You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@flink.apache.org by "Stephan Ewen (JIRA)" <ji...@apache.org> on 2017/07/28 14:42:02 UTC

[jira] [Commented] (FLINK-7285) allow submission of big job graphs

    [ https://issues.apache.org/jira/browse/FLINK-7285?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16105047#comment-16105047 ] 

Stephan Ewen commented on FLINK-7285:
-------------------------------------

I would not do anything with just now.

All client to JobManager communication is going REST (so no akka limits) and may come out of the container's launch context in many setups (Yarn / Kubernetes). Any JobGraph and submit call optimization now is going against all other current development threads.

It is worth making sure that large REST POST requests to start jobs work on the dispatchers, but that is blocked on the REST client API for now.

I would actually close this issue.

> allow submission of big job graphs
> ----------------------------------
>
>                 Key: FLINK-7285
>                 URL: https://issues.apache.org/jira/browse/FLINK-7285
>             Project: Flink
>          Issue Type: Improvement
>          Components: Distributed Coordination, Job-Submission, Network
>    Affects Versions: 1.4.0
>            Reporter: Nico Kruber
>            Assignee: Nico Kruber
>
> In order to support submission of big job graphs, i.e. from user programs with a lot of code that does not fit into {{akka.framesize}}, we could try to store the {{JobGraph}} in the {{BlobServer}}.
> Note that in the communication between the job and task managers, we also need the offloading of the {{TaskDeploymentDescriptor}} (FLINK-6046) to complete this feature.



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)