[jira] [Commented] (FLINK-6685) SafetyNetCloseableRegistry is closed prematurely in Task::triggerCheckpointBarrier

["ASF GitHub Bot (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/FLINK-6685?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16046552#comment-16046552 ] 

ASF GitHub Bot commented on FLINK-6685:
---------------------------------------

GitHub user StefanRRichter opened a pull request:

    https://github.com/apache/flink/pull/4108

    [FLINK-6685] Adjust scopes of SafetyNetCloseableRegistry usage

    This PR addresses remaining issues with the `SafetyNetCloseableregistry`:
    
    - reset the registry to `null` after usage in `Task`.
    - use the registry in the scope of the async checkpointing thread. 


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

    $ git pull https://github.com/StefanRRichter/flink adjustSafetyNetScope

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

    https://github.com/apache/flink/pull/4108.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 #4108
    
----
commit e2031db5f9171d692a41f04d1f3d1ebf592ef6b1
Author: Stefan Richter <s....@data-artisans.com>
Date:   2017-06-12T09:48:15Z

    [FLINK-6685] Adjust scopes of SafetyNetCloseableRegistry usage

----


> SafetyNetCloseableRegistry is closed prematurely in Task::triggerCheckpointBarrier
> ----------------------------------------------------------------------------------
>
>                 Key: FLINK-6685
>                 URL: https://issues.apache.org/jira/browse/FLINK-6685
>             Project: Flink
>          Issue Type: Bug
>    Affects Versions: 1.3.0
>            Reporter: Stefan Richter
>            Assignee: Stefan Richter
>            Priority: Blocker
>
> The {{SafetyNetCloseableRegistry}} is closed to early in {{triggerCheckpointBarrier(...)}}. Right now, it seems like the code assumes that {{statefulTask.triggerCheckpoint(...)}} is blocking - which it is not. Like this, the registry can be closed while the checkpoint is still running.



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