You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@samza.apache.org by "Chris Riccomini (JIRA)" <ji...@apache.org> on 2015/02/11 19:21:13 UTC

[jira] [Commented] (SAMZA-555) Write a ZK-based samza-standalone JobRunner

    [ https://issues.apache.org/jira/browse/SAMZA-555?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14316706#comment-14316706 ] 

Chris Riccomini commented on SAMZA-555:
---------------------------------------

We should also version the ZK znode data so that a rolling upgrade won't break.

> Write a ZK-based samza-standalone JobRunner
> -------------------------------------------
>
>                 Key: SAMZA-555
>                 URL: https://issues.apache.org/jira/browse/SAMZA-555
>             Project: Samza
>          Issue Type: Sub-task
>          Components: container
>    Affects Versions: 0.9.0
>            Reporter: Chris Riccomini
>
> SAMZA-516 documents the design of a standalone Samza job, which uses ZK for liveness monitoring and partition assignment rather than YARN. The first step to implementing this design is to add a samza-standalone submodule in Samza, which implements a JobRunner that uses ZK to elect a leader for the job (JobCoordinator), and then start a container. When failures occur, the leader should shift partitions within ZK to move partitions off of the dead container. Likewise, when a new container is added, the leader should shift partitions to the new container. The SamzaContainer should remain immutable, and simply be bounced when partitions are added or removed from it.



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