[
https://issues.apache.org/jira/browse/FLINK-23535?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17389843#comment-17389843
]
tim yu commented on FLINK-23535:
--------------------------------
Hello [~yunta], A flink job that is runnning on YARN often exits normally or
exceptionally,
The monitor system will submit it automatically again when finding this job
exits. Zookeeper-based HA enables this job to recover its state from the latest
checkpoint, but requiring a fixed cluster ID. I find the
high-availability.cluster-id is set only using zookeeper namespace option now
through reading the code. The high-availability.cluster-id will be replaced
improperly with application ID if setting it in flink-conf.yaml. The
FLINK-19493 causes this issue is very obvious.
The high-availability.cluster-id could be configured explicitly in
flink-conf.yaml for submitting the job to YARN or another.
> The value of option high-availability.cluster-id in flink-conf.yaml is
> replaced improperly
> ------------------------------------------------------------------------------------------
>
> Key: FLINK-23535
> URL: https://issues.apache.org/jira/browse/FLINK-23535
> Project: Flink
> Issue Type: Bug
> Components: Client / Job Submission, Table SQL / Client
> Affects Versions: 1.12.3
> Reporter: tim yu
> Priority: Major
>
> The value of option high-availability.cluster-id in flink-conf.yaml is
> replaced improperly with application ID when submitting a job to YARN.
> I guess that maybe the FLINK-19493 causes this problem.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
