Institut Pasteur Configuration
All nf-core pipelines have been successfully configured for use on the tars cluster at the Institut Pasteur.
To use, run the pipeline with -profile pasteur. This will download and launch the pasteur.config which has been pre-configured with a setup suitable for the Pasteur cluster. Using this profile, a docker image containing all of the required software will be downloaded, and converted to a Singularity image before execution of the pipeline.
Running the workflow on the Pasteur cluster
Nextflow is not installed by default on the Pasteur cluster.
- Install Nextflow : here
Nextflow manages each process as a separate job that is submitted to the cluster by using the sbatch command.
Nextflow shouldn’t run directly on the submission node but on a compute node.
The compute nodes don’t have access to internet so you need to run it offline.
To do that:
- Create a virtualenv to install nf-core
module purge
module load Python/3.6.0
module load java
module load singularity
cd /path/to/nf-core/workflows
virtualenv .venv -p python3
. .venv/bin/activate
- Install nf-core: here
- Get nf-core pipeline and container: here
- Get the nf-core Pasteur profile: here
- Run nextflow on a compute node:
# create a terminal
tmux
# Get a compute node
salloc
# Load the dependencies if not done before
module purge
module load java
module load singularity
# Run nextflow workflow
nextflow run \\
/path/to/pipeline-dir/from/step/3/workflow \\
-resume
-profile pasteur \\
-with-singularity /path/to/pipeline-dir/from/step/3/singularity-images/singularity.img \\
--email my-email@pasteur.fr \\
--custom_config_base /path/to/configs/from/step/4/ \\
-c my-specific.config
Config file
//Profile config names for nf-core/configs
params {
config_profile_description = 'The Institut Pasteur HPC cluster profile'
config_profile_contact = 'Remi Planel (@rplanel)'
config_profile_url = 'https://research.pasteur.fr/en/service/tars-cluster'
}
singularity {
enabled = true
autoMounts = true
runOptions = '-B /local/scratch:/tmp'
}
process {
resourceLimits = [
memory: 256.GB,
cpus: 28,
time: 24.h
]
executor = 'slurm'
}
params {
igenomes_ignore = true
igenomesIgnore = true
//deprecated
max_memory = 256.GB
max_cpus = 28
max_time = 24.h
}
executor
contact
