diff --git a/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/ResourceManager.java b/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/ResourceManager.java index d392410..aada69f 100644 --- a/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/ResourceManager.java +++ b/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/ResourceManager.java @@ -817,6 +817,33 @@ public void handle(RMAppAttemptEvent event) { LOG.error("Error in handling event type " + event.getType() + " for applicationAttempt " + appAttemptId, t); } + } else if (rmApp.getApplicationSubmissionContext() != null + && rmApp.getApplicationSubmissionContext() + .getKeepContainersAcrossApplicationAttempts() + && event.getType() == RMAppAttemptEventType.CONTAINER_FINISHED) { + // For work-preserving AM restart, failed attempts are still + // capturing CONTAINER_FINISHED events and record the finished + // containers which will be used by current attempt. + // We just keep 'yarn.resourcemanager.am.max-attempts' in + // RMStateStore. If the finished container's attempt is deleted, we + // use the first attempt in app.attempts to deal with these events. + + RMAppAttempt previousFailedAttempt = + rmApp.getAppAttempts().values().iterator().next(); + if (previousFailedAttempt != null) { + try { + LOG.debug("Event " + event.getType() + " handled by " + + previousFailedAttempt); + previousFailedAttempt.handle(event); + } catch (Throwable t) { + LOG.error("Error in handling event type " + event.getType() + + " for applicationAttempt " + appAttemptId + + " with " + previousFailedAttempt, t); + } + } else { + LOG.error("Event " + event.getType() + + " not handled, because previousFailedAttempt is null"); + } } } } diff --git a/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/ResourceTrackerService.java b/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/ResourceTrackerService.java index bd24b25..902244b 100644 --- a/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/ResourceTrackerService.java +++ b/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/ResourceTrackerService.java @@ -230,6 +230,11 @@ void handleNMContainerStatus(NMContainerStatus containerStatus, NodeId nodeId) { } RMAppAttempt rmAppAttempt = rmApp.getRMAppAttempt(appAttemptId); + if (rmAppAttempt == null) { + LOG.info("Ignoring not found attempt " + appAttemptId); + return; + } + Container masterContainer = rmAppAttempt.getMasterContainer(); if (masterContainer.getId().equals(containerStatus.getContainerId()) && containerStatus.getContainerState() == ContainerState.COMPLETE) { diff --git a/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/FileSystemRMStateStore.java b/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/FileSystemRMStateStore.java index a1cebf5..021ca36 100644 --- a/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/FileSystemRMStateStore.java +++ b/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/FileSystemRMStateStore.java @@ -482,6 +482,18 @@ public synchronized void updateApplicationAttemptStateInternal( } @Override + public synchronized void removeApplicationAttemptInternal( + ApplicationAttemptId appAttemptId) + throws Exception { + Path appDirPath = + getAppDir(rmAppRoot, appAttemptId.getApplicationId()); + Path nodeRemovePath = getNodePath(appDirPath, appAttemptId.toString()); + LOG.info("Removing info for attempt: " + appAttemptId + " at: " + + nodeRemovePath); + deleteFileWithRetries(nodeRemovePath); + } + + @Override public synchronized void removeApplicationStateInternal( ApplicationStateData appState) throws Exception { diff --git a/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/LeveldbRMStateStore.java b/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/LeveldbRMStateStore.java index afc6721..045e258 100644 --- a/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/LeveldbRMStateStore.java +++ b/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/LeveldbRMStateStore.java @@ -575,6 +575,28 @@ protected void updateApplicationAttemptStateInternal( } @Override + public synchronized void removeApplicationAttemptInternal( + ApplicationAttemptId attemptId) + throws IOException { + String attemptKey = getApplicationAttemptNodeKey(attemptId); + try { + WriteBatch batch = db.createWriteBatch(); + try { + batch.delete(bytes(attemptKey)); + if (LOG.isDebugEnabled()) { + LOG.debug("Removing state for attempt " + attemptId + " at " + + attemptKey); + } + db.write(batch); + } finally { + batch.close(); + } + } catch (DBException e) { + throw new IOException(e); + } + } + + @Override protected void removeApplicationStateInternal(ApplicationStateData appState) throws IOException { ApplicationId appId = diff --git a/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/MemoryRMStateStore.java b/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/MemoryRMStateStore.java index ce6addb..caaea7e 100644 --- a/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/MemoryRMStateStore.java +++ b/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/MemoryRMStateStore.java @@ -142,6 +142,19 @@ public synchronized void updateApplicationAttemptStateInternal( } @Override + public synchronized void removeApplicationAttemptInternal( + ApplicationAttemptId appAttemptId) throws Exception { + ApplicationStateData appState = + state.getApplicationState().get(appAttemptId.getApplicationId()); + ApplicationAttemptStateData attemptState = + appState.attempts.remove(appAttemptId); + LOG.info("Removing state for attempt: " + appAttemptId); + if (attemptState == null) { + throw new YarnRuntimeException("Application doesn't exist"); + } + } + + @Override public synchronized void removeApplicationStateInternal( ApplicationStateData appState) throws Exception { ApplicationId appId = diff --git a/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/NullRMStateStore.java b/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/NullRMStateStore.java index 96f77f5..f6fd6fe 100644 --- a/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/NullRMStateStore.java +++ b/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/NullRMStateStore.java @@ -132,6 +132,12 @@ protected void updateApplicationAttemptStateInternal(ApplicationAttemptId attemp } @Override + public synchronized void removeApplicationAttemptInternal( + ApplicationAttemptId attemptId) throws Exception { + // Do nothing + } + + @Override public void checkVersion() throws Exception { // Do nothing } diff --git a/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/RMStateStore.java b/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/RMStateStore.java index ec42cbe..ae17aaa 100644 --- a/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/RMStateStore.java +++ b/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/RMStateStore.java @@ -137,6 +137,10 @@ new UpdateAppAttemptTransition()) .addTransition(RMStateStoreState.ACTIVE, EnumSet.of(RMStateStoreState.ACTIVE, RMStateStoreState.FENCED), + RMStateStoreEventType.REMOVE_APP_ATTEMPT, + new RemoveAppAttemptTransition()) + .addTransition(RMStateStoreState.ACTIVE, + EnumSet.of(RMStateStoreState.ACTIVE, RMStateStoreState.FENCED), RMStateStoreEventType.STORE_MASTERKEY, new StoreRMDTMasterKeyTransition()) .addTransition(RMStateStoreState.ACTIVE, @@ -552,6 +556,32 @@ private static RMStateStoreState finalState(boolean isFenced) { return isFenced ? RMStateStoreState.FENCED : RMStateStoreState.ACTIVE; } + private static class RemoveAppAttemptTransition implements + MultipleArcTransition { + @Override + public RMStateStoreState transition(RMStateStore store, + RMStateStoreEvent event) { + if (!(event instanceof RMStateStoreRemoveAppAttemptEvent)) { + // should never happen + LOG.error("Illegal event type: " + event.getClass()); + return RMStateStoreState.ACTIVE; + } + boolean isFenced = false; + ApplicationAttemptId attemptId = + ((RMStateStoreRemoveAppAttemptEvent) event).getApplicationAttemptId(); + ApplicationId appId = attemptId.getApplicationId(); + LOG.info("Removing attempt " + attemptId + " from app: " + appId); + try { + store.removeApplicationAttemptInternal(attemptId); + } catch (Exception e) { + LOG.error("Error removing attempt: " + attemptId, e); + isFenced = store.notifyStoreOperationFailedInternal(e); + } + return finalState(isFenced); + } + } + public RMStateStore() { super(RMStateStore.class.getName()); ReentrantReadWriteLock lock = new ReentrantReadWriteLock(); @@ -983,6 +1013,29 @@ public void removeApplication(RMApp app) { protected abstract void removeApplicationStateInternal( ApplicationStateData appState) throws Exception; + /** + * Non-blocking API + * ResourceManager services call this to remove an attempt from the state + * store + * This does not block the dispatcher threads + * There is no notification of completion for this operation. + */ + @SuppressWarnings("unchecked") + public synchronized void removeApplicationAttempt( + ApplicationAttemptId applicationAttemptId) { + dispatcher.getEventHandler().handle( + new RMStateStoreRemoveAppAttemptEvent(applicationAttemptId)); + } + + /** + * Blocking API + * Derived classes must implement this method to remove the state of specified + * attempt. + */ + protected abstract void removeApplicationAttemptInternal( + ApplicationAttemptId attemptId) throws Exception; + + // TODO: This should eventually become cluster-Id + "AM_RM_TOKEN_SERVICE". See // YARN-1779 public static final Text AM_RM_TOKEN_SERVICE = new Text( diff --git a/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/RMStateStoreEventType.java b/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/RMStateStoreEventType.java index 492826d..b34634d 100644 --- a/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/RMStateStoreEventType.java +++ b/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/RMStateStoreEventType.java @@ -24,6 +24,7 @@ UPDATE_APP, UPDATE_APP_ATTEMPT, REMOVE_APP, + REMOVE_APP_ATTEMPT, FENCED, // Below events should be called synchronously diff --git a/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/RMStateStoreEventType.java b/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/RMStateStoreRemoveAppAttemptEvent.java similarity index 62% copy from hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/RMStateStoreEventType.java copy to hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/RMStateStoreRemoveAppAttemptEvent.java index 492826d..7455c39 100644 --- a/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/RMStateStoreEventType.java +++ b/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/RMStateStoreRemoveAppAttemptEvent.java @@ -18,21 +18,20 @@ package org.apache.hadoop.yarn.server.resourcemanager.recovery; -public enum RMStateStoreEventType { - STORE_APP_ATTEMPT, - STORE_APP, - UPDATE_APP, - UPDATE_APP_ATTEMPT, - REMOVE_APP, - FENCED, +import org.apache.hadoop.yarn.api.records.ApplicationAttemptId; - // Below events should be called synchronously - STORE_MASTERKEY, - REMOVE_MASTERKEY, - STORE_DELEGATION_TOKEN, - REMOVE_DELEGATION_TOKEN, - UPDATE_DELEGATION_TOKEN, - UPDATE_AMRM_TOKEN, - STORE_RESERVATION, - REMOVE_RESERVATION, +/** + * A event used to remove an attempt. + */ +public class RMStateStoreRemoveAppAttemptEvent extends RMStateStoreEvent { + private ApplicationAttemptId applicationAttemptId; + + RMStateStoreRemoveAppAttemptEvent(ApplicationAttemptId applicationAttemptId) { + super(RMStateStoreEventType.REMOVE_APP_ATTEMPT); + this.applicationAttemptId = applicationAttemptId; + } + + public ApplicationAttemptId getApplicationAttemptId() { + return applicationAttemptId; + } } diff --git a/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/ZKRMStateStore.java b/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/ZKRMStateStore.java index ca0f4ac..ddb8a0b 100644 --- a/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/ZKRMStateStore.java +++ b/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/ZKRMStateStore.java @@ -659,6 +659,22 @@ public synchronized void updateApplicationAttemptStateInternal( } @Override + public synchronized void removeApplicationAttemptInternal( + ApplicationAttemptId appAttemptId) + throws Exception { + String appId = appAttemptId.getApplicationId().toString(); + String appIdRemovePath = getNodePath(rmAppRoot, appId); + String attemptIdRemovePath = getNodePath(appIdRemovePath, + appAttemptId.toString()); + + if (LOG.isDebugEnabled()) { + LOG.debug("Removing info for attempt: " + appAttemptId + " at: " + + attemptIdRemovePath); + } + safeDelete(attemptIdRemovePath); + } + + @Override public synchronized void removeApplicationStateInternal( ApplicationStateData appState) throws Exception { diff --git a/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/records/ApplicationStateData.java b/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/records/ApplicationStateData.java index 1d199ed..a439fa8 100644 --- a/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/records/ApplicationStateData.java +++ b/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/records/ApplicationStateData.java @@ -80,6 +80,16 @@ public ApplicationAttemptStateData getAttempt( return attempts.get(attemptId); } + public int getAttemptStartId() { + int min = Integer.MAX_VALUE; + for(ApplicationAttemptId attemptId : attempts.keySet()) { + if (attemptId.getAttemptId() < min) { + min = attemptId.getAttemptId(); + } + } + return min == Integer.MAX_VALUE ? 1 : min; + } + public abstract ApplicationStateDataProto getProto(); /** 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 c4c8d2e..8c687af 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 @@ -149,6 +149,8 @@ private long startTime; private long finishTime = 0; private long storedFinishTime = 0; + private int startAttemptIdInStateStore = 1; + private int currentAttemptId = 1; // This field isn't protected by readlock now. private volatile RMAppAttempt currentAttempt; private String queue; @@ -809,6 +811,8 @@ public void recover(RMState state) { this.storedFinishTime = appState.getFinishTime(); this.startTime = appState.getStartTime(); this.callerContext = appState.getCallerContext(); + this.startAttemptIdInStateStore = appState.getAttemptStartId(); + this.currentAttemptId = startAttemptIdInStateStore; // send the ATS create Event sendATSCreateEvent(this, this.startTime); @@ -822,7 +826,7 @@ public void recover(RMState state) { private void createNewAttempt() { ApplicationAttemptId appAttemptId = - ApplicationAttemptId.newInstance(applicationId, attempts.size() + 1); + ApplicationAttemptId.newInstance(applicationId, currentAttemptId++); BlacklistManager currentAMBlacklist; if (currentAttempt != null) { @@ -1304,6 +1308,9 @@ public RMAppState transition(RMAppImpl app, RMAppEvent event) { + app.attemptFailuresValidityInterval + " milliseconds " : " ") + "is " + numberOfFailure + ". The max attempts is " + app.maxAppAttempts); + + removeExcessAttempts(app); + if (!app.submissionContext.getUnmanagedAM() && numberOfFailure < app.maxAppAttempts) { if (initialState.equals(RMAppState.KILLING)) { @@ -1340,6 +1347,28 @@ public RMAppState transition(RMAppImpl app, RMAppEvent event) { return RMAppState.FINAL_SAVING; } } + + private void removeExcessAttempts(RMAppImpl app) { + while (app.currentAttemptId - app.startAttemptIdInStateStore + > app.maxAppAttempts) { + // attempts' first element is oldest attempt because it is a + // LinkedHashMap + ApplicationAttemptId attemptId = ApplicationAttemptId.newInstance( + app.getApplicationId(), app.startAttemptIdInStateStore); + app.startAttemptIdInStateStore++; + RMAppAttempt oldestAttempt = app.getRMAppAttempt(attemptId); + if (oldestAttempt != null) { + removeAppAttemptFromStateStore(app, oldestAttempt); + } + } + } + + private void removeAppAttemptFromStateStore(RMAppImpl app, + RMAppAttempt appAttempt) { + ApplicationAttemptId attemptId = appAttempt.getAppAttemptId(); + LOG.info("Remove attempt from state store : " + attemptId); + app.rmContext.getStateStore().removeApplicationAttempt(attemptId); + } } @Override 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 acacc40..ee97885 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 @@ -857,6 +857,9 @@ public void testRMAppAttemptFailuresValidityInterval() throws Exception { @SuppressWarnings("resource") MockRM rm2 = new MockRM(conf, memStore); rm2.start(); + ApplicationStateData app1State = + memStore.getState().getApplicationState().get(app1.getApplicationId()); + Assert.assertEquals(1, app1State.getAttemptStartId()); // re-register the NM nm1.setResourceTrackerService(rm2.getResourceTrackerService()); @@ -869,6 +872,7 @@ public void testRMAppAttemptFailuresValidityInterval() throws Exception { nm1.registerNode(Collections.singletonList(status), null); rm2.waitForState(attempt3.getAppAttemptId(), RMAppAttemptState.FAILED); + Assert.assertEquals(2, app1State.getAttemptCount()); rm2.waitForState(app1.getApplicationId(), RMAppState.ACCEPTED); @@ -882,6 +886,7 @@ public void testRMAppAttemptFailuresValidityInterval() throws Exception { nm1 .nodeHeartbeat(am4.getApplicationAttemptId(), 1, ContainerState.COMPLETE); am4.waitForState(RMAppAttemptState.FAILED); + Assert.assertEquals(2, app1State.getAttemptCount()); // can launch the 5th attempt successfully rm2.waitForState(app1.getApplicationId(), RMAppState.ACCEPTED); @@ -895,6 +900,7 @@ public void testRMAppAttemptFailuresValidityInterval() throws Exception { nm1 .nodeHeartbeat(am5.getApplicationAttemptId(), 1, ContainerState.COMPLETE); am5.waitForState(RMAppAttemptState.FAILED); + Assert.assertEquals(2, app1State.getAttemptCount()); rm2.waitForState(app1.getApplicationId(), RMAppState.FAILED); rm1.stop();