0

I used to have the worker logs as such:

./slurm-<id>.out
...

So I wanted to have SLURMCluster writes the worker logs in a separate directory (as opposed to the current working dir), so I provided "log_directory" as an input argument as such.

from dask_jobqueue import SLURMCluster
c = SLURMCluster(queue_name, log_directory="dirA")
c.scale_up(1)

And the logs end up like this:

./dirA/dask-worker-<id>.out
./dirA/dask-worker-<id>.err

And all the worker logs are piped to the .err file while the .out file is empty.

What am I missing in the SLURMCluster configuration to set it correctly, such that the .out should contain the logs and the .err should contain worker errors.

Rob
  • 14,746
  • 28
  • 47
  • 65
michaelgbj
  • 290
  • 1
  • 10
  • the title no longer makes sense without any context to what log and log_directory is, when the reference to SLURMCluster is removed – michaelgbj Apr 03 '23 at 19:19
  • Could you check the job submission script created by dask-jobqueue (print(cluster.job_script())), normally, setting log_directory will add two options to the batch submission script: #SBATCH -e log_directory/worker-id.err and #SBATCH -o log_directory/worker-id.err. In the previous setup, did you have the logs in the .out file? – Guillaume EB Apr 07 '23 at 05:54

0 Answers0