[
https://issues.apache.org/jira/browse/FLINK-16278?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17044676#comment-17044676
]
Andrey Zagrebin commented on FLINK-16278:
-----------------------------------------
There is [chapter in release
notes|https://ci.apache.org/projects/flink/flink-docs-stable/release-notes/flink-1.10.html#memory-management]
about memory changes. It could contain this option. On the other hand, it is
hard to list all changes there. Therefore it refers to [new memory
docs|https://ci.apache.org/projects/flink/flink-docs-stable/ops/memory/mem_setup.html]
which has a [migration
guide|https://ci.apache.org/projects/flink/flink-docs-stable/ops/memory/mem_setup.html]
from previous versions. The new memory docs and migration guide mention this
option
[here|https://ci.apache.org/projects/flink/flink-docs-release-1.10/ops/memory/mem_detail.html#overview],
[here|https://ci.apache.org/projects/flink/flink-docs-release-1.10/ops/memory/mem_detail.html#overview]
and
[here|https://ci.apache.org/projects/flink/flink-docs-release-1.10/ops/memory/mem_detail.html#overview].
The option mentioning also refers to the main config doc page which has the
[default
value|https://ci.apache.org/projects/flink/flink-docs-release-1.10/ops/config.html#taskmanager-memory-jvm-metaspace-size].
We could improve docs the following way:
* The migration guide could be mentioned directly in release notes.
* The migration guide (and possibly release note) could mention that the JVM
meta space is now strictly limited.
> Document new settings for MaxMetaSpace in 1.10
> ----------------------------------------------
>
> Key: FLINK-16278
> URL: https://issues.apache.org/jira/browse/FLINK-16278
> Project: Flink
> Issue Type: Improvement
> Components: Documentation
> Affects Versions: 1.10.0
> Reporter: None none
> Priority: Major
>
> I was running Flink 1.8 and decided to upgrade to 1.10. I started to get
> OutOfMemory: Metaspace on Task nodes and they would shutdown.
> Deploying 3-4 jobs was ok, deploying 5th job, the task manager would shutdown
> after a couple of minutes.
> Some investigation shows, that the Flink startup scripts set this to 100m.
> Setting taskmanager.memory.jvm-metaspace.size: 256m seems to alleviate the
> problem. Looking at the docs, this setting seems new. But it is not
> documented anywhere that the default is now 100m
> I would like to add the exact same setup/jobs on Flink 1.8 was working fine.
> Here is a conversation on the mailing list as well:
> [https://mail.google.com/mail/u/1/#label/Apache+Flink/KtbxLrjNZhsRGptKFCvHHlTWSdxbqKXHBB]
> I think the upgrade/migration docs should indicate this and also maybe the
> default size.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
