1
0
Fork 0
mirror of https://github.com/MillironX/nf-configs.git synced 2024-11-23 00:39:56 +00:00
nf-configs/docs/mpcdf.md
James A. Fellows Yates 7088ed196f
Update mpcdf.md
2020-10-22 11:08:29 +02:00

1.6 KiB

nf-core/configs: MPCDF Configuration

All nf-core pipelines have been successfully configured for use on the HPCs at Max Planck Computing and Data Facility.

⚠️ these profiles are not officially supported by the MPCDF.

To run Nextflow, the jdk module must be loaded. To use the nf-core profile(s), run the pipeline with -profile <cluster>,mpcdf.

Currently the following clusters are supported: cobra, raven

NB: Nextflow will need to submit the jobs via SLURM to the clusters and as such the commands above will have to be executed on one of the head nodes. Check the MPCDF documentation.

cobra

Cobra does not currently support singularity, therefore the anaconda/module is loaded for each process.

Due to this, we also recommend setting the $NXF_CONDA_CACHEDIR to a location of your choice to store all environments (so to prevent nextflow building the environment on every run).

To use: -profile cobra,mpcdf

Sets the following parameters:

  • Maximum parallel running jobs: 8
  • Max. memory: 750.GB
  • Max. CPUs: 80
  • Max. walltime: 24.h

draco

🛠️ under testing.

raven

Raven does not currently support singularity, therefore module load anaconda/3/2020.02 is loaded for each process.

Due to this, we also recommend setting the $NXF_CONDA_CACHEDIR to a location of your choice to store all environments (so to prevent nextflow building the environment on every run).

To use: -profile raven,mpcdf

Sets the following parameters:

  • Maximum parallel running jobs: 8
  • Max. memory: 368.GB
  • Max. CPUs: 192
  • Max. walltime: 24.h