Closed StevenWhitaker closed 8 months ago
In case this info helps, sometimes the jl_set_task_tid == 0
error occurs on a different worker (e.g., worker 4 instead of worker 2).
Here is more info about the other machine:
NAME="Red Hat Enterprise Linux Server"
VERSION="7.9 (Maipo)"
ID="rhel"
ID_LIKE="fedora"
My laptop runs Ubuntu 22.04 on WSL 2.
Seems like it might be that I'm not understanding how that function should be used - I'm gathering details on Slack and will try to push a fix today.
@jpsamaroo Any updates?
https://github.com/JuliaParallel/Dagger.jl/pull/462 should resolve this, potentially with a small performance impact if we keep failing to set the task TID (since it will give up and just the task run wherever Julia's scheduler decides to place it).
This is great news! I'll try it out and let you know if the issue is resolved.
@jpsamaroo it appears your PR resolved the issue I was seeing on the other Linux machine. Thanks!
I ran across an error when trying to run some code on a different machine. I don't know if this error has occurred on my laptop; if it has, it is not (yet) reproducible and only occurs rarely, whereas on the other machine the error happens every time. My laptop runs Windows 11 (WSL 2); the other machine runs some form of Linux.
Here is the (reformatted) error:
I am in the process of trying to more fully mimic the setup of the other machine (different Julia patch version, more Julia processes, different
--heap-size-hint
) to see if I can reproduce the error.Do you have any idea what would cause the
jl_set_task_tid == 0
error?