[
https://issues.apache.org/jira/browse/FLINK-26248?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17495867#comment-17495867
]
zhisheng commented on FLINK-26248:
----------------------------------
thanks [~wangyang0918] , I have a question to ask you, if one of my jobs uses
the same cluster.id all the time, stops the job repeatedly and then resumes the
job from the state, is there any problem with the job state and HA-related data
in this case? Because i find the zk data、hdfs checkpoint data、hdfs ha data
will allways in the same folder? Do I need to use a new cluster.id every time I
start the same job? Similar to yarn, will use new application id, so this way
these directories(zk data/hdfs checkpoint data/hdfs ha data) are a unique
> flink job not recover from latest checkpoint on native k8s
> ----------------------------------------------------------
>
> Key: FLINK-26248
> URL: https://issues.apache.org/jira/browse/FLINK-26248
> Project: Flink
> Issue Type: Bug
> Components: Deployment / Kubernetes
> Affects Versions: 1.12.0
> Reporter: zhisheng
> Priority: Major
> Attachments: image-2022-02-22-10-26-53-699.png
>
>
> flink version: 1.12.0
> ha: zk
> mode: application mode
> native k8s
>
> 1、if flink job start with savepoint, running for a while，then i delete the JM
> pod, the job will restart with the old savepoint not the latest checkpoint.
> this is not what i want
> if i delete the TM pod, the job will restart with the latest the
> checkpoint，this is what i want
>
> 2、if start job without savepoint, running for a while，then i delete the JM
> pod, the job will restart from earliest not the latest checkpoint
--
This message was sent by Atlassian Jira
(v8.20.1#820001)
