flink-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "YuFeng Shen (JIRA)" <j...@apache.org>
Subject [jira] [Comment Edited] (FLINK-10274) The stop-cluster.sh cannot stop cluster properly when there are multiple clusters running
Date Fri, 31 Aug 2018 16:39:00 GMT

    [ https://issues.apache.org/jira/browse/FLINK-10274?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16598983#comment-16598983
] 

YuFeng Shen edited comment on FLINK-10274 at 8/31/18 4:38 PM:
--------------------------------------------------------------

After doing some checking I found Flink stop its components(JM&&TM) by reading the
pid file firstly then run cmd kill pid, and Till Rohrmann mentioned by default the pid file
is written to /tmp and has the name flink-<USER>-<FLINK_COMPONENT>.pid and
can control the dir by setting the env.pid.dir configuration in flink-conf.yaml to avoid
this issue temporarily. 

By default should the Flink use the  flink-<USER><*CLUSERID*><FLINK_COMPONENT>.pid
instead? And also the non-public config item env.pid.dir  should be in the official document.


was (Author: shenyufeng):
After doing some checking I found Flink stop its components(JM&&TM) by reading the
pid file firstly then run cmd kill pid, and Till Rohrmann mentioned by default the pid file
is written to /tmp and has the name flink-<USER>-<FLINK_COMPONENT>.pid and
can control the dir by setting the env.pid.dir configuration in flink-conf.yaml to avoid
this issue temporarily. 

By default should the Flink use the  flink-<USER>-<*CLUSERID*>-<FLINK_COMPONENT>.pid
instead? And also the non-public config item env.pid.dir  should be in the official document.

> The stop-cluster.sh cannot stop cluster properly when there are multiple clusters running
> -----------------------------------------------------------------------------------------
>
>                 Key: FLINK-10274
>                 URL: https://issues.apache.org/jira/browse/FLINK-10274
>             Project: Flink
>          Issue Type: Bug
>          Components: Configuration
>    Affects Versions: 1.5.1, 1.5.2, 1.5.3, 1.6.0
>            Reporter: YuFeng Shen
>            Priority: Major
>
> **When you are prepare to do a Flink framework version upgrading by using the strategy
[shadow copy|https://ci.apache.org/projects/flink/flink-docs-release-1.6/ops/upgrading.html#upgrading-the-flink-framework-version]
, then you need to run multiple clusters concurrently,  however when you want to stop the
old version cluster after upgrading ,you would find the stop-cluster.sh wouldn't work as you
expected: the following is the details to duplicate the issue:
>  # There is a running Flink 1.5.x cluster instance
>  # Installing another Flink 1.6.x cluster instance at the same cluster machines
>  # Migrating the jobs from Flink 1.5.x  to Flink 1.6.x 
>  # go to the bin dir of the Flink 1.5.x cluster instance and run stop-cluster.sh 
> You expect the old Flink 1.5.x cluster instance would be stopped ,right? Unfortunately
the stopped cluster is the new installed Flink 1.6.x cluster instance!



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

Mime
View raw message