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.

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

WORK inodes limitations / troubleshooting / work-arounds #69

Open
lesteve opened this issue Jun 3, 2021 · 7 comments
Open

WORK inodes limitations / troubleshooting / work-arounds #69

lesteve opened this issue Jun 3, 2021 · 7 comments

Comments

@lesteve
Copy link
Member

lesteve commented Jun 3, 2021

From https://gitter.im/jean-zay-users/jean-zay-doc?at=60ae90153d43a15a65c9fad0

High-level overview (quick to run <1s, but the info may be stale, it is refreshed periodically IIRC). It also tells you your inodes quota:

idrquota -w

Exact method to count the inodes in some folder but slower (a few seconds), -d1 is to get the info about one-level deep folders:

du -d1 --inodes $WORK

In practice, the main limitation happens if you have a miniconda install in the WORK. If that's the case conda clean -tipy may remove some unused files and save you some inodes.

Feel free to comment if you have additional tips and tricks!

You can ask to increase your inodes quota, but it seems like there is a special procedure that I don't know about ... maybe @guhur you can add some info here if you manage to do it 😉

@lesteve
Copy link
Member Author

lesteve commented Jul 6, 2021

For the conda environment use case, I am thinking maybe there is a convenient way to use conda-pack to create .tar.gz or even .squashfs that use a single inode and then you can untar this in the SCRATCH.

This is not as convenient as having a conda environment always accessible but maybe an OK work-around?

With the squashfs option there maybe an easy way to use it without unpacking, I need to ask user support about this, I think at one point they said it is supported on Jean-Zay.

@RemiLacroix-IDRIS
Copy link
Contributor

RemiLacroix-IDRIS commented Jul 6, 2021

With the squashfs option there maybe an easy way to use it without unpacking

It should be possible but as far as I can tell conda-pack does not support SquashFS natively.

@lesteve
Copy link
Member Author

lesteve commented Jul 6, 2021

It does actually but for some reason their doc website is not up-to-date ... look at one of the last commit 😉

@lesteve
Copy link
Member Author

lesteve commented Jul 6, 2021

There you go: https://github.com/conda/conda-pack/pull/177/files

@RemiLacroix-IDRIS by the way, since you are here, it would be great if you can post the way to mount a squashfs image on Jean-Zay (if I remember correctly you can even do this without singularity).

@RemiLacroix-IDRIS
Copy link
Contributor

It does actually but for some reason their doc website is not up-to-date ... look at one of the last commit wink

This is indeed really new, it's probably not in an official release yet hence not in the doc on the website.

it would be great if you can post the way to mount a squashfs image on Jean-Zay

This feature is still in beta stage but I will let you know as soon as we release it.

@lesteve
Copy link
Member Author

lesteve commented Jul 6, 2021

This is indeed really new, it's probably not in an official release yet hence not in the doc on the website.

In Python open-source (especially true for small projects) this is often the case that doc is built from the master branch but not in this case indeed, they deploy the doc only for tags: https://github.com/conda/conda-pack/blob/0bdd462691e8dc31edb0dcdb86a332cc6ddaeda9/.github/workflows/main.yml#L155-L160

@lesteve
Copy link
Member Author

lesteve commented Jul 6, 2021

it would be great if you can post the way to mount a squashfs image on Jean-Zay

This feature is still in beta stage but I will let you know as soon as we release it.

Ah OK I thought it was already possible, let me know if I can help for beta-testing 😉

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants