I've demultiplexed a NovaSeq X run today, using the 1.7.0 release. On Seqera Platform, using fusion. Running bclconvert went fine, but then the pipeline just sat around for 4 hours without starting anything further. Stopping and resuming didn't help. What did help, was modifying the bcl_demultiplex subworkflow, to remove the section where it tries to read each of the fastq files to extract the read group information. Do we really need this? I don't think we use the readgroup information at all within the pipeline

We really rely on that info in our pipelines. We’ve been seeing the same issue since our move to gcp from local cluster. I’m guessing it has to do with file transfer from the bucket to the head node, since the function is in groovy. This is especially an issue with large flowcells.

But there is no reason to do that in the demultiplex pipeline, because that pipeline doesn't use it in any way

I do agree with that part. The fastq parsing is part of the

bcl_demultiplex

subworkflow though, so I would suggest a local patch of the subworkflow for the time being