Hi, I did some further investigation. This is not a configuration problem, but a bug that was introduced in 2.2.5 and is still present in the latest 2.2 branch.
Moe Jette <[email protected]> writes: > Your test is working fine for me with the v2.3.2 code. The bug was merged into 2.3 before 2.3.0-pre5, and fixed right after 2.3.0-rc2. This means all released (not pre/rc) 2.3 versions have been ok. slurm-2.2: Introduce the bug: 2011-04-02 103c7ce7324abe90cabd1476da00c355afe3680e reduce frequency that we sent SIGKILL to jobs being killed. slurm-2.3: Merged the bug: 2011-04-05 736e18369c4a0e4981054c3cf85777f5c271bb38 svn merge -r22954:23010 https://eris.llnl.gov/svn/slurm/branches/slurm-2.2 Fixed the bug: 2011-08-24 fe46ecc09eb82311800cf7ee6eb78db61a408837 Improve enforcement of KillWait time For 2.2, either reverting 103c7ce7324abe90cabd1476da00c355afe3680e or cherry-picking fe46ecc09eb82311800cf7ee6eb78db61a408837 will resolve the problem. In our local 2.2 branch I have done the later, it would be nice if you would do this on the official 2.2 branch as well. I am guessing that we are not the only ones that have not had time to upgrade to 2.3 yet. Kind regards, Pär Andersson NSC
pgpY84P4ffsKG.pgp
Description: PGP signature
