You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@samza.apache.org by "Boris Shkolnik (JIRA)" <ji...@apache.org> on 2017/07/12 20:34:00 UTC

[jira] [Commented] (SAMZA-1360) Zookeeper session state handler for LocalApplicationRunner

    [ https://issues.apache.org/jira/browse/SAMZA-1360?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16084640#comment-16084640 ] 

Boris Shkolnik commented on SAMZA-1360:
---------------------------------------

Yeah, this should be added for the LocalApplicaitonRunner too.
But I think first we need to see if we should get rid of the CoordinationServiceUtils (move the instantiation of ZkUtils to the LocalApplication runner) or combine them.

> Zookeeper session state handler for LocalApplicationRunner
> ----------------------------------------------------------
>
>                 Key: SAMZA-1360
>                 URL: https://issues.apache.org/jira/browse/SAMZA-1360
>             Project: Samza
>          Issue Type: Improvement
>            Reporter: Shanthoosh Venkataraman
>            Priority: Minor
>             Fix For: 0.14.0
>
>
> When zookeeper client reestablishes connection with zookeeper server, it should not receive notifications/events for previous session. Acting on zookeeper events without distinction if they belong to current session leads to incorrect behavior. However, it's discovered from testing that this contract isn't honored(zookeeper client receives events from earlier sessions).
> SAMZA-1336 introduces sessionListeners(which gets triggered onNewSession, newEvent) for ZkJobCoordinator handling stale events. Similar mechanism should be followed for ZkCoordinationService used by LocalApplicationRunner for leader election.



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