[
https://issues.apache.org/jira/browse/FLINK-9465?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17423539#comment-17423539
]
Feifan Wang commented on FLINK-9465:
------------------------------------
Hi [~trohrmann], thanks for reply, I think we can add the "savepoint-timeout"
parameter in the following four places:
REST API :
*
[/jobs/:jobid/savepoints|https://nightlies.apache.org/flink/flink-docs-release-1.14/docs/ops/rest_api/#jobs-jobid-savepoints]
*
[/jobs/:jobid/stop|https://nightlies.apache.org/flink/flink-docs-release-1.14/docs/ops/rest_api/#jobs-jobid-stop]
Command-Line Interface :
* [Creating a
Savepoint|https://nightlies.apache.org/flink/flink-docs-release-1.14/docs/deployment/cli/#creating-a-savepoint]
* [Stopping a Job Gracefully Creating a Final
Savepoint|https://nightlies.apache.org/flink/flink-docs-release-1.14/docs/deployment/cli/#stopping-a-job-gracefully-creating-a-final-savepoint]
BTW, I noticed that there are different styles of parameter formats in rest api
and cli, some are in camel case, and others are in kebab case. Should we use a
uniform format ?
> Specify a separate savepoint timeout option via CLI
> ---------------------------------------------------
>
> Key: FLINK-9465
> URL: https://issues.apache.org/jira/browse/FLINK-9465
> Project: Flink
> Issue Type: Improvement
> Components: Runtime / Checkpointing
> Affects Versions: 1.5.0
> Reporter: Truong Duc Kien
> Assignee: Feifan Wang
> Priority: Minor
> Labels: auto-deprioritized-major, pull-request-available
> Time Spent: 10m
> Remaining Estimate: 0h
>
> Savepoint can take much longer time to perform than checkpoint, especially
> with incremental checkpoint enabled. This leads to a couple of troubles:
> * For our job, we currently have to set the checkpoint timeout much large
> than necessary, otherwise we would be unable to perform savepoint.
> * During rush hour, our cluster would encounter high rate of checkpoint
> timeout due to backpressure, however we're unable to migrate to a larger
> configuration, because savepoint also timeout.
> In my opinion, the timeout for savepoint should be configurable separately,
> both in the config file and as parameter to the savepoint command.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
