Sample Header Ad - 728x90

Make a job depend on another but only if timeout

0 votes
0 answers
132 views
When submitting jobs with qsub , we can make sure that a job only starts after another. More so, we can only execute it depending on the status of that other job. Perhaps only run the new job if the other one fails, or only if exits OK. But in my case, I want to start a job when it has "failed" with a status of CANCELLED,TIMEOUT. The use case is that I sometimes have long training runs and our sysadmin only allows jobs of 32 hours. If a job "times out", that means that the training run was not finished, so a new job should take up where the timed out job left. The usual syntax is as follows.
qsub myjob.pbs -W depend=afterok:
Or afterany, afternotok, etc. Is there a way to make this work for specific statuses, so in my case the TIMEOUT status of an ended job?
Asked by Bram Vanroy (183 rep)
Jul 25, 2023, 12:05 PM
Last activity: Jul 26, 2023, 03:22 PM