[ 
https://issues.apache.org/jira/browse/FLINK-5556?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15828422#comment-15828422
 ] 

ASF GitHub Bot commented on FLINK-5556:
---------------------------------------

GitHub user uce opened a pull request:

    https://github.com/apache/flink/pull/3164

    [FLINK-5556] [checkpointing] Report correct buffered bytes during alignment

    During stream alignment via the `BarrierTracker`, we were reporting the 
wrong number of spilled bytes, because the `bytesWritten` in `BufferSpiller` 
were reset before the checkpoint notification happens.
    
    Instead of `bufferSpiller.getBytesWritten()` I use 
`SpilledBufferOrEventSequence.size()` if `currentBuffered != null`.
    
    I adjusted one `BarrierBufferTest` to check the expected number of buffered 
bytes. Before the check was only for `>= 0`.

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

    $ git pull https://github.com/uce/flink 5556-reset_bytes

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

    https://github.com/apache/flink/pull/3164.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 #3164
    
----
commit 591fa4c600ef70432e8162584ab89dd5c0508da6
Author: Ufuk Celebi <u...@apache.org>
Date:   2017-01-18T17:14:16Z

    [FLINK-5556] [checkpointing] Report correct buffered bytes during alignment

----


> BarrierBuffer resets bytes written on spiller roll over
> -------------------------------------------------------
>
>                 Key: FLINK-5556
>                 URL: https://issues.apache.org/jira/browse/FLINK-5556
>             Project: Flink
>          Issue Type: Bug
>          Components: State Backends, Checkpointing
>            Reporter: Ufuk Celebi
>            Assignee: Ufuk Celebi
>            Priority: Minor
>
> When rolling over a spilled sequence of buffers, the tracker bytes written 
> are reset to 0. They are reported to the checkpoint listener right after this 
> operation, which results in the reported buffered bytes always being 0.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Reply via email to