Orientation Diagnostics does not scale resources with threads (slurm)

Troubleshooting
1

Dear Devs,

One of our users encountered a problem with the Orientation Diagnostics job (version 4.5.1), ending with an OOM kill.

In a quick investigation, I observed that this job does not scale resources reserved (this is a cluster setup with slurm lanes) with the number of threads declared in options.

In fact, build_orientation_diagnostics.py has a hardcoded single CPU and 8GB of RAM, regardless of the number of threads (which defaults to 24, which is also quite high for default value, in my opinion). As this behavior doesn’t make much sense, I suspect that this is a bug (easy to fix bug :slight_smile: )

2

Hey @bsobol –

Thanks for reporting! This is indeed a bug. Noted and tracked :slight_smile:

Valentin