Prabhu Joseph created YARN-11285:
------------------------------------

             Summary: LocalizedResources are leaked and its LocalPath are not 
cleared
                 Key: YARN-11285
                 URL: https://issues.apache.org/jira/browse/YARN-11285
             Project: Hadoop YARN
          Issue Type: Bug
          Components: nodemanager
    Affects Versions: 3.2.1
            Reporter: Prabhu Joseph
            Assignee: Prabhu Joseph


LocalizedResources are leaked and its LocalPath are not cleared from NM Local 
Directories. When multiple containers are initialized at same time, 
LocalResourcesTrackerImpl REQUEST handler could create and handle multiple 
LocalizedResource object for the same input path due to race condition in 
[below 
code|https://github.com/apache/hadoop/blob/trunk/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-nodemanager/src/main/java/org/apache/hadoop/yarn/server/nodemanager/containermanager/localizer/LocalResourcesTrackerImpl.java#L149]

{code}
    case REQUEST:
    LocalResourceRequest req = event.getLocalResourceRequest();
    LocalizedResource rsrc = localrsrc.get(req);
     ....
      if (null == rsrc) {
        rsrc = new LocalizedResource(req, dispatcher);
        localrsrc.put(req, rsrc);
      }
     rsrc.handle(event);
{code}


Each container will have separate LocalizedResource object and separate local 
path like below.
{code}
   /mnt/yarn/usercache/hive/filecache/6/2552419:
   total 28456
   -r-x------ 1 yarn yarn 29135164 Aug  7 10:24 
hive-exec-2.3.4.50-3fd48f33b0c0b82ab431013f0fe794dfe75c31a5027567e6865cccbb49de862b.jar

   /mnt/yarn/usercache/hive/filecache/6/2552420:
   total 28456
   -r-x------ 1 yarn yarn 29135164 Aug  7 10:24 
hive-exec-2.3.4.50-3fd48f33b0c0b82ab431013f0fe794dfe75c31a5027567e6865cccbb49de862b.jar

   /mnt/yarn/usercache/hive/filecache/6/2552421:
   total 28456
   -r-x------ 1 yarn yarn 29135164 Aug  7 10:24 
hive-exec-2.3.4.50-3fd48f33b0c0b82ab431013f0fe794dfe75c31a5027567e6865cccbb49de862b.jar

   /mnt/yarn/usercache/hive/filecache/6/2552422:
   total 28456
{code}





--
This message was sent by Atlassian Jira
(v8.20.10#820010)

---------------------------------------------------------------------
To unsubscribe, e-mail: yarn-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: yarn-issues-h...@hadoop.apache.org

Reply via email to