Job Dependencies#
This pages describes the SLURM depencency feature.
This feature is used when you need to chain jobs, due to dependencies. For example:
-
A preprocessing job with 1 core should be followed by a simulation with 16 cores. The results should be post-processed with a single main job.
-
A post-processing job must be sent after all tasks in a job matrix are completed.
Usage#
The dependent job must be launched using the sbatch command by specifying the --dependency=<type>:<listOfJobIDs> option. The type can be:
after: the job is launched after the first job enters into execution, i.e. it is RUNNING.afterok: the job is launched after the first job has been successfully completed.afterany: the job is launched after the first job has finished, regardless of whether it has failed or not.afternotok: the job is launched after the first job has finished on failure.aftercorr: a job in an array of jobs can start executing after the ID of another job in the array has been successfully completed.singleton: the job can start running when any other job with the same name and from the same user has finished.
Example:
The underlying job (on which this job depends) must be submitted first. The related job ID can be captured by collecting the output of the sbatch command with the --parsable option: