Storage and retention¶
Sequencing depth can make raw and intermediate data substantially larger than a sample count alone suggests. Omnomnomics provides explicit retention policies and a soft project-size guard.
Retention policies¶
| Policy | Behavior |
|---|---|
all |
keep all outputs and intermediates |
pruned |
keep FASTQ and reusable downstream products; remove bulky upstream intermediates after successful use |
minimal |
keep FASTQ and only terminal outputs required by the requested branches |
omnomnomics rna \
-i EXPERIMENT \
-g GRCh38 \
--retention-policy pruned
Cleanup is stage-aware. trimmed_FASTQ/ is not eligible until lane merging has
completed, and BAM/ is not eligible until filtered BAM creation has completed.
Lightweight trim and mapper metrics are copied to run_logs/flow_qc_cache/
before their source directories are removed, preserving later reporting.
Soft size guard¶
omnomnomics rna \
-i EXPERIMENT \
-g GRCh38 \
--retention-policy minimal \
--max-project-size 300G
The guard measures project usage at storage checkpoints and attempts only safe cleanup. If essential FASTQs and filtered BAMs already exceed the cap, the run records a warning. Space-heavy BigWig/track-hub branches may be skipped while requested count-table work continues.
This is a soft workflow policy, not a reservation. It cannot prevent failures caused by another project consuming the same filesystem quota or by temporary job-local scratch exhaustion.