Are queue arguement in profiles, additive or do th...
# configs
r
Are queue arguement in profiles, additive or do they overwrite each other? (I'm assuming the overwrite each other Example
Copy code
profiles {
   queue1 {
     process {
       executor = 'slurm'
       queue = 'queue1' 
     }
   }
   queue2 {
     process {
       executor = 'slurm'
       queue = 'queue2' 
     }
   }
}
Copy code
-profile queue1,queue2
What is passed to slurm here:
--partition queue1,queue2
or
--parition queue2
This also might be a better question for Nextflow slack, let me know if I should ask there
p
I don't recall the rules, but am pretty certain it's either
queue1
or
queue2
- no merging there. It's actually a possible problem that it can replace other parts if you use the
{ ... }
format - whereas e.g.
process.executor = 'slurm'
will not have that problem.
👍 1
r
For submitting to nf-core/configs on a HPC system that has 10+ partitions would it be preferred to have a profile for each partition? or have users set a env variable when multiple partitions are being used?
p
Depends on the situation (how partitions are used). E.g. for dardel we decide on one or a few partitions (queue is not used, but that doesn't really matter, could be a separate closure for that).
Similarly, if different partitions are for different groups, I think that ideally should be managed automatically.
But if not possible; not environment variables - either different profiles, but probably rather pick up from
params
.
👍 1
e
Ooo I actually figured something out like this
Two things you might find useful. 1. Checking the groups a user is in and determining queues off that. 2. Multiple queues per job with some fancy stuff I didn't know you could do with slurm.
Idk how your queues are configured but if it's just a condo system of who owns what you could do:
Copy code
def membership = "groups".execute().text.join(',')
process.queue = membership