[ 
https://issues.apache.org/jira/browse/FLINK-5800?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15866736#comment-15866736
 ] 

ASF GitHub Bot commented on FLINK-5800:
---------------------------------------

GitHub user StephanEwen opened a pull request:

    https://github.com/apache/flink/pull/3312

    [FLINK-5800] [checkpointing] Create CheckpointSteamFactory only once per 
operator

    Previously, the factory was created once per checkpoint, and its repeated 
initialization logic
    (like ensuring existence of base paths) caused heavy load on some 
filesystems at very large scale.
    
    This issue was reported by (and the solution suggested by) Steven Wu.
    
    ## Core Changes
    
    The `CheckpointStreamFactory` is now created once per operator upon 
initialization of the operator. The factory is also stored in the operator, 
alongside the `KeyedStateBackend` and the `OperatorStateBackend`.
    
    ## Tests
    
    This change does not introduce new functionality, but simply re-arranges 
current tests to handle and check the changes instantiation of the 
CheckpointStreamFactory.

You can merge this pull request into a Git repository by running:

    $ git pull https://github.com/StephanEwen/incubator-flink stream_factory

Alternatively you can review and apply these changes as the patch at:

    https://github.com/apache/flink/pull/3312.patch

To close this pull request, make a commit to your master/trunk branch
with (at least) the following in the commit message:

    This closes #3312
    
----

----


> Make sure that the CheckpointStreamFactory is instantiated once per operator 
> only
> ---------------------------------------------------------------------------------
>
>                 Key: FLINK-5800
>                 URL: https://issues.apache.org/jira/browse/FLINK-5800
>             Project: Flink
>          Issue Type: Improvement
>          Components: State Backends, Checkpointing
>    Affects Versions: 1.2.0
>            Reporter: Stephan Ewen
>            Assignee: Stephan Ewen
>             Fix For: 1.3.0, 1.2.1
>
>
> Previously, the {{CheckpointSteamFactory}} was created once per checkpoint, 
> and its repeated initialization logic (like ensuring existence of the parent 
> directory) caused unnecessary load on some file systems at very large scale.



--
This message was sent by Atlassian JIRA
(v6.3.15#6346)

Reply via email to