UPPMAX: Memory overflow error
parlundin opened this issue · 4 comments
parlundin commented
SLURM job killed by an out-of-memory condition.
Running SBATCH with -p node and -n 16
Seems to be a combination of bwa and Samtools that uses to much ram.
/pica/v9/a2014205_nobackup/perlu/ANALYSIS/ND-0522/.queue/tmp/.exec8811658286087080852: line 2: 6923 Broken pipe /sw/apps/bioinfo/bwa/0.7.5a/milou/bin/bwa mem -M -t 15 -R "@RG\tID:AC45JGANXX.NA11993_Nano.7\\tSM:NA11993_Nano\\tCN:NGI\\tLB:SX398_NA11993_Nano\\tPL:Illumina\\tPU:AC45JGANXX.NA11993_Nano.7" /proj/a2014205/piper_references/gatk_bundle/2.8/b37/human_g1k_v37.fasta /pica/v9/a2014205_nobackup/perlu/Subsampled2/DATA/ND-0522/NA11993_Nano/SX398_NA11993_Nano/140821_D00458_0029_AC45JGANXX/NA11993_Nano_TAGCTT_L007_R1_001.fastq.gz /pica/v9/a2014205_nobackup/perlu/Subsampled2/DATA/ND-0522/NA11993_Nano/SX398_NA11993_Nano/140821_D00458_0029_AC45JGANXX/NA11993_Nano_TAGCTT_L007_R2_001.fastq.gz
6924 | /sw/apps/bioinfo/samtools/0.1.19/milou/bin/samtools view -Su -
6925 Killed | /sw/apps/bioinfo/samtools/0.1.19/milou/bin/samtools sort -@ 15 -m 7G - /pica/v9/a2014205_nobackup/perlu/ANALYSIS/ND-0522/pipeline_output/01_raw_alignments/NA11993_Nano.AC45JGANXX.NA11993_Nano.7
mariogiov commented
@francesco didn't you encounter something like this once? I recall you manually setting the Java memory parameters somehow,
vezzi commented
I do not think the things are connected.... this seems the node to go out of memory my problem was running too many piper jobs on the login node
johandahlberg commented
I've been talking to @parlundin about this, and it's bwa + samtools that cause the problem. I'll fix it for the next release.
johandahlberg commented
I think that this should be ok now. So I'll close this.