[jira] [Commented] (YARN-4893) Fix some intermittent test failures in TestRMAdminService

[Junping Du (JIRA)]

    [ 
https://issues.apache.org/jira/browse/YARN-4893?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15217959#comment-15217959
 ] 

Junping Du commented on YARN-4893:
----------------------------------

Thanks [~brahmareddy] for updating the patch.
Instead of removing exception throw in drianEvents():
{noformat}
@@ -153,9 +153,7 @@ public void drainEvents() {
     Dispatcher rmDispatcher = getRmDispatcher();
     if (rmDispatcher instanceof DrainDispatcher) {
       ((DrainDispatcher) rmDispatcher).await();
-    } else {
-      throw new UnsupportedOperationException("Not a Drain Dispatcher!");
-    }
+    } 
   }
{noformat}

Shall we replace AsyncDispatcher in TestNodeListManager with DrainDispatcher 
instead?
{code}
  private Dispatcher getDispatcher() {
    Dispatcher dispatcher = new AsyncDispatcher() {
{code}

IMO, the previous way of fix could make drainEvents() failed silently (if not a 
DrainDispatcher), that potentially make test failure harder to debug in future.

> Fix some intermittent test failures in TestRMAdminService
> ---------------------------------------------------------
>
>                 Key: YARN-4893
>                 URL: https://issues.apache.org/jira/browse/YARN-4893
>             Project: Hadoop YARN
>          Issue Type: Bug
>            Reporter: Junping Du
>            Assignee: Brahma Reddy Battula
>            Priority: Blocker
>         Attachments: YARN-4893-002.patch, YARN-4893.patch
>
>
> As discussion in YARN-998, we need to add rm.drainEvents() after 
> rm.registerNode() or some of test could get failed intermittently. Also, we 
> can consider to add rm.drainEvents() within rm.registerNode() that could be 
> more convenient.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)