[
https://issues.apache.org/jira/browse/STORM-233?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13998943#comment-13998943
]
ASF GitHub Bot commented on STORM-233:
--------------------------------------
Github user nathanmarz commented on the pull request:
https://github.com/apache/incubator-storm/pull/98#issuecomment-43241008
The supervisor gives the worker 30 seconds by default to heartbeat after
the first time. That's controlled by the config supervisor.worker.timeout.secs.
I'd be fine with removing the ZK heartbeat line, thus allowing the worker
to finish launching itself, but I think we should still keep the one that
heartbeats to the supervisor. The sooner it heartbeats to the supervisor the
sooner the supervisor can move on to other tasks.
> Avoid worker killed on startup because of heavy ZK load
> -------------------------------------------------------
>
> Key: STORM-233
> URL: https://issues.apache.org/jira/browse/STORM-233
> Project: Apache Storm (Incubating)
> Issue Type: Sub-task
> Reporter: Robert Joseph Evans
>
> When under a very heavy load ZK can be very slow. If it takes longer than 3
> seconds for a worker to heatbeat in through ZK the first time, this can cause
> the supervisor to time out the worker process and shoot it before it comes up
> fully.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
