[
https://issues.apache.org/jira/browse/BEAM-7684?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17122758#comment-17122758
]
Beam JIRA Bot commented on BEAM-7684:
-------------------------------------
This issue is P2 but has been unassigned without any comment for 60 days so it
has been labeled "stale-P2". If this issue is still affecting you, we care!
Please comment and remove the label. Otherwise, in 14 days the issue will be
moved to P3.
Please see https://beam.apache.org/contribute/jira-priorities/ for a detailed
explanation of what these priorities mean.
> WindmillStateCache grossly misunderestimates object size in its weighting
> function
> ----------------------------------------------------------------------------------
>
> Key: BEAM-7684
> URL: https://issues.apache.org/jira/browse/BEAM-7684
> Project: Beam
> Issue Type: Bug
> Components: runner-dataflow
> Reporter: Steve Niemitz
> Priority: P2
> Labels: stale-P2
>
> As a little experiment, we set the WindmillStateCache size to 2 GB. After we
> let a job run for a few hours, the workers all died from OOMs. Digging into
> the memory usage, over 10 GB of heap was used by the WindmillStateCache. The
> total weight the cache had calculated however was only ~800 MB.
> It looks like the StateCacheEntry calculation is way off. The weight of an
> entry was calculated as only ~200, while the actual heap usage of it (as
> reported by YourKit at least) was around 2,400 bytes.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
