Execution context

weave is capable of automatically distributing its pipeline jobs across a slurm cluster. The context for it's initial execution can be varied as well.

The context is also centrally related to the configuration and setup of a particular cluster. Right now weave is configured to work with NIH clusters skyline, biowulf, and bigsky.

Typical contexts of execution include:

srun (real time execution) (non-interactive)¶

The weave pipeline can be triggered from a head node in a non-interactive fashion:

Bigsky/Skyline¶

Note

Dependency files for skyline and bigsky differ
Bigsky: /gs1/RTS/OpenOmics/bin/dependencies.sh
Skyline: /data/openomics/bin/dependencies.sh

source ${dependencies}
srun --export=ALL "weave run [keyword args] ${run_id}"

Note

srun by default exports all environmental variables from the executing environment and --export=ALL can be left off

Biowulf¶

srun --export=ALL "module load snakemake singularity; weave run [keyword args] ${run_id}"

srun (real time execution) (interactive)¶

Bigsky/Skyline¶

Note

Dependency files for skyline and bigsky differ
Bigsky: /gs1/RTS/OpenOmics/bin/dependencies.sh
Skyline: /data/openomics/bin/dependencies.sh

> # <head node>
srun --pty bash
> # <compute node>
source ${dependencies}
weave run [keyword args] ${run_id}

Biowulf¶

> # <head node>
sinteractive
> # <compute node>
module purge
module load snakemake singularity
weave run [keyword args] ${run_id}

Biowulf uses environmental modules to control software. After executing the above you should see a message similar to:

[+] Loading snakemake 7.XX.X on cnXXXX
[+] Loading singularity 4.X.X on cnXXXX

sbatch (later time execution)¶

Bigsky/Skyline¶

sbatch tempalte¶

bigsky-skyline sbatch template

#!/bin/bash
#SBATCH --job-name=<job_name>
#SBATCH --export=ALL
#SBATCH --time=01-00:00:00
#SBATCH --cpus-per-task=1
#SBATCH --ntasks=1
#SBATCH --mem=8g
#SBATCH --output=<stdout_file>_%j.out
source ${dependencies}
weave run \
-s /sequencing/root/dir \
-o output_dir \
<run_id>

This above script can serve as a template to create an sbatch script for weave. Update the psuedo-variables in the script to suit your particular needs then execute using sbatch command:

sbatch weave_script.sbatch

Biowulf¶

sbatch tempalte¶

biowulf sbatch template

#!/bin/bash
#SBATCH --job-name=<job_name>
#SBATCH --export=ALL
#SBATCH --time=01-00:00:00
#SBATCH --cpus-per-task=1
#SBATCH --ntasks=1
#SBATCH --mem=8g
#SBATCH --output=<stdout_file>_%j.out
module purge
module load snakemake singularity
weave run \
-s /sequencing/root/dir \
-o output_dir \
<run_id>

Same sbatch execution as bigsky/skyline.

sbatch weave_script.sbatch