You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@samza.apache.org by "Jagadish (JIRA)" <ji...@apache.org> on 2018/04/19 21:17:00 UTC

[jira] [Commented] (SAMZA-1668) StreamProcessor shutdown waits indefinitely for SamzaContainer to stop.

    [ https://issues.apache.org/jira/browse/SAMZA-1668?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16444813#comment-16444813 ] 

Jagadish commented on SAMZA-1668:
---------------------------------

Shanthoosh,
I'd like to frame this problem a bit differently. 

>> If there's a infinite loop(wait) in any step of SamzaContainer shutdown sequence, the whole StreamProcessor shutdown is stalled.

Here are the threads involved:
t1:
1. samzaContainer.shutdown() --> this calls runLoop.shutdown which sets a flag
2. samzaContainer.awaitShutdown(shutdownTimeout);

t2-runLoop-thread:
1. await for currently running operation to complete
2. poll the flag
3. initiate an orderly shutdown of the components

If step-3 in the runLoop-thread takes longer, step-2 in t1 will still proceed and terminate the coordinator right? So, I'm not sure there's a stall here. 

There's a separate question of whether the coordinator should indeed be terminated or we should re-instantiate a new container+coordinator pair.

> StreamProcessor shutdown waits indefinitely for SamzaContainer to stop.
> -----------------------------------------------------------------------
>
>                 Key: SAMZA-1668
>                 URL: https://issues.apache.org/jira/browse/SAMZA-1668
>             Project: Samza
>          Issue Type: Bug
>            Reporter: Shanthoosh Venkataraman
>            Assignee: Shanthoosh Venkataraman
>            Priority: Major
>
> In the current implementation, StreamProcessor shutdown sequence is the following:
>  # Stop the SamzaContainer.
>  # In the SamzaContainer shutdown callback handler, stop the JobCoordinator.
>  # In JobCoordinator callback handler, trigger the shutdown sequence of LocalApplicationRunner.
> If there's a infinite loop(wait) in any step of SamzaContainer shutdown sequence, the whole StreamProcessor shutdown is stalled.



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)