Need to run a job on the RCS compute service that could take several days? Or looking to execute a workflow consisting of many jobs, each taking less than 72 hours?
Each RCS account is able to submit one job to the long queue, which can take up to 1000 hours to complete rather than the usual limit of 3 days.
As an example, this could be useful for running a Nextflow pipeline where you know that each step will execute within the limits for the regular queues, but the coordinating job could take much longer. The relevant job file would look something like this:
#PBS -l select=1:ncpus=1:mem=1g
#PBS -l walltime=96:00:00
module load nextflow
cd $PBS_O_WORKDIR
nextflow run hello
You should always ensure that such long-running jobs are resumable.
Other options for long running jobs include self-service job extension or Express Access.
Further resources
- Job sizing guidance on the RCS web pages