ISISNeutronMuon
[BUG] Kill MDANSE job when 'No space left on device' errors occur.
Description of the error When running an MDANSE job with no disk space left, MDANSE jobs continue to run. Errors are seen in the console.
Suggested fix When the user runs out of disk space the MDANSE jobs should fail immediately with a disk space error.
Looks like this is a particularly strange behaviour with h5py. It doesn't seem to raise an exception when it fails to write to disk when there is no longer any space. This happens when writing to a dataset that has the chunks property set. When I remove the chunks setting from this part of the code
https://github.com/ISISNeutronMuon/MDANSE/blob/26487f13c2ca0c1308041081e670067d943c1fbe/MDANSE/Src/MDANSE/MolecularDynamics/Trajectory.py#L749-L820
which is used for trajectory conversion, and if I run trajectory conversion on a disk with insufficient space, then h5py will raise an error and cause the job to fail.
This is kinda strange, I tried to create a minimal reproducible example, but h5py seems to raise the OSError when the chunking is or is not set. The error seems to be very specific to something going on in MDANSE.