[
https://issues.apache.org/jira/browse/FLINK-34907?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17829473#comment-17829473
]
Rui Fan commented on FLINK-34907:
---------------------------------
I see the jobRunningTs is generated from JobDetailsInfo instead of kubernetes
operator. The JobDetailsInfo is got via rest api in autoscaler module, so it
may be a common issue.
Feel free to correct me if my understanding is wrong, thanks a lot.
> jobRunningTs should be the timestamp that all tasks are running
> ---------------------------------------------------------------
>
> Key: FLINK-34907
> URL: https://issues.apache.org/jira/browse/FLINK-34907
> Project: Flink
> Issue Type: Improvement
> Components: Autoscaler
> Reporter: Rui Fan
> Assignee: Rui Fan
> Priority: Major
> Fix For: kubernetes-operator-1.9.0
>
>
> Currently, we consider the timestamp that JobStatus is changed to RUNNING as
> jobRunningTs. But the JobStatus will be RUNNING once job starts schedule, so
> it doesn't mean all tasks are running.
> It will let the isStabilizing or estimating restart time are not accurate.
> Solution: jobRunningTs should be the timestamp that all tasks are running.
> It can be got from SubtasksTimesHeaders rest api.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
