Skip to content

Set JTREG concurrency based on cores and memory size - #1427

Merged
smlambert merged 1 commit into
adoptium:masterfrom
jerboaa:fix_openjdk_concurrency
Nov 7, 2019
Merged

smlambert merged 1 commit into
adoptium:masterfrom
jerboaa:fix_openjdk_concurrency

Conversation

@jerboaa

@jerboaa jerboaa commented Nov 6, 2019

Copy link
Copy Markdown
Contributor

Upstream OpenJDK test framework, roughly, sets it to
min(NPROCS/2, MEM_IN_GB/2). Set it to a similar value
so as to avoid test reliability issues due to memory
limits.

Upstream OpenJDK test framework, roughly, sets it to
min(NPROCS/2, MEM_IN_GB/2). Set it to a similar value
so as to avoid test reliability issues due to memory
limits.
@jerboaa

jerboaa commented Nov 6, 2019

Copy link
Copy Markdown
Contributor Author

Grinder on the big Aarch64 box with this:
https://ci.adoptopenjdk.net/job/Grinder_Sandbox/884/console

It's rather conservative but certainly better than a concurrency value of NPROCS. I'll do some more testing on other boxes and may amend it. Hence, the WIP suffix.

@jerboaa

jerboaa commented Nov 6, 2019

Copy link
Copy Markdown
Contributor Author

@smlambert Thoughts?

@smlambert

Copy link
Copy Markdown
Contributor

Thanks @jerboaa - have been meaning to look at what we were currently doing, based on the troubles seen running on the 'big box' / test-packet-armv8-ubuntu-16-04 reported in #1203 and seen in a bunch of the recent release runs (seen more frequently now that there are less aarch64 machines in the pool.

Agree its an improvement, will be good to see it run through the other platforms to make sure it doesn't hose anything, thanks for looking at this !

@jerboaa

jerboaa commented Nov 7, 2019 •

Copy link
Copy Markdown
Contributor Author

@jerboaa

jerboaa commented Nov 7, 2019

Copy link
Copy Markdown
Contributor Author

Another run on the 96 core aarch64 box (on upstream aarch64 JDK 11.0.6+1 EA build):
https://ci.adoptopenjdk.net/job/Grinder_Sandbox/889/

Contrast this to flaky runs earlier on the same JDK: 11.0.6+1 EA. For example:
https://ci.adoptopenjdk.net/view/Test_upstream/job/Test_openjdk11_hs_sanity.openjdk_aarch64_linux_upstream/23/consoleFull

@jerboaa

jerboaa commented Nov 7, 2019

Copy link
Copy Markdown
Contributor Author

This seems to build OK on Windows/Linux/Mac. I'm removing the WIP prefix.

@jerboaa jerboaa changed the title [WIP] Set JTREG concurrency based on cores and memory size Set JTREG concurrency based on cores and memory size Nov 7, 2019
@jerboaa

jerboaa commented Nov 7, 2019

Copy link
Copy Markdown
Contributor Author

@smlambert More testing of this completed. Make file seems OK. I don't think any of the failures are related to this patch. Mac/Win (no JRE 8 got downloaded, hence failure). Aarch64 linux on the big box now passes (or is significantly better than before in terms of reliability). ppcle 64 bit linux failures for JDK 13 also seem unrelated.

I'd appreciate somebody reviewing this! Thanks!

@smlambert
smlambert merged commit e0b7dd0 into adoptium:master Nov 7, 2019
@jerboaa

jerboaa commented Nov 8, 2019

Copy link
Copy Markdown
Contributor Author

Thanks, @smlambert!

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants