Open mathiasbio opened 3 weeks ago
From Eva 2024-06-14: Another leaky rule seems to be picard_umiaware:
OpenJDK 64-Bit Server VM warning: Insufficient space for shared memory file:
75256
Try using the -Djava.io.tmpdir= option to select an alternate temp location.
Alright, last one. Most failed jobs are either due to picard_umiaware or to cadd_annotate. But there is also one failing job (also in comp 27) failing on BALSAMIC.bettercrappie.cnvkit_segment_CNV_research.142.sh_6710424.err With similar errors, so this might also be another rule to look at:
RuntimeError: Subprocess command failed:
$ Rscript --no-restore --no-environ /var/tmp/tmpdqpd9oxy
b"Fatal error: cannot create 'R_TempDir'\n"
Description
At the moment there's a bit of a mix in the use of tmp dirs in the rules in our snakemake workflows. Some rules have this in params:
tmpdir = tempfile.mkdtemp(prefix=tmp_dir)
And this in the command:Other rules don't have anything regarding tmpdirs, such as:
Which failed in production due to too full tmp on the worknode compute-0-27 (which was close to full due to a leaky rule in my development branch which saved a bunch of inprogress bamfile-chunks in there)
We should try to find the correct way to assign tmpdirs and make this consistent across all of our rules
How to reproduce
No response
Expected behaviour
No response
Anything else?
No response
Pipeline version
15.0.0