Known Issues

A list of currently known issues with workarounds, if availble.

If an issue is resolved, please disregard the workaround.

Out-Of-Memory error messages even though job completed successfully

There is what appears to be a Slurm issue where it reports an "Out Of Memory" error, despite a job completing successfully.

Examine the job's output to see if it completed as expected. It may be useful to put an "echo" statement at the end of the job script to indicate that it has completed:

echo "Job done."

If the output of the job is as expected, the Out Of Memory error message can be safely ignored.

There is currently (July 3, 2021) no fix for this issue. An update to Slurm is planned for next month, which may resolve this issue.

RESOLVED - TMP directory not created with srun (interactive job)

When an interactive job is run, e.g.

[juser@picotte001 ~]$ srun -p def --mem=1000M -n 1 -t 600 --pty /bin/bash

the TMP local scratch directory is not created:

[juser@node042 ~]$ ls -l $TMP /bin/ls: cannot access '/local/scratch/80301': No such file or directory

This can cause interactive sessions to behave oddly, and can cause GUI applications to fail.

Workaround

Manually create the directory:

[juser@node042 ~]$ mkdir $TMP