Job Dependencies¶

It is possible to schedule jobs to run based on the status of other jobs. For example, you might schedule a preprocessing job to run; start a computation job when the preprocessing job is complete; then start a post-processing job when the computation is done.

One way to schedule such a series or chain of jobs is to use qsub -W [job-dependency-expression] to specify the job dependencies you need. This can also be accomplished by submitting subsequent jobs from inside another job.

PBS job dependencies (-W depend)¶

Let's say you have you have three scripts to submit and run consecutively:

1. pre.pbs: a preprocessing job
2. main.pbs: a computation job
3. post.pbs: a post-processing job

The main job can be run only when the preprocessing job finishes, and the post-processing job can be run only when the computation job finishes.

Submit the first job, placing it on hold. (If it starts before the dependent jobs are submitted, the dependent jobs might never start.):

JOBID1=$(qsub -h pre.pbs)

JOBID2=$(qsub -W depend=afterok:$JOBID1 main.pbs)

JOBID3=$(qsub -W depend=afterok:$JOBID2 post.pbs)

qrls $JOBID1

The complete sequence is shown below for additional clarity:

JOBID1=$(qsub -h pre.pbs)
JOBID2=$(qsub -W depend=afterok:$JOBID1 main.pbs)
JOBID3=$(qsub -W depend=afterok:$JOBID2 post.pbs)
qrls $JOBID1

set JOBID1=`qsub -h pre.pbs`
set JOBID2=`qsub -W depend=afterok:$JOBID1 main.pbs`
set JOBID3=`qsub -W depend=afterok:$JOBID2 post.pbs`
qrls $JOBID1

In the example above, the clause afterok is used to indicate the subsequent jobs should only begin after the preceding job completes successfully. This is likely the most common use case, however other clauses are available, and jobs may depend on multiple predecessors. Some of the more common dependency clauses are listed below, where <arg_list> is usually a single PBS job id as shown above, but can be a colon-separated list of IDs if appropriate. See man qsub for a full listing and additional details.