When using shifter as the container platform, pulling a container for the first time can take much longer than the requested time through SLURM. The check_max + task.attempt code in nextflow.config might help with this in some cases but not always.
Not sure of a good solution. One idea could be to pull all required containers at the start of the pipeline in a dedicated process, but while this might work for shifter, it might not work for other container platforms. It would also require a list of containers to be kept, which isn't a bad idea for software versions etc.
When using
shifter
as the container platform, pulling a container for the first time can take much longer than the requested time through SLURM. Thecheck_max
+task.attempt
code innextflow.config
might help with this in some cases but not always.Not sure of a good solution. One idea could be to pull all required containers at the start of the pipeline in a dedicated process, but while this might work for
shifter
, it might not work for other container platforms. It would also require a list of containers to be kept, which isn't a bad idea for software versions etc.