[
https://issues.apache.org/jira/browse/STORM-1058?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14940553#comment-14940553
]
Zhuo Liu edited comment on STORM-1058 at 10/1/15 10:43 PM:
-----------------------------------------------------------
Hello rashmikant,
I downloaded the 0.9.5 storm src release, and add the kill_workers codes in,
compile and deploy in one node.
Launched a wordcount topology and then tried to run "storm kill_workers", it
can kill the running workers. However,
the supervisor is still active, it will restart the workers with different PIDs.
BTW, I didn't met the problem of "my-worker-ids returns empty".
And I have all configuration as defaults (no storm.yaml added).
For better test, you can manually kill supervisor process, and then run the
"storm kill_workers".
I did this, and it cleanly killed the running workers (this is also one of the
main purpose for having this CLI,
cleaning up the isolated workers in problematic node).
Zhuo
was (Author: zhuoliu):
Hello rashmikant,
I tried to download the 0.9.5 storm src release, and add the kill_workers codes
in, compile and deploy in one node.
Launched a wordcount topology and then tried to run "storm kill_workers", it
can kill the running workers. However,
the supervisor is still active, it will restart the workers with different PIDs.
BTW, I didn't met the problem of "my-worker-ids returns empty".
And I have all configuration as defaults (no storm.yaml added).
For better test, you can manually kill supervisor process, and then run the
"storm kill_workers".
I did this, and it cleanly killed the running workers (this is also one of the
main purpose for having this CLI,
cleaning up the isolated workers in problematic node).
Zhuo
> create CLI kill_workers to kill workers on a supervisor node
> ------------------------------------------------------------
>
> Key: STORM-1058
> URL: https://issues.apache.org/jira/browse/STORM-1058
> Project: Apache Storm
> Issue Type: Improvement
> Components: storm-core
> Reporter: Zhuo Liu
> Assignee: Zhuo Liu
> Fix For: 0.11.0
>
>
> Kill all the workers running on this supervisor. Currently there is no direct
> way to kill all the running workers from a command line. In addition, this
> CLI can be called by future node health-check script to cleanly wipe out a
> supervisor on an problematic node.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
