Qsiprep recon dsi studio error

Dear experts,

I am running qsiprep recon with dsi studio workflow and every time I run, recon crashes with error at random stages every time I try to rerun. I don’t seem to find a definitive error. May you please guide what is going on? Crash report is as attached for reference.


docker run -ti --rm -v /home/ubuntu/tracktbi/niftis:/data:ro -v /home/ubuntu/tracktbi/derivatives:/out -v ${fslicense}:/opt/freesurfer/license.txt -v /home/ubuntu/tracktbi/work:/work -v /home/ubuntu/tracktbi/config_files:/config_files pennbbl/qsiprep:0.14.3 /data /out participant --participant-label 011007s20140402 --skip-bids-validation --output-resolution 2 --recon-spec dsi_studio_gqi --recon-input /out/qsiprep --low-mem -w /work

Thank you,
Snehacrash-20211210-204840-root-calc_connectivity.a0-56d207b3-8d00-4145-ae4c-73bff12d221d.txt (82.4 KB)

Error code 137 usually indicates out of memory. How much memory/cpu are you devoting to the job?

Thank you @Steven, I am running it on GCP, for the recon part, I am using a non GPU, n1-standard-4 instance, that would be: Standard 4 CPU machine type with 4 virtual CPUs and 15 GB of memory. I am running it sequentially, so just one subject at a time. What would be ideal configuration?

I usually run QSIPrep/Recon with 32GB and either 8 or 16 CPU (depending on how many parallel jobs I have). It may be overkill but I rarely get out of memory crashes.

Perfect, thank you, let me try with similar config. Also, is there a possibility to supply multiple --recon-spec flag to run multiple workflows?

I think you can only supply one recon spec at a time.

1 Like