[
https://issues.apache.org/jira/browse/HBASE-24126?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17077358#comment-17077358
]
Michael Stack commented on HBASE-24126:
---------------------------------------
To be clear (since my application has been messy), this issue has three commits
against it on branch-2; an original commit followed by two addendums:
{code}
commit 01715fbbf6e00edbb4497bfa72e3c2a5b9937088
Author: stack <[email protected]>
Date: Mon Apr 6 18:23:35 2020 -0700
HBASE-24126 Up the container nproc uplimit from 10000 to 12500
Addendum number 2; docker doesn't like the ulimit arg as is.
commit 25773c681b35e231152b3bb2347e7f55f03ba2e4
Author: stack <[email protected]>
Date: Mon Apr 6 16:29:46 2020 -0700
HBASE-24126 Up the container nproc uplimit from 10000 to 12500
ADDENDUM: Add to hbase-personality
commit ed24998b85dec1dddedd636b01d87bcdb0445ef0 (2.up.nproc)
Author: stack <[email protected]>
Date: Mon Apr 6 13:44:53 2020 -0700
HBASE-24126 Up the container nproc uplimit from 10000 to 12500
{code}
A related push against HBASE-24072 is complimentary making use of the changes
made in this JIRA:
{code}
commit 9c60c5fddbd9b1910fa59c44592249696948041a (HEAD -> 2, origin/branch-2)
Author: stack <[email protected]>
Date: Thu Apr 2 18:29:05 2020 -0700
HBASE-24072 Nightlies reporting OutOfMemoryError: unable to create new
native thread.
Up the nproc in hbase-personality. Add test that prints out env.
{code}
Will push everywhere after confirmed all works.
> Up the container nproc uplimit from 10000 to 12500
> --------------------------------------------------
>
> Key: HBASE-24126
> URL: https://issues.apache.org/jira/browse/HBASE-24126
> Project: HBase
> Issue Type: Sub-task
> Components: build
> Reporter: Michael Stack
> Priority: Major
> Attachments:
> 0001-HBASE-24126-Up-the-container-nproc-uplimit-from-1000.patch
>
>
> Let me up the container nproc count from 10000. If many tests running in
> parallel, container could breach the 10000 limit: i.e. if a bunch of
> heavy-duty long tests run at same time (0.5C on a 16CPU machine means 8
> possible concurrents) and we spin up 1-2000 threads in each test, then could
> hit the 10000 limit. Move up the limit some expecially after INFRA just upped
> our limit to 30000.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
