Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Update known issue about lmod hook in host-injection #183

Merged
merged 5 commits into from
Jun 25, 2024
Merged

Update known issue about lmod hook in host-injection #183

Changes from 1 commit
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
70004f9
Update known issue about lmod hook in host-injection
xinan1911 Jun 4, 2024
afb7c24
Update mkdocs.yml for structure and remove version pages of known issues
xinan1911 Jun 18, 2024
eb02210
Adding reference to Lmod hooks
xinan1911 Jun 18, 2024
0b6c2e8
Update docs/known_issues/eessi-2023.06.md
xinan1911 Jun 19, 2024
1adfa66
Merge pull request #214 from EESSI/main
xinan1911 Jun 20, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
16 changes: 15 additions & 1 deletion docs/known_issues/eessi-2023.06.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -7,7 +7,7 @@

<p>This is an error that occurs with OpenMPI after updating to OFED 23.10.</p>

<p>Their is an upstream issue on this problem opened with EasyBuild.
<p>There is an upstream issue on this problem opened with EasyBuild.
See: https://github.com/easybuilders/easybuild-easyconfigs/issues/20233</p>

<b>Workarounds</b>
Expand All @@ -26,3 +26,17 @@ export OMPI_MCA_pml='ucx'
export OMPI_MCA_mtl='^ofi'
```
</div>

### `Bug in EESSI initialization and priority mechanisms: site OpenMPI or UCX not loaded`
Copy link
Collaborator

@casparvl casparvl Jun 4, 2024 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I don't think these are two seperate issues, right? I mean #456 was about site-specific tuning in general, but it only became an 'issue' when people hit the Failed-to-modify error. Shouldn't we just add to the "Failed to modify UD QP to INIT... etc" that the OMPI_MCA_* environment variables can be done in host_injections/...?

I was browsing the docs, I think we should have some general explaination of host_injections under Advanced usage. I.e. that this directory is used for site specific tuning. And then, in a paragraph there, we can add how it can be used to execute LMOD hooks for site specific tuning. We can give an example there of a hook (might as well use the example of setting these three OpenMPI environment variables). It's a good point, I wanted to make this documentation, but didn't get to it before my leave. Writing it as a todo for this afternoon... We can then refer to that from the "Failed to modify UD QP to INIT" known issue and give the code for the LMOD hook that can be used as a workaround for this particular issue.

Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Docs PR for Lmod hooks made here. I'd simply put an example hook to resolve this issue under the existing Failed to modify UD QP to INIT on mlx5_0: Operation not permitted header, and reference the new docs on Lmod hooks for further information.

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@casparvl Adjusted on the known issues pages. Could you review it and if everything is okay, merge it?

<div style="padding-left: 30px;">

<p>This error may occur when bugs resolving or site-specific tuning is needed for OpenMPI or UCX.</p>

<p>There is an issue on this problem opened with EESSI software layer repository.
See: https://github.com/EESSI/software-layer/issues/456</p>

<b>Workarounds</b>

<p>The workaround is to specify site properties and allow defining lmod hooks in host injections (see https://github.com/EESSI/software-layer/pull/525).
</div>

Loading