PBS executor not working for task.config.getCpus() == 1 if no default set cluster

[fbartusch]

Dear Nextflow team,

Due to licensing issues we had to go back from MOAB/Torque to plain Torque on our cluster.
Because of the migration we lost the default resource configuration for each queue, thus processes with cpus = { 1} in the configuration do not run anymore. The -l nodes=1:ppn=1 line is missing in the jobscript and our lovely scheduler refuses to run the job.

I think the problem could be solved if you don't assume that there is a default values for nodes and cores set for the queues in PBS, e.g. also consider the case task.config.getCpus() == 1 here

[bentsherman]

Both the PBS and MOAB executors have this condition, so I'm hesitant to remove it. If it's always been that way, does that mean your cluster is behaving in a non-standard way?

You can also use clusterOptions to set the cpus manually

[fbartusch]

Our cluster does not behave in a non-standard way, it was just not optimally configured. Due to this I found this "issue".

When a Nextflow task required only one CPU, the PBS executor did not add the appropriate -l nodes=1:ppn=1 to the job script. As a consequence, these tasks did not start on our cluster. We have now set a proper default in our configuration, and the tasks are running correctly.

My point is that you cannot assume this default is set everywhere. I don’t know every detail of the codebase, but I don’t see any issues with changing this line

if( task.config.getCpus() > 1 ) {

to

if( task.config.getCpus() >= 1 ) {

This ensures that -l nodes=1:ppn=1 is always included in the job script, preventing reliance on a scheduler default that may or may not be set.