[
https://issues.apache.org/jira/browse/FLINK-4735?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15552140#comment-15552140
]
ASF GitHub Bot commented on FLINK-4735:
---------------------------------------
Github user StephanEwen commented on the issue:
https://github.com/apache/flink/pull/2588
Looks good, thanks, will merge this.
I would suggest two followups:
- We can remove the "dispose savepoint" method, because that is
functionality disappearing right now. Savepoints need no special remove logic
any more, they are just a directory in a filesystem.
- We can reduce the number of message classes (like
`TriggerSavepointResponse`) by using exceptions more prominently. The
"triggerSavepoint()" method could simply return a String (success) or throw an
exception (failure). Exceptions are propagated by the RPC, so the caller future
will always get it as a response.
> Migrate some job execution related akka messages to rpc calls
> -------------------------------------------------------------
>
> Key: FLINK-4735
> URL: https://issues.apache.org/jira/browse/FLINK-4735
> Project: Flink
> Issue Type: Sub-task
> Components: Cluster Management
> Reporter: Kurt Young
> Assignee: Kurt Young
>
> This includes the following operations about job execution:
> 1. checkpointing
> 2. kvstate
> 3. savepoint
> 4. classloading props
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
