YARN-4087. Followup fixes after YARN-2019 regarding RM behavior when state-store error occurs. Contributed by Jian He

(cherry picked from commit 9f7fcb54e798cf4fda1ea7972dd96491976e1857)

hadoop-yarn-project/CHANGES.txt

@@ -21,6 +21,9 @@ Release 2.6.2 - UNRELEASED
 
   BUG FIXES
 
+    YARN-4087. Followup fixes after YARN-2019 regarding RM behavior when
+    state-store error occurs. (Jian He via xgong)
+
 Release 2.6.1 - 2015-09-23
 
   INCOMPATIBLE CHANGES

hadoop-yarn-project/hadoop-yarn/hadoop-yarn-api/src/main/java/org/apache/hadoop/yarn/conf/YarnConfiguration.java

@@ -381,7 +381,7 @@ public class YarnConfiguration extends Configuration {
   public static final boolean DEFAULT_RM_RECOVERY_ENABLED = false;
 
   public static final String YARN_FAIL_FAST = YARN_PREFIX + "fail-fast";
-  public static final boolean DEFAULT_YARN_FAIL_FAST = true;
+  public static final boolean DEFAULT_YARN_FAIL_FAST = false;
 
   public static final String RM_FAIL_FAST = RM_PREFIX + "fail-fast";
 

hadoop-yarn-project/hadoop-yarn/hadoop-yarn-common/src/main/resources/yarn-default.xml

@@ -286,9 +286,12 @@
 
   <property>
     <description>Should YARN fail fast if it encounters any errors.
+      This is a global config for all other components including RM,NM etc.
+      If no value is set for component-specific config (e.g yarn.resourcemanager.fail-fast),
+      this value will be the default.
     </description>
     <name>yarn.fail-fast</name>
-    <value>true</value>
+    <value>false</value>
   </property>
 
   <property>

hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/recovery/RMStateStore.java

@@ -42,6 +42,7 @@ import org.apache.hadoop.yarn.api.records.Container;
 import org.apache.hadoop.yarn.api.records.ContainerExitStatus;
 import org.apache.hadoop.yarn.api.records.FinalApplicationStatus;
 import org.apache.hadoop.yarn.api.records.impl.pb.ApplicationSubmissionContextPBImpl;
+import org.apache.hadoop.yarn.conf.HAUtil;
 import org.apache.hadoop.yarn.conf.YarnConfiguration;
 import org.apache.hadoop.yarn.event.AsyncDispatcher;
 import org.apache.hadoop.yarn.event.Dispatcher;
@@ -822,17 +823,19 @@ public abstract class RMStateStore extends AbstractService {
    */
   protected void notifyStoreOperationFailed(Exception failureCause) {
     LOG.error("State store operation failed ", failureCause);
-    if (failureCause instanceof StoreFencedException) {
+    if (HAUtil.isHAEnabled(getConfig())) {
+      LOG.warn("State-store fenced ! Transitioning RM to standby");
       Thread standByTransitionThread =
           new Thread(new StandByTransitionThread());
       standByTransitionThread.setName("StandByTransitionThread Handler");
       standByTransitionThread.start();
+    } else if (YarnConfiguration.shouldRMFailFast(getConfig())) {
+      LOG.fatal("Fail RM now due to state-store error!");
+      rmDispatcher.getEventHandler().handle(
+          new RMFatalEvent(RMFatalEventType.STATE_STORE_OP_FAILED,
+              failureCause));
     } else {
-      if (YarnConfiguration.shouldRMFailFast(getConfig())) {
-        rmDispatcher.getEventHandler().handle(
-            new RMFatalEvent(RMFatalEventType.STATE_STORE_OP_FAILED,
-                failureCause));
-      }
+      LOG.warn("Skip the state-store error.");
     }
   }