelegationTokenRenewerEvent extends + private static class DelegationTokenRenewerEvent extends AbstractEvent { private ApplicationId appId; diff --git a/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/test/java/org/apache/hadoop/yarn/server/resourcemanager/MockRM.java b/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/test/java/org/apache/hadoop/yarn/server/resourcemanager/MockRM.java index 50b42db3dd6..1f3b80cb04b 100644 --- a/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/test/java/org/apache/hadoop/yarn/server/resourcemanager/MockRM.java +++ b/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/test/java/org/apache/hadoop/yarn/server/resourcemanager/MockRM.java @@ -497,7 +497,7 @@ public class MockRM extends ResourceManager { // override to disable webapp } - public static void finishApplicationMaster(RMApp rmApp, MockRM rm, MockNM nm, + public static void finishAMAndVerifyAppState(RMApp rmApp, MockRM rm, MockNM nm, MockAM am) throws Exception { FinishApplicationMasterRequest req = FinishApplicationMasterRequest.newInstance( diff --git a/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/test/java/org/apache/hadoop/yarn/server/resourcemanager/RMHATestBase.java b/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/test/java/org/apache/hadoop/yarn/server/resourcemanager/RMHATestBase.java index 6cbea253774..58258ac0b4a 100644 --- a/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/test/java/org/apache/hadoop/yarn/server/resourcemanager/RMHATestBase.java +++ b/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/test/java/org/apache/hadoop/yarn/server/resourcemanager/RMHATestBase.java @@ -30,7 +30,6 @@ import org.apache.hadoop.yarn.conf.HAUtil; import org.apache.hadoop.yarn.conf.YarnConfiguration; import org.apache.hadoop.yarn.exceptions.YarnException; import org.apache.hadoop.yarn.server.resourcemanager.recovery.ZKRMStateStore; -import org.apache.hadoop.yarn.server.resourcemanager.recovery.RMStateStore.RMState; import org.apache.hadoop.yarn.server.resourcemanager.rmapp.RMApp; import org.apache.hadoop.yarn.server.resourcemanager.rmapp.RMAppImpl; import org.apache.hadoop.yarn.server.resourcemanager.rmapp.RMAppState; @@ -142,7 +141,7 @@ public class RMHATestBase extends ClientBaseWithFixes{ @Override protected void submitApplication( ApplicationSubmissionContext submissionContext, long submitTime, - String user, boolean isRecovered, RMState state) throws YarnException { + String user) throws YarnException { //Do nothing, just add the application to RMContext RMAppImpl application = new RMAppImpl(submissionContext.getApplicationId(), this.rmContext, diff --git a/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/test/java/org/apache/hadoop/yarn/server/resourcemanager/TestAppManager.java b/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/test/java/org/apache/hadoop/yarn/server/resourcemanager/TestAppManager.java index 94db331faf0..19031f91356 100644 --- a/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/test/java/org/apache/hadoop/yarn/server/resourcemanager/TestAppManager.java +++ b/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/test/java/org/apache/hadoop/yarn/server/resourcemanager/TestAppManager.java @@ -178,7 +178,7 @@ public class TestAppManager{ ApplicationSubmissionContext submissionContext, String user) throws YarnException { super.submitApplication(submissionContext, System.currentTimeMillis(), - user, false, null); + user); } } diff --git a/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/test/java/org/apache/hadoop/yarn/server/resourcemanager/TestRM.java b/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/test/java/org/apache/hadoop/yarn/server/resourcemanager/TestRM.java index 7747368dfe0..7906373b5e3 100644 --- a/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/test/java/org/apache/hadoop/yarn/server/resourcemanager/TestRM.java +++ b/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/test/java/org/apache/hadoop/yarn/server/resourcemanager/TestRM.java @@ -414,7 +414,7 @@ public class TestRM { new MockNM("127.0.0.1:1234", 15120, rm1.getResourceTrackerService()); nm1.registerNode(); MockAM am1 = MockRM.launchAndRegisterAM(app1, rm1, nm1); - MockRM.finishApplicationMaster(app1, rm1, nm1, am1); + MockRM.finishAMAndVerifyAppState(app1, rm1, nm1, am1); // a failed app RMApp app2 = rm1.submitApp(200); diff --git a/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/test/java/org/apache/hadoop/yarn/server/resourcemanager/TestRMRestart.java b/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/test/java/org/apache/hadoop/yarn/server/resourcemanager/TestRMRestart.java index ad2e17f730b..484f231927d 100644 --- a/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/test/java/org/apache/hadoop/yarn/server/resourcemanager/TestRMRestart.java +++ b/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/test/java/org/apache/hadoop/yarn/server/resourcemanager/TestRMRestart.java @@ -1709,6 +1709,63 @@ public class TestRMRestart { rm2.stop(); } + // Test Delegation token is renewed synchronously so that recover events + // can be processed before any other external incoming events, specifically + // the ContainerFinished event on NM re-registraton. + @Test (timeout = 20000) + public void testSynchronouslyRenewDTOnRecovery() throws Exception { + conf.setInt(YarnConfiguration.RM_AM_MAX_ATTEMPTS, 2); + conf.set(CommonConfigurationKeysPublic.HADOOP_SECURITY_AUTHENTICATION, + "kerberos"); + MemoryRMStateStore memStore = new MemoryRMStateStore(); + memStore.init(conf); + + // start RM + MockRM rm1 = new MockRM(conf, memStore); + rm1.start(); + final MockNM nm1 = + new MockNM("127.0.0.1:1234", 15120, rm1.getResourceTrackerService()); + nm1.registerNode(); + RMApp app0 = rm1.submitApp(200); + final MockAM am0 = MockRM.launchAndRegisterAM(app0, rm1, nm1); + + MockRM rm2 = new MockRM(conf, memStore) { + @Override + protected ResourceTrackerService createResourceTrackerService() { + return new ResourceTrackerService(this.rmContext, + this.nodesListManager, this.nmLivelinessMonitor, + this.rmContext.getContainerTokenSecretManager(), + this.rmContext.getNMTokenSecretManager()) { + @Override + protected void serviceStart() throws Exception { + // send the container_finished event as soon as the + // ResourceTrackerService is started. + super.serviceStart(); + nm1.setResourceTrackerService(getResourceTrackerService()); + List status = new ArrayList(); + ContainerId amContainer = + ContainerId.newInstance(am0.getApplicationAttemptId(), 1); + status.add(ContainerStatus.newInstance(amContainer, + ContainerState.COMPLETE, "AM container exit", 143)); + nm1.registerNode(status); + } + }; + } + }; + // Re-start RM + rm2.start(); + + // wait for the 2nd attempt to be started. + RMApp loadedApp0 = + rm2.getRMContext().getRMApps().get(app0.getApplicationId()); + int timeoutSecs = 0; + while (loadedApp0.getAppAttempts().size() != 2 && timeoutSecs++ < 40) { + Thread.sleep(200); + } + MockAM am1 = MockRM.launchAndRegisterAM(loadedApp0, rm2, nm1); + MockRM.finishAMAndVerifyAppState(loadedApp0, rm2, nm1, am1); + } + private void writeToHostsFile(String... hosts) throws IOException { if (!hostFile.exists()) { TEMP_DIR.mkdirs(); diff --git a/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/test/java/org/apache/hadoop/yarn/server/resourcemanager/applicationsmanager/TestAMRestart.java b/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/test/java/org/apache/hadoop/yarn/server/resourcemanager/applicationsmanager/TestAMRestart.java index 66d7acb9792..d2f35380ebc 100644 --- a/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/test/java/org/apache/hadoop/yarn/server/resourcemanager/applicationsmanager/TestAMRestart.java +++ b/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/test/java/org/apache/hadoop/yarn/server/resourcemanager/applicationsmanager/TestAMRestart.java @@ -223,7 +223,7 @@ public class TestAMRestart { ((CapacityScheduler) rm1.getResourceScheduler()) .getCurrentAttemptForContainer(containerId2); // finish