We’re having an issue with a segmentation fault during one of our actions.
We’re testing nextflow pipelines on minimal datasets to ensure they are still functional with new code, versions etc. All of the pipelines are running within docker containers.
As the title suggests, completely non-deterministically, some of these pipeline runs are running into a segmentation fault at exactly the same step. Other pipeline runs which run the same step do not segfault, and furthermore, if we re-run the failed tests enough, they will eventually pass without a segfault. We have never seen this segfault outside of the github actions runs.
A couple of examples:
More specifically, this particular part of the pipeline is running a python process from the package Scanpy.
At first we considered that we might be falling onto a VM with older hardware and a different instruction set, but this doesn’t seem to be the case.
Any thoughts, or help here would be appreciated.