yarn.resourcemanager.fail-fast is used inconsistently. Contributed by Yuanbo Liu.

(cherry picked from commit d9ba6f3656e8dc97d2813181e27d12e52dca4328)

hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/conf/capacity-scheduler.xml

@@ -207,4 +207,14 @@
     </description>
   </property>
 
+
+  <property>
+    <name>yarn.scheduler.capacity.application.fail-fast</name>
+    <value>false</value>
+    <description>
+      Whether RM should fail during recovery if previous applications'
+      queue is no longer valid.
+    </description>
+  </property>
+
 </configuration>

hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/capacity/CapacityScheduler.java

@@ -749,7 +749,7 @@ public class CapacityScheduler extends
       if (queue == null) {
         //During a restart, this indicates a queue was removed, which is
         //not presently supported
-        if (!YarnConfiguration.shouldRMFailFast(getConfig())) {
+        if (!getConfiguration().shouldAppFailFast(getConfig())) {
           this.rmContext.getDispatcher().getEventHandler().handle(
               new RMAppEvent(applicationId, RMAppEventType.KILL,
                   "Application killed on recovery as it"
@@ -770,7 +770,7 @@ public class CapacityScheduler extends
       if (!(queue instanceof LeafQueue)) {
         // During RM restart, this means leaf queue was converted to a parent
         // queue, which is not supported for running apps.
-        if (!YarnConfiguration.shouldRMFailFast(getConfig())) {
+        if (!getConfiguration().shouldAppFailFast(getConfig())) {
           this.rmContext.getDispatcher().getEventHandler().handle(
               new RMAppEvent(applicationId, RMAppEventType.KILL,
                   "Application killed on recovery as it was "
@@ -829,7 +829,7 @@ public class CapacityScheduler extends
           return autoCreateLeafQueue(placementContext);
         } catch (YarnException | IOException e) {
           if (isRecovery) {
-            if (!YarnConfiguration.shouldRMFailFast(getConfig())) {
+            if (!getConfiguration().shouldAppFailFast(getConfig())) {
               LOG.error("Could not auto-create leaf queue " + queueName +
                   " due to : ", e);
               this.rmContext.getDispatcher().getEventHandler().handle(

hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/capacity/CapacitySchedulerConfiguration.java

@@ -248,6 +248,12 @@ public class CapacitySchedulerConfiguration extends ReservationSchedulerConfigur
   public static final String SCHEDULE_ASYNCHRONOUSLY_MAXIMUM_PENDING_BACKLOGS =
       SCHEDULE_ASYNCHRONOUSLY_PREFIX + ".maximum-pending-backlogs";
 
+  @Private
+  public static final String APP_FAIL_FAST = PREFIX + "application.fail-fast";
+
+  @Private
+  public static final boolean DEFAULT_APP_FAIL_FAST = false;
+
   @Private
   public static final Integer
       DEFAULT_SCHEDULE_ASYNCHRONOUSLY_MAXIMUM_PENDING_BACKLOGS = 100;
@@ -1287,6 +1293,10 @@ public class CapacitySchedulerConfiguration extends ReservationSchedulerConfigur
     return getBoolean(LAZY_PREEMPTION_ENABLED, DEFAULT_LAZY_PREEMPTION_ENABLED);
   }
 
+  public boolean shouldAppFailFast(Configuration conf) {
+    return conf.getBoolean(APP_FAIL_FAST, DEFAULT_APP_FAIL_FAST);
+  }
+
   private static final String PREEMPTION_CONFIG_PREFIX =
       "yarn.resourcemanager.monitor.capacity.preemption.";
 

hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/test/java/org/apache/hadoop/yarn/server/resourcemanager/TestWorkPreservingRMRestart.java

@@ -760,6 +760,7 @@ public class TestWorkPreservingRMRestart extends ParameterizedSchedulerTestBase
       MockMemoryRMStateStore memStore, RMState state) throws Exception {
     // Restart RM with fail-fast as false. App should be killed.
     csConf.setBoolean(YarnConfiguration.RM_FAIL_FAST, false);
+    csConf.setBoolean(CapacitySchedulerConfiguration.APP_FAIL_FAST, false);
     rm2 = new MockRM(csConf, memStore);
     rm2.start();
 
@@ -794,6 +795,7 @@ public class TestWorkPreservingRMRestart extends ParameterizedSchedulerTestBase
 
     // Now restart RM with fail-fast as true. QueueException should be thrown.
     csConf.setBoolean(YarnConfiguration.RM_FAIL_FAST, true);
+    csConf.setBoolean(CapacitySchedulerConfiguration.APP_FAIL_FAST, true);
     MockRM rm = new MockRM(csConf, memStore2);
     try {
       rm.start();