Problem with generating STAR genome index using cellranger

Question

[Zqall]

Hi,

I'm trying to run cellranger mkref using this tutorial (https://support.10xgenomics.com/single-cell-gene-expression/software/pipelines/latest/using/tutorial_mr#getfiles) I reached to this step

#Run mkref
cellranger mkref \
  --genome=mRatBN7 \
  --fasta=Rattus_norvegicus.mRatBN7.2.dna.toplevel.fa \
  --genes=Rattus_norvegicus.mRatBN7.2.105.filtered.gtf \
  --ref-version=1.0.0

And I got this error message

Generating STAR genome index (may take over 8 core hours for a 3Gb genome)...
Jun 30 14:08:04 ..... started STAR run
Jun 30 14:08:04 ... starting to generate Genome files
Jun 30 14:09:23 ... starting to sort Suffix Array. This may take a long time...
Jun 30 14:09:29 ... sorting Suffix Array chunks and saving them to disk...
Failed to make genome index with STAR.  This can occasionally be caused by setting the argument `memgb` too low.
Error was from running command '/out/Analysis/CellRanger/cellranger-7.0.0/lib/bin/STAR'
Command '['/out/Analysis/CellRanger/cellranger-7.0.0/lib/bin/STAR', '--runMode', 'genomeGenerate', '--genomeDir', '/out/Analysis/CellRanger/Mmul_10/star', '--runThreadN', '1', '--genomeFastaFiles', '/out/Analysis/CellRanger/Mmul_10/fasta/genome.fa', '--sjdbGTFfile', '/out/Analysis/CellRanger/Mmul_10/genes/genes.gtf', '--limitGenomeGenerateRAM', '17179869184', '--genomeSAsparseD', '3', '--genomeSAindexNbases', '14', '--genomeChrBinNbits', '18']' died with <Signals.SIGKILL: 9>.

how to solve this?

[GenoMax]

STAR index generation can require a lot of memory. In this case it can be 30-40GB. Do you have that much free memory available?

[Zqall]

Yes I have

[GenoMax]

Are you running this job directly on the server command line or are you submitting to a job scheduler? If latter, make sure you ask for additional RAM allocation.

[Zqall]

I'm running it directly on the server command line

[Ram]

Don't do that. HPC are usually designed so login nodes don't support high memory/long run-time tasks. Use a compute node using a job scheduler instead of running it directly on the server command line.