Setup
- pgsc_calc: v2.1.0
- Nextflow: 25.10.4
- Profile:
gcb (Google Cloud Batch)
- Input: Per-chromosome GT VCF files (22 chromosomes, single sample from 1000 Genomes)
- Scoring: PGS002209 with
--run_ancestry
Samplesheet:
sampleset,path_prefix,chrom,format,vcf_genotype_field
HG00096,/mnt/disks/work-volume/HG00096.chr1,1,vcf,GT
HG00096,/mnt/disks/work-volume/HG00096.chr2,2,vcf,GT
...
HG00096,/mnt/disks/work-volume/HG00096.chr22,22,vcf,GT
Problem
Processes that use storeDir fail with Missing output file(s) versions.yml even though the command completes successfully (exit code 0).
This happens because all 22 chromosome tasks write versions.yml to the same storeDir path concurrently. The data files don't collide (they include chromosome in the filename), but versions.yml is always the same name. On GCS FUSE this causes stale file handles or missing files.
Retries eventually succeed (cached outputs are found in storeDir and tasks are skipped), but this adds 30+ minutes per retry and is non-deterministic.
Errors
ERROR ~ Error executing process > 'PGSCATALOG_PGSCCALC:PGSCCALC:MAKE_COMPATIBLE:PLINK2_VCF (HG00096 chromosome 2)'
Caused by:
Missing output file(s) `versions.yml` expected by process `PGSCATALOG_PGSCCALC:PGSCCALC:MAKE_COMPATIBLE:PLINK2_VCF (HG00096 chromosome 2)'
Command exit status:
0
ERROR ~ Error executing process > 'PGSCATALOG_PGSCCALC:PGSCCALC:ANCESTRY_PROJECT:INTERSECT_VARIANTS (HG00096 chromosome 16)'
Caused by:
Missing output file(s) `versions.yml` expected by process `PGSCATALOG_PGSCCALC:PGSCCALC:ANCESTRY_PROJECT:INTERSECT_VARIANTS (HG00096 chromosome 16)'
Command exit status:
0
Steps to reproduce
- Prepare per-chromosome VCF files for any sample
- Run pgsc_calc with
-profile gcb on Google Cloud Batch
- Pipeline fails at
PLINK2_VCF or ancestry modules (INTERSECT_VARIANTS, FILTER_VARIANTS, etc.)
- The failing chromosome varies between runs
Affected processes
All modules using storeDir where multiple chromosomes write to the same directory:
PLINK2_VCF, PLINK2_RELABELPVAR, PLINK2_RELABELBIM
INTERSECT_VARIANTS, FILTER_VARIANTS, EXTRACT_DATABASE
RELABEL_IDS, RELABEL_AFREQ
FRAPOSA_PCA, FRAPOSA_PROJECT, PLINK2_MAKEBED, INTERSECT_THINNED, PLINK2_ORIENT
Setup
gcb(Google Cloud Batch)--run_ancestrySamplesheet:
Problem
Processes that use
storeDirfail withMissing output file(s) versions.ymleven though the command completes successfully (exit code 0).This happens because all 22 chromosome tasks write
versions.ymlto the samestoreDirpath concurrently. The data files don't collide (they include chromosome in the filename), butversions.ymlis always the same name. On GCS FUSE this causes stale file handles or missing files.Retries eventually succeed (cached outputs are found in
storeDirand tasks are skipped), but this adds 30+ minutes per retry and is non-deterministic.Errors
Steps to reproduce
-profile gcbon Google Cloud BatchPLINK2_VCFor ancestry modules (INTERSECT_VARIANTS,FILTER_VARIANTS, etc.)Affected processes
All modules using
storeDirwhere multiple chromosomes write to the same directory:PLINK2_VCF,PLINK2_RELABELPVAR,PLINK2_RELABELBIMINTERSECT_VARIANTS,FILTER_VARIANTS,EXTRACT_DATABASERELABEL_IDS,RELABEL_AFREQFRAPOSA_PCA,FRAPOSA_PROJECT,PLINK2_MAKEBED,INTERSECT_THINNED,PLINK2_ORIENT