You are viewing a plain text version of this content. The canonical link for it is here.
Posted to yarn-commits@hadoop.apache.org by vi...@apache.org on 2014/02/21 03:20:21 UTC
svn commit: r1570415 - in /hadoop/common/trunk/hadoop-yarn-project:
CHANGES.txt
hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/capacity/LeafQueue.java
Author: vinodkv
Date: Fri Feb 21 02:20:20 2014
New Revision: 1570415
URL: http://svn.apache.org/r1570415
Log:
YARN-1398. Fixed a deadlock in ResourceManager between users requesting queue-acls and completing containers. Contributed by Vinod Kumar Vavilapalli.
Modified:
hadoop/common/trunk/hadoop-yarn-project/CHANGES.txt
hadoop/common/trunk/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/capacity/LeafQueue.java
Modified: hadoop/common/trunk/hadoop-yarn-project/CHANGES.txt
URL: http://svn.apache.org/viewvc/hadoop/common/trunk/hadoop-yarn-project/CHANGES.txt?rev=1570415&r1=1570414&r2=1570415&view=diff
==============================================================================
--- hadoop/common/trunk/hadoop-yarn-project/CHANGES.txt (original)
+++ hadoop/common/trunk/hadoop-yarn-project/CHANGES.txt Fri Feb 21 02:20:20 2014
@@ -321,6 +321,9 @@ Release 2.4.0 - UNRELEASED
YARN-713. Fixed ResourceManager to not crash while building tokens when DNS
issues happen transmittently. (Jian He via vinodkv)
+ YARN-1398. Fixed a deadlock in ResourceManager between users requesting
+ queue-acls and completing containers. (vinodkv)
+
Release 2.3.1 - UNRELEASED
INCOMPATIBLE CHANGES
Modified: hadoop/common/trunk/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/capacity/LeafQueue.java
URL: http://svn.apache.org/viewvc/hadoop/common/trunk/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/capacity/LeafQueue.java?rev=1570415&r1=1570414&r2=1570415&view=diff
==============================================================================
--- hadoop/common/trunk/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/capacity/LeafQueue.java (original)
+++ hadoop/common/trunk/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/capacity/LeafQueue.java Fri Feb 21 02:20:20 2014
@@ -50,7 +50,6 @@ import org.apache.hadoop.yarn.api.record
import org.apache.hadoop.yarn.api.records.QueueUserACLInfo;
import org.apache.hadoop.yarn.api.records.Resource;
import org.apache.hadoop.yarn.api.records.ResourceRequest;
-import org.apache.hadoop.yarn.api.records.Token;
import org.apache.hadoop.yarn.factories.RecordFactory;
import org.apache.hadoop.yarn.factory.providers.RecordFactoryProvider;
import org.apache.hadoop.yarn.server.resourcemanager.rmcontainer.RMContainer;
@@ -1410,12 +1409,14 @@ public class LeafQueue implements CSQueu
FiCaSchedulerApp application, FiCaSchedulerNode node, RMContainer rmContainer,
ContainerStatus containerStatus, RMContainerEventType event, CSQueue childQueue) {
if (application != null) {
+
+ boolean removed = false;
+
// Careful! Locking order is important!
synchronized (this) {
Container container = rmContainer.getContainer();
- boolean removed = false;
// Inform the application & the node
// Note: It's safe to assume that all state changes to RMContainer
// happen under scheduler's lock...
@@ -1441,13 +1442,14 @@ public class LeafQueue implements CSQueu
" absoluteUsedCapacity=" + getAbsoluteUsedCapacity() +
" used=" + usedResources +
" cluster=" + clusterResource);
- // Inform the parent queue
- getParent().completedContainer(clusterResource, application,
- node, rmContainer, null, event, this);
}
}
-
+ if (removed) {
+ // Inform the parent queue _outside_ of the leaf-queue lock
+ getParent().completedContainer(clusterResource, application, node,
+ rmContainer, null, event, this);
+ }
}
}