SLURM developers, we upgraded a cluster at NSC to SLURM 2.2.4 today (from 2.1.15).
We saw no immediate problems, but during the hours following the upgrade, we received two user complaints about sbatch failing with sbatch: error: Batch job submission failed: Pathname of a file, directory or other parameter too long We found the reason soon in the log file: [2011-05-10T13:51:23] job_create_request: strlen(script) too big (538703 > 131072) The users have been submitting huge batch scripts under 2.1, were it worked. The commit that introduced the check seems to be: > commit 721e60668ce4e496532594cd64f5456f9bf4acfa > Author: Moe Jette <[email protected]> > Date: Thu Apr 15 21:56:06 2010 +0000 > > no change in functionality, but cleaner/smaller logic by adding > subroutine Perhaps the limit ought to be raised (we have compiled a locally patched version with a larger limit)? At least, new limits introduced in 2.2 should probably be mentioned in the release notes. BTW, we are really happy that the git repo is available :-) -- Kent Engström, National Supercomputer Centre [email protected], +46 13 28 4444
