Could somebody with access to Seqera Platform plea...
# configs
m
Could somebody with access to Seqera Platform please double-check something for me? It seems that there is a regression regarding the institutional configs in the newer (3.x.x) pipeline templates. I can no longer reference those profiles in Seqera Platform GUI or specify them on the
tw
CLI (
tw launch --profile
). I think the problem is, that the centralized configs are now added with a conditional
includeConfig
that Seqera platform can no longer correctly resolve
includeConfig !System.getenv('NXF_OFFLINE') && params.custom_config_base
. Any test helps, but particularly those people with access to Platform and using local offline pipelines on a HPC via an agent would be relevant here. 🙏
😱 1
j
To be clear you mean someone who uses platform to excute a pipeline on their local HPC, not just any execution ?
p
I'm not sure I understand what the problem is either.
j
I think it's like you can't pick e.g. uppmax from seqera platform
m
Exactly. For older pipelines, I have all the institutional profiles available for selection, but for newer ones only those that are directly declared in the
nextflow.config
inside the pipeline repository. So I can't launch a pipeline with the
uppmax
profile for example, because it will tell me, that I selected an invalid profile.
👍 1
p
But then, what does the offline have to do with anything except for the bit of code.
m
Possibly nothing. But the whole agent communication is simply a black box for me. It clones the pipeline into the work directory and then presumably runs
nextflow config
on that, to determine the available profiles. But if you change the revision or commit in the GUI, that must be rerun and the same applies, when you manually enter additional config in the GUI. Regardless what exactly that communication looks like, the ultimate config is pieced together on Platform as an ephemeral single file, so it might not be able to evaluate
System.getenv()
correctly.