[
https://issues.apache.org/jira/browse/FLINK-10292?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16627538#comment-16627538
]
vinoyang commented on FLINK-10292:
----------------------------------
[~uce] Thank you for verifying, my understanding of "image". Then I give some
of my thoughts about the questions you asked before. In fact, I think we can't
help but think about running in image(container), and we can't just consider
the image scene. In my opinion, there is no essential difference between
standalone and running in docker/k8s (virtualization can be seen as a
micro-standalone model? Is it also dependent on Zookeeper in the same
environment).
* Regarding what you said: Killing rather than canceling the scene, the
resulting metadata persists, and the same happens in standalone, which can't be
avoided, because this is not a normal operation.
* Regarding what you said "update the image with a modified application" may
cause the new jobGraph to fail to overwrite the old jobgraph, it should not
happen. Because this can be seen as a scenario of a job upgrade. In order to
achieve this, our first step should be to execute "cancel with savepoint", then
the job will be resubmitted (of course we should recognize it and update the
metadata of the jobgraph stored in Zookeeper), then restore based on savepoint .
My personal opinion is that Flink on k8s does not seem to need to be considered
separately on this issue.
> Generate JobGraph in StandaloneJobClusterEntrypoint only once
> -------------------------------------------------------------
>
> Key: FLINK-10292
> URL: https://issues.apache.org/jira/browse/FLINK-10292
> Project: Flink
> Issue Type: Improvement
> Components: Distributed Coordination
> Affects Versions: 1.6.0, 1.7.0
> Reporter: Till Rohrmann
> Assignee: vinoyang
> Priority: Major
> Fix For: 1.7.0, 1.6.2
>
>
> Currently the {{StandaloneJobClusterEntrypoint}} generates the {{JobGraph}}
> from the given user code every time it starts/is restarted. This can be
> problematic if the the {{JobGraph}} generation has side effects. Therefore,
> it would be better to generate the {{JobGraph}} only once and store it in HA
> storage instead from where to retrieve.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
