Hi,
checkpoints are only used in failover for one job. Once a job is cancelled,
the related checkpoint-count metadata (stored on HA) will be removed.
But the checkpoint data could be retained if you configured it.
IIUC, the redeploy/update job will cancel the old job and then start a new
one.
The
Hey Hang,
I am deploying my Flink Job in HA application mode, Whenever I redeploy my
job, or deploy an updated version of the job, it's using the same job_id. I
haven't configured anywhere to use a fixed job id, I think it's doing it by
default. Can you share where I can configure this? I tried it
Hi Weihua,
I am deploying my flink job in HA application mode on a kubernetes cluster.
I am using an external nfs mount for storing checkpoints. For some reason,
whenever I deploy an updated version of my application, it uses the same
job_id for the new job as for the previous job. Thus the flink
Hi,
if for some reason there exists a checkpoint by same name.
>
Could you give more details about your scenarios here?
>From your description, I guess this problem occurred when a job restart,
does this restart is triggered personally?
In common restart processing, the job will retrieve the late
Hi, amenreet,
As Hangxiang said, we should use a new checkpoint dir if the new job has
the same jobId as the old one . Or else you should not use a fixed jobId
and the checkpoint dir will not conflict.
Best,
Hang
Hangxiang Yu 于2023年5月10日周三 10:35写道:
> Hi,
> I guess you used a fixed JOB_ID, and
Hi,
I guess you used a fixed JOB_ID, and configured the same checkpoint dir as
before ?
And you may also start the job without before state ?
The new job cannot know anything about before checkpoints, that's why the
new job will fail when it tries to generate a new checkpoint.
I'd like to suggest y
Hi all,
Is there any way to prevent restart of flink job, or override the
checkpoint metadata, if for some reason there exists a checkpoint by same
name. I get the following exception and my job restarts, have been trying
to find solution for a very long time but havent found anything useful yet,