this application - MockRM.finishApplicationMaster(app1, rm1, nm1, am2); + MockRM.finishAMAndVerifyAppState(app1, rm1, nm1, am2); // the 2nd attempt released the 1st attempt's running container, when the // 2nd attempt finishes. diff --git a/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/test/java/org/apache/hadoop/yarn/server/resourcemanager/security/TestDelegationTokenRenewer.java b/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/test/java/org/apache/hadoop/yarn/server/resourcemanager/security/TestDelegationTokenRenewer.java index a6ad9b68437..786552eb063 100644 --- a/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/test/java/org/apache/hadoop/yarn/server/resourcemanager/security/TestDelegationTokenRenewer.java +++ b/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/test/java/org/apache/hadoop/yarn/server/resourcemanager/security/TestDelegationTokenRenewer.java @@ -353,7 +353,7 @@ public class TestDelegationTokenRenewer { // register the tokens for renewal ApplicationId applicationId_0 = BuilderUtils.newApplicationId(0, 0); - delegationTokenRenewer.addApplication(applicationId_0, ts, true, false); + delegationTokenRenewer.addApplicationAsync(applicationId_0, ts, true); waitForEventsToGetProcessed(delegationTokenRenewer); // first 3 initial renewals + 1 real @@ -393,7 +393,7 @@ public class TestDelegationTokenRenewer { ApplicationId applicationId_1 = BuilderUtils.newApplicationId(0, 1); - delegationTokenRenewer.addApplication(applicationId_1, ts, true, false); + delegationTokenRenewer.addApplicationAsync(applicationId_1, ts, true); waitForEventsToGetProcessed(delegationTokenRenewer); delegationTokenRenewer.applicationFinished(applicationId_1); waitForEventsToGetProcessed(delegationTokenRenewer); @@ -429,7 +429,7 @@ public class TestDelegationTokenRenewer { // register the tokens for renewal ApplicationId appId = BuilderUtils.newApplicationId(0, 0); - delegationTokenRenewer.addApplication(appId, ts, true, false); + delegationTokenRenewer.addApplicationAsync(appId, ts, true); int waitCnt = 20; while (waitCnt-- >0) { if (!eventQueue.isEmpty()) { @@ -473,7 +473,7 @@ public class TestDelegationTokenRenewer { ApplicationId applicationId_1 = BuilderUtils.newApplicationId(0, 1); - delegationTokenRenewer.addApplication(applicationId_1, ts, false, false); + delegationTokenRenewer.addApplicationAsync(applicationId_1, ts, false); waitForEventsToGetProcessed(delegationTokenRenewer); delegationTokenRenewer.applicationFinished(applicationId_1); waitForEventsToGetProcessed(delegationTokenRenewer); @@ -540,7 +540,7 @@ public class TestDelegationTokenRenewer { // register the tokens for renewal ApplicationId applicationId_0 = BuilderUtils.newApplicationId(0, 0); - localDtr.addApplication(applicationId_0, ts, true, false); + localDtr.addApplicationAsync(applicationId_0, ts, true); waitForEventsToGetProcessed(localDtr); if (!eventQueue.isEmpty()){ Event evt = eventQueue.take(); @@ -617,7 +617,7 @@ public class TestDelegationTokenRenewer { // register the tokens for renewal ApplicationId applicationId_0 = BuilderUtils.newApplicationId(0, 0); - localDtr.addApplication(applicationId_0, ts, true, false); + localDtr.addApplicationAsync(applicationId_0, ts, true); localDtr.applicationFinished(applicationId_0); waitForEventsToGetProcessed(delegationTokenRenewer); //Send another keep alive. @@ -718,14 +718,14 @@ public class TestDelegationTokenRenewer { Thread submitThread = new Thread() { @Override public void run() { - dtr.addApplication(mock(ApplicationId.class), creds1, false, false); + dtr.addApplicationAsync(mock(ApplicationId.class), creds1, false); } }; submitThread.start(); // wait till 1st submit blocks, then submit another startBarrier.await(); - dtr.addApplication(mock(ApplicationId.class), creds2, false, false); + dtr.addApplicationAsync(mock(ApplicationId.class), creds2, false); // signal 1st to complete endBarrier.await(); submitThread.join();