[
https://issues.apache.org/jira/browse/FLINK-25646?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17859595#comment-17859595
]
Piotr Nowojski edited comment on FLINK-25646 at 6/24/24 7:39 AM:
-----------------------------------------------------------------
Yes, that would be the best. Neither me nor [~akalashnikov] are currently
looking into it. From our past investigation it looked like subtasks in the
cluster were ending up oscillating in terms of CPU usage/business. They were
100% busy for some time, then idle for a short period, then 100% busy again.
Busy/idle as in terms of Flink's busy/idle metrics. It was strange, clearly
bogus, but we haven't managed to nail it down why is it happening.
was (Author: pnowojski):
Yes, that would be the best. Neither me nor [~akalashnikov] are currently
looking into it. From our past investigation it looked like subtasks in the
cluster were ending up oscillating in terms of CPU usage/business. They were
100% busy for some time, then idle for a short period, then 100% busy again. It
was strange, clearly bogus, but we haven't managed to nail it down why is it
happening.
> Document buffer debloating issues with high parallelism
> -------------------------------------------------------
>
> Key: FLINK-25646
> URL: https://issues.apache.org/jira/browse/FLINK-25646
> Project: Flink
> Issue Type: Improvement
> Components: Documentation, Runtime / Network
> Affects Versions: 1.14.0
> Reporter: Anton Kalashnikov
> Assignee: Anton Kalashnikov
> Priority: Major
> Labels: pull-request-available
> Fix For: 1.15.0
>
>
> According to last benchmarks, there are some problems with buffer debloat
> when job has high parallelism. The high parallelism means the different value
> from job to job but in general it is more than 200. So it makes sense to
> document that problem and propose the solution - increasing the number of
> buffers.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
