YARN-7511. NPE in ContainerLocalizer when localization failed for run…

…ning container. Contributed by Tao Yang (cherry picked from commit 83798f1)
justinuang · Mar 2, 2018 · 2d313c8 · 2d313c8
1 parent 2599b97
commit 2d313c8
Show file tree

Hide file tree

Showing 2 changed files with 30 additions and 0 deletions.
diff --git a/...ava/org/apache/hadoop/yarn/server/nodemanager/containermanager/localizer/ResourceSet.java b/...ava/org/apache/hadoop/yarn/server/nodemanager/containermanager/localizer/ResourceSet.java
@@ -138,6 +138,10 @@ public Set<String> resourceLocalized(LocalResourceRequest request,
   }
 
   public void resourceLocalizationFailed(LocalResourceRequest request) {
+    // Skip null request when localization failed for running container
+    if (request == null) {
+      return;
+    }
     pendingResources.remove(request);
     resourcesFailedToBeLocalized.add(request);
   }

diff --git a/...a/org/apache/hadoop/yarn/server/nodemanager/containermanager/container/TestContainer.java b/...a/org/apache/hadoop/yarn/server/nodemanager/containermanager/container/TestContainer.java
@@ -378,6 +378,27 @@ public void testLocalizationFailureAtDone() throws Exception {
     }
   }
 
+  @Test
+  @SuppressWarnings("unchecked")
+  public void testLocalizationFailureWhileRunning()
+      throws Exception {
+    WrappedContainer wc = null;
+    try {
+      wc = new WrappedContainer(6, 314159265358979L, 4344, "yak");
+      wc.initContainer();
+      wc.localizeResources();
+      wc.launchContainer();
+      reset(wc.localizerBus);
+      assertEquals(ContainerState.RUNNING, wc.c.getContainerState());
+      // Now in RUNNING, handle ContainerResourceFailedEvent, cause NPE before
+      wc.handleContainerResourceFailedEvent();
+    } finally {
+      if (wc != null) {
+        wc.finished();
+      }
+    }
+  }
+
   @Test
   @SuppressWarnings("unchecked") // mocked generic
   public void testCleanupOnKillRequest() throws Exception {
@@ -1132,6 +1153,11 @@ public void resourceFailedContainer() {
       drainDispatcherEvents();
     }
 
+    public void handleContainerResourceFailedEvent() {
+      c.handle(new ContainerResourceFailedEvent(cId, null, null));
+      drainDispatcherEvents();
+    }
+
     // Localize resources 
     // Skip some resources so as to consider them failed
     public Map<Path, List<String>> doLocalizeResources(