[
https://issues.apache.org/jira/browse/FLINK-23466?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17405633#comment-17405633
]
Yingjie Cao commented on FLINK-23466:
-------------------------------------
I have opened a PR which fix the issue of recycle logic and tested the fix
locally, the problem did not produce. For the requestMemorySegmentFromGlobal
logic, it seems not easy to fix. Previously, I was thinking that we can unify
the buffer lister to the availability listener. But I am afraid that it may
cause performance issue when the number of input channels is large. Another way
is to refactor the code and split the LocalBufferPool into two parts (two pool
implementations), one contains the buffer listener logic and serves the input
gates, the other contains the availability logic and serves the result
partition. The logic of LocalBufferPool is pretty complicated. I think it is
better to do some simplification. [~pnowojski] What do you think? Any
suggestions?
> UnalignedCheckpointITCase hangs on Azure
> ----------------------------------------
>
> Key: FLINK-23466
> URL: https://issues.apache.org/jira/browse/FLINK-23466
> Project: Flink
> Issue Type: Bug
> Components: Runtime / Checkpointing
> Affects Versions: 1.14.0
> Reporter: Dawid Wysakowicz
> Assignee: Yingjie Cao
> Priority: Blocker
> Labels: pull-request-available, test-stability
> Fix For: 1.14.0
>
>
> https://dev.azure.com/apache-flink/apache-flink/_build/results?buildId=20813&view=logs&j=a57e0635-3fad-5b08-57c7-a4142d7d6fa9&t=2ef0effc-1da1-50e5-c2bd-aab434b1c5b7&l=16016
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
