Hello,
We are using OOD with Slurm. A job status shows as “CANCELLED” in Slurm when a user terminates a session. Is there a way to treat it as “COMPLETED” ?
We have a billing system but it doesn’t charge SUs for “CANCELLED” jobs.
Thank you
Hello,
We are using OOD with Slurm. A job status shows as “CANCELLED” in Slurm when a user terminates a session. Is there a way to treat it as “COMPLETED” ?
We have a billing system but it doesn’t charge SUs for “CANCELLED” jobs.
Thank you
Thanks for the question!
The place this gets set in the code is in the ood_core
repo and the states are explicitly set at the line below:
But oddly that does show the cancelled code as showing a completed value (using the ruby symbol :completed
). So, I guess I’d next ask where exactly does the billing system look for that string? Is it using some log output or are you making a call to ood somehow?
Hello
The billing system is getting the state info from Slurm Database. I see the status as “CANCELLED” in Slurm using “sacct” command.
-bash-4.2$sacct -u gp4r -j 40722795
JobID JobName Partition Account AllocCPUS State ExitCode
------------ ---------- ---------- ---------- ---------- ---------- --------
40722795 ood_jupyt+ standard hpc_build 1 CANCELLED+ 0:0
40722795.ba+ batch hpc_build 1 CANCELLED 0:15
I assume OOD is using Slurm scancel
command to terminate a session and it set the state as “CANCELLED” in Slurm.
Thank you,
This topic was automatically closed 180 days after the last reply. New replies are no longer allowed.