Closed Juke34 closed 4 years ago
What did this person run exactly?
Also when running on Uppmax, the general idea is to open a screen session and run the nextflow pipeline in it. Nextflow then submits slurm job scripts for each process itself and monitors it.
Do we have need for an Uppmax specific profile?
Was preprocessing pipeline. What I understood it was the Conda stuff that was slow (create the env)... It is true create a conda env on uppmax can often take forever.
Yes, if they're using conda then it's going to take ages. There is a singularity profile which they can use and should be faster.
You mean instead of -profile nbis,conda
using -profile nbis,singularity
Yes.
But also they should use a profile for uppmax. E.g.
profiles {
uppmax {
executor {
name = 'slurm'
}
process {
scratch = '$SNIC_TMP'
}
includeConfig "$baseDir/config/compute_resources.config"
env.PERL5LIB='${PERL5LIB:-""}:${HOME}/git/NBIS/GAAS/annotation/'
env.PATH='${PATH}:${HOME}/git/NBIS/GAAS/annotation/Tools/bin'
singularity.enabled = true
includeConfig "$baseDir/config/software_packages.config"
singularity.envWhitelist = 'SNIC_TMP'
}
}
(put this in your custom config)!
Then run like so:
module load bioinfo-tools Nextflow
NXF_VER=20.01.0 nextflow run -c <config> -profile uppmax <nextflow_script>
Oh, and modify these lines too:
env.PERL5LIB='${PERL5LIB:-""}:${HOME}/git/NBIS/GAAS/annotation/'
env.PATH='${PATH}:${HOME}/git/NBIS/GAAS/annotation/Tools/bin'
GAAS is now on conda we should update that. Can we save the uppmax profile among our other profiles too?
Yes.
I'm going to propose a minor folder restructure too. It seems silly to have the pipelines subfolder there. I'll merge the README's.
One user reported he was not able to run it on uppmax. It was taking forever. He ran directly the nextflow pipeline after
module load nextflow
. Could we provide an example of sbatch file to run it through Slurm in the readme?