Closed briannadon closed 6 months ago
This is a bug -- thanks for raising it. You can work around it by not using --binariesMode docker
. So either
PATH
to include the cactus binaries. Then run with --binariesMode local
just to be sure. Ordocker run <...> cactus-pangenome <...>
. Thanks for the prompt reply. I actually did start a run entirely within docker before posting this issue and you are correct - it worked.
I am running an alignment of about 90 human pangenome samples with cactus-pangenome, and the pipeline fails nearly every time at the
odgi squeeze
stage. My region is fairly small, about 50kbp, so I don't think it's a memory or resource problem. Oddly, about a week ago on the same machine on the same data and same install of cactus, it worked all the way through, so it seems like this is a problem with toil's job partitioning.My command is:
And the relevant error in the output log is:
The error is claiming
/mnt/results/bnadon/cactus/drb_pan/workdir/toilwf-55733a5be1cf5033b76e0a03c516bb60/7417/b0dd/tmpu364gnhz/dr_region.full.og
does not exist. However, when I look at my directory structure, that file does indeed exist.Restarting the job with
--restart
does not work and results in the same error.Any advice?