Dalia FAQ

⚠ INFORMATION
This page was translated by an AI (LLM) with a cursory human check and is awaiting full review.

✉️ For any technical questions related to the operation of Dalia, you can contact IDRIS support at assist@idris.fr by prefixing the subject of your email with "[Dalia]". For example:

To : assist@idris.fr
Subject : [Dalia] Question about Apptainer container usage
Content : Blablabla

Segfault on a job using more than 11 nodes

By default, the resources allocated to system processes are unlimited (visible with the command ulimit -s). But in reality, the value unlimited corresponds to an allocation of 2 MB of memory on the stack per thread, which is too little. The solution is to increase the allocated resources as follows:

ulimit -s 8192

Source: https://docs.nvidia.com/deeplearning/nccl/user-guide/docs/troubleshooting.html#stack-size.