Brian Huynh
2014-07-17 14:47:31 UTC
Hi,
I have two jobs encapsulated in a single bash script. I would like the second to be able to queue but not start until the first has completed. As of now, I can only get SLURM to implement this if I have the exact job ID but I will not be able to readily guess this ID once SLURM is active with normal usage.
Previously, in SGE, I used the tag -N previousjob_$1 in the qsub line for the first job and -hold_jid previousjob_$1 in the qsub line for the second job to create the job dependency.
However, SLURM does not allow for me to pass in arguments into the tag --depend=afterok:jobID while jobID contains an input argument such as $1. How should I best implement this?
Both job submissions work independently of each other without the dependency included.
Thanks for all the help and advice.
Best,
Brian
I have two jobs encapsulated in a single bash script. I would like the second to be able to queue but not start until the first has completed. As of now, I can only get SLURM to implement this if I have the exact job ID but I will not be able to readily guess this ID once SLURM is active with normal usage.
Previously, in SGE, I used the tag -N previousjob_$1 in the qsub line for the first job and -hold_jid previousjob_$1 in the qsub line for the second job to create the job dependency.
However, SLURM does not allow for me to pass in arguments into the tag --depend=afterok:jobID while jobID contains an input argument such as $1. How should I best implement this?
Both job submissions work independently of each other without the dependency included.
Thanks for all the help and advice.
Best,
Brian