gputtley / htcondor

Snakemake profile for HTCondor

HTCondor Snakemake profile

This profile configures Snakemake to submit jobs to a HTCondor cluster.

Prerequisites

The profile makes use of the HTCondor python bindings (and snakemake-executor-plugin-cluster-generic for snakemake > 8) which can be installed with

pip install --user htcondor snakemake-executor-plugin-cluster-generic 

or using Anaconda with

conda install -c conda-forge -c bioconda python-htcondor snakemake-executor-plugin-cluster-generic

Deploy profile

To deploy this profile run

mkdir -p ~/.config/snakemake
cd ~/.config/snakemake
cookiecutter https://github.com/Snakemake-Profiles/htcondor.git

You will be asked for the name of the profile and for a path where the HTCondor logs will be stored. The logs will be used to update the status of submitted jobs (as recommended in the documentation of the HTCondor Python bindings).

Then, you can run Snakemake with

snakemake --profile htcondor ...

so that jobs are submitted to the cluster. If Snakemake is killed and restarted afterwards, it will automatically resume still running jobs.

Tests

The tests are heavily inspired by the tests for the slurm snakemake profile. They can be run from the base directory by

pytest

Because the tests will try to submit jobs they need to be started from a HTCondor submit node. To run the tests from non-cluster machines or from github CI the HTCondor/mini docker container can be started by:

DOCKER_COMPOSE=tests/docker-compose.yaml ./tests/deploystack.sh

Migration to snakemake v8

If using snakemake version 8 or higher, refer to the migration guide. For this profile, use config.v8+.yaml instead of config.yaml