Fasterq Dump Issue

Updated: October 6, 2023

Edit this Page via GitHub       Comment by Filing an Issue      Have Questions? Ask them here.

fasterq-dump is a command-line utility used to extract FASTQ or FASTA data from SRA-accessions.

There is a known issue with the way the application interacts with the version of filesystem used for Scratch storage on the cluster. This bug causes cluster nodes to become unresponsive and require a reboot to restore to normal operation when fasterq-dump is used with Scratch for reading or writing data.

Scratch should not be used with the fasterq-dump application. To run fasterq-dump on the Gizmo cluster please use one of the following workarounds:

Workarounds

Use Fast instead of Scratch

One option is to use Fast storage instead of Scratch for temporary data. A major downside to this approach is that it can end up creating a significant amount of wasted space in backups, particularly when data is frequently created and deleted. Additionally, performance may be significantly lower than with Scratch when used for temporary data.

Use local scratch space on the cluster nodes

Another alternative is to use the local scratch space available on the cluster nodes. This is available to cluster jobs at /loc/scratch/<jobid> where <jobid> is replaced by the actual SLURM job id. This location is also stored in the TMPDIR and SCRATCH_LOCAL environment variables.

It is important to note that this location only exists for the duration of the job, so any data here that needs to be kept would need to be copied to persistent storage (e.g. Scratch or Fast) by the job before it exits.

The amount of local scratch space needed should be specified with the –tmp argument to sbatch when submitting a job.

More information on node local scratch space is available at the Job local storage page

Use parallel-fastq instead

parallel-fastq can speed some operations- this is available in the module parallel-fastq-dump/0.6.7-GCCcore-11.2.0.

Updated: October 6, 2023

Edit this Page via GitHub       Comment by Filing an Issue      Have Questions? Ask them here.