Snakemake Profile for Running on AWS ParallelCluster with the Slurm Scheduler

This is a Snakemake profile for running on AWS ParallelCluster with the Slurm scheduler. It is heavily based on John Blischak's smk-simple-slurm profile and as such this repository uses the same license.

Default ParallelCluster Slurm installations have two peculiarities that make them differ from most other Slurm installations:

• The --mem argument to sbatch is not supported, and if used sends nodes into the DRAINED state.
• The sacct accounting system is not installed by default.

This profile accomodates these peculiarities.

Usage

Obtain the profile

Clone this repository to a convenient location on the head node

git clone https://github.com/SequenceBio/sb-slurm.git

Edit the profile

The profile in this repository should be sufficient for many basic use cases at SB. If needed, you can edit this profile by changing parameters in config.yaml. For example, you may want to add --rerun-incomplete to the snakemake command, or you may want to adjust the default partition.

Set up a wrapper script

Copy the script wrapper.sh from this repository into your pipeline directory. Be sure that the path after --profile points back to this repository.

Run your pipeline

Navigate into your pipeline directory. Submit the wrapper script to the job scheduler

sbatch wrapper.sh

Evaluate your pipeline run

Monitor your jobs with squeue. To continually refresh squeue at 10 second intervals, run watch -n10 squeue

Upon pipeline completion, check the following to examine your run:

• Look in results/ to see the files produced by the pipeline run
• Look in log_<datestamp>.out to see the Snakemake stdout - the final few lines should include an indication of whether the pipeline finished successfully or not
• Look in the logs/ directory, which contains the stdout and stderr for every rule that was run. This is helpful for troubleshooting if a pipeline did not complete successfully.