Closed alanocallaghan closed 2 years ago
My naive solution was to try removing the bit where the job properties are written to the jobscript
@Alanocallaghan In case it could be helpful, I put together an example with my smk-simple-slurm profile that uses your idea of a custom jobscript to submit a job with 150k input files:
https://github.com/jdblischak/smk-simple-slurm/tree/main/examples/many-input-files
@Alanocallaghan indeed I think this require some custom solution as @jdblischak suggests. Does his solution solve your problem?
Didn't test it yet tbh, I just used a temp hack to get around it
Yeah works okay, I ended up refactoring the code to avoid this issue anyways... closing the issue as I don't think it'll be resolved.
Thanks @Alanocallaghan and @jdblischak. This issue was really helpful!
FYI https://github.com/snakemake/snakemake/pull/2149 will fix this when/if it is merged
snakemake v7.24.0 should not have this problem now 🎉
I think this may be a general snakemake issue rather than with this profile specifically, but I'm not a python expert.
I have a rule that runs c. 100k jobs and uses all of them together as input to a new rule. Because the input is
expand
ed to 100k items, and the job properties (including input) are written to the jobscript, this means that the jobscript is way over the file size limit forsbatch
. This means you get the error :My naive solution was to try removing the bit where the job properties are written to the jobscript, but this doesn't work because that's where the properties are later read from to set cluster parameters by
slurm-utils
. Other than that I'm not quite sure what to do, other than maybedel
ing some of the properties and then re-writing the jobscript just before submission.