nig-uge

Snakemake profile for running jobs on the NIG Supercomputer.
Derived from
jaicher/snakemake-qsub
and
broadinstitute/snakemake-broad-uger.

Requires Cookiecutter to use.

Installation/Setup

To create a new profile, first make sure that the configuration directory that Snakemake searches for profiles exists by running:

mkdir -p ~/.config/snakemake

then create the profile using Cookiecutter:

cookiecutter --output-dir ~/.config/snakemake gh:shjenkins94/nig-uge

After giving the profile a name, the following Snakemake parameters can be set:

• directory
• conda-prefix
• restart-times
• use-conda
• use-singularity
• keep-going
• printshellcmds
• jobs
• latency-wait

Along with parameters for cluster execution:

• default_mem_gb: default amount of memory reserved by each job.
• default_queue: which queue to use if no resources are specified.
• cpu_hung_min_time: minimum amount of time before a job can be checked for hanging.
• cpu_hung_max_ratio: the maximum ratio between cpu time and wallclock time for a job to be considered hanging.
• missing_job_wait: the amount of time the cluster should wait before checking on a missing job.
• default_cluster_logdir: directory to store stdout and stderr output from cluster.
• cluster_dir: directory that stores files tracking job statuses during cluster jobs.

Usage

After creating a profile, you can run workflows with the parameters you specified using Snakemake's profile flag like so:

snakemake --profile PROFILE-NAME {...}

Rule Settings

Individual Snakemake rules can have the following parameters specified in the Snakemake file:

• threads: the number of threads needed for the job. If not specified, assumed to be 1.
• resources
  • mem_gb: the memory required for the rule in gigabytes, which will be requested if present
  • mpi: If using mpi, set to the number of mpi processes to run
  • use_java: set to True for rules that use java. Prevents java from using too much memory.