You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@flink.apache.org by "Till Rohrmann (JIRA)" <ji...@apache.org> on 2019/01/10 08:10:00 UTC
[jira] [Commented] (FLINK-11298) Scheduling job in the unit of
concurrent groups
[ https://issues.apache.org/jira/browse/FLINK-11298?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16739192#comment-16739192 ]
Till Rohrmann commented on FLINK-11298:
---------------------------------------
Hi [~tiemsn], thanks for opening this issue. I think we first need to finish FLINK-10429 before we can start working on this issue.
> Scheduling job in the unit of concurrent groups
> -----------------------------------------------
>
> Key: FLINK-11298
> URL: https://issues.apache.org/jira/browse/FLINK-11298
> Project: Flink
> Issue Type: Improvement
> Components: JobManager
> Reporter: shuai.xu
> Assignee: shuai.xu
> Priority: Major
>
> Now flink only support two scheduling modes, that's scheduling all tasks Eager for streaming jobs and scheduling all task Lazy_from_source for batch jobs. This is not flexible enough for the various requirements of different job such as FLINK-10240. We proposal a new ConcurrentSchedulingGroup based scheduling strategy which first split a job into serval concurrent groups and then schedule it in the unit of concurrent groups. This strategy will support not only the existing EAGER and LAZY_FROM_SOURCE mode but also other situation such as the Build/Probe in FLINK-10240.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)