[ 
https://issues.apache.org/jira/browse/DRILL-3449?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16160223#comment-16160223
 ] 

ASF GitHub Bot commented on DRILL-3449:
---------------------------------------

Github user vrozov commented on a diff in the pull request:

    https://github.com/apache/drill/pull/934#discussion_r137941124
  
    --- Diff: 
exec/java-exec/src/main/java/org/apache/drill/exec/work/fragment/FragmentStatusReporter.java
 ---
    @@ -113,4 +120,14 @@ void fail(final UserException ex) {
         sendStatus(status);
       }
     
    +  @Override
    +  public void close()
    +  {
    +    final ControlTunnel tunnel = this.tunnel.getAndSet(null);
    +    if (tunnel != null) {
    +      logger.debug("Closing {}", this);
    --- End diff --
    
    We are closing FragmentStatusReporter, not the `tunnel` that it references. 
The ControlTunnel is not Closable even though it has a reference to a resource 
that is Closable and should provide a way to release the resource it holds. 
Please let me know if a comment is required here, but I do plan to make 
ControlTunnel Closable. As it requires code refactoring not directly related to 
the JIRA/PR, I plan to do this in a separate PR.


> When Foreman node dies, the FragmentExecutor still tries to send status 
> updates to Foreman
> ------------------------------------------------------------------------------------------
>
>                 Key: DRILL-3449
>                 URL: https://issues.apache.org/jira/browse/DRILL-3449
>             Project: Apache Drill
>          Issue Type: Bug
>          Components: Execution - Flow
>            Reporter: Sudheesh Katkam
>            Assignee: Vlad Rozov
>            Priority: Minor
>
> FragmentExecutor.FragmentDrillbitStatusListener discovers that the Foreman 
> died and cancels the fragment which tries to update the Foreman with new 
> CANCELLATION_REQUESTED status. This fails at RPC level. However, this error 
> is not logged nor is it thrown. This just silently happens.
> + The fragment moves to CANCELED state, which is also reported to the Foreman 
> and this also fails.



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)

Reply via email to