After the recent move to bookworm, we're hitting lots of errors related to /tmp filling up

[rdbisme]

I'm not 100% sure this is related to the image update, but do you know if there's something that has changed in bookworm that could lead to TMPDIR filling up much more than in bullseye when pip installing?

(I'm using pipx):

           INFO     Some possibly relevant errors from pip install:                                                                                poller.py:111
           INFO         ERROR: Could not install packages due to an OSError: [Errno 28] No space left on device    

[tianon]

That's weird; is this the only unusual behavior you're seeing?

(My best guess is something misinterpreting a seccomp denial as "no space left" but that seems odd -- however, you'd probably also be seeing issues with apt-get update in that case 🤔)

[yosifkit]

If it is seccomp related, then see #837 (comment).

[yosifkit]

If it is seccomp related, then see #837 (comment).

Can you reproduce the failure if you run the container with --security-opt seccomp=unconfined? If it does not reproduce, then it is libseccomp related and the two solutions are to either update Docker & libseccomp & runc or to use the -bullseye images.

[rdbisme]

@yosifkit Thanks for your reply. I'm having the problems on Gitlab runners, so no much control on how docker runs stuff. But I could try to update it and see if still reproduces.

[tianon]

Unfortunately, without a solid/minimal reproducer there's not much more we can do here (so I'm going to close). 🙇

[rdbisme]

I can confirm that updating libseccomp on our machines fixed the problem :)