Slurm memory overflow on variant effect prediction

[dnil]

Running a regular wgs bam with --steps manta,vep

nextflow run main.nf --project b2014152 --bam /proj/b2014152/private/delivery/P2109/P2109_150/03-BAM/P2109_150.clean.dedup.recal.bam --steps manta,vep

Works up until: [1c/4f4270] Submitted process > variant_effect_predictor (4df25178-37e0-40ea-a02c-114782119240 - manta_masked.vcf)

Then got an email back with: UPPMAX: Memory overflow error on host m116 for job 9775456 (/nf-variant_effect_predictor_(4df25178-37e0-40ea-a)

At about 8 Gb;

the offending line happens to be

  gzip: /sw/data/uppnex/vep/84/homo_sapiens_merged/84_GRCh37/16/11000001-12000000.gz: Cannot allocate memory
  .command.sh: line 15:  9836 Killed                  variant_effect_predictor.pl -i "$INFILE" --format "$FORMAT" -cache --dir "$VEP_CACHE" -o "$OUTFILE" --vcf --merged --regulatory --force_overwrite --sift b --polyphen b --symbol --numbers --biotype --total_length --canonical --ccds --fields Consequence,Codons,Amino_acids,Gene,SYMBOL,Feature,EXON,PolyPhen,SIFT,Protein_position,BIOTYPE --assembly "$ASSEMBLY" --offline

Did it work ok for you? Allocate more mem / cores?

[viklund]

Tested a bit. Running vep on 4 cores instead of 1 (with --forks 4 as parameter) solves this.

[glormph]

71 fixes this