Hello All,
So, never got SLURM 14.11 to work, upgrading to 14.11.1 didn't help; with
our setup, