Hello, I am running nf-core/rnaseq on SLURM, and my pipeline keeps breaking at SORTMERNA. 25/35 human samples have been processed without any problems, but the rest of the 10 samples keep causing:

'NFCORE_RNASEQ:RNASEQ:FASTQ_QC_TRIM_FILTER_SETSTRANDEDNESS:SORTMERNA' terminated with an error exit status (137) -- Execution is retried (2)

and OOM happens with 144GB and 24CPU. Should I increase the memory for the process to e.x. 240GB or lower memory pressure to avoid crashes - e.x. 8 CPU and 64GB (jobs will then run longer but might complete without crashes)? I used very simple config before, as I've processed 5 samples max with the pipeline, and didn't have these problems before:

process {

resourceLimits = [

memory: 256.GB,

cpus: 32,

time: 48.h

]

executor       = 'slurm'

queueSize     = 20

}

singularity {

enabled = true

autoMounts = true

}

I would highly appreciate any help and advice. Thanks!

Generally, you'd need to request more memory - for the process, since you're not yet reaching your maximum from

resourceLimits

. There are some tools that need significant extra amounts of memory per thread, but those are fairly rare. (But most bioinformatics tool scale rather bad so more cores generally don't do much good anyway, don't recall about sormerna though.)

Okay, thank you for the reply!

Docs: https://nf-co.re/docs/running/configuration/nextflow-for-your-system#customize-process-resources

sortmerna is a bit of a beast like this, I think some samples with different composition can make the memory spike. You might have more luck with one of the other options for rrna removal- folks sometimes have good results with good old bowtie2 https://nf-co.re/rnaseq/3.25.0/parameters#ribo_removal_tool

Thanks!