[slurm-users] 21.08: Removing batch scripts from the database

Kevin Buckley Kevin.Buckley at pawsey.org.au
Tue Oct 5 06:45:11 UTC 2021


Trying to get my head around the extremely useful addition, for 21.08
onwards, as regards storing the batch scripts in the accounting database,
so that the script can be obtained in cases where say, the user submitting
the job may no longer have access to them, and wanted to ask:

Is there anything preventing the Slurm administrator clearing out batch
scripts from the database ?

I can't see that there's anything obvious in the code that flags whether
a Job has had its batch script stored or not - it's either there or it
isn't - so expect a "Nothing stopping you from doing that, but take care"
response, but thre may be something that's not obvious to me?

I'm asking so that the database doesn't run the risk of getting full up
with batch scripts that might, for example, contain the commented out
development history of the script since first written.


I suppose, along similar lines, if one could identify that the batch script
fields containing identical data, is there anything stopping the administrator
form binning all bar one script and just putting the JobID of that one into
the batch_script field of the duplicates? Modulo that against job visibility
concerns where all the users in a research group are running identical scripts
provided by their PI.

In the way I think of things (well, at the moment anyway) nothing makes use
of the batch_script field in the <clustername>_job_table table, other than
the client-side "show me the batch script for this JobID" options but maybe
there are plans to have some kind of

  sbatch/srun --rerun_the_script_from_jobid <job_id>

functionality in the future, in which case pulling the batch script rug
out from underneath the users might not be a good thing.

Any thoughts/insights welcome,
Kevin Buckley
-- 
Supercomputing Systems Administrator
Pawsey Supercomputing Centre



More information about the slurm-users mailing list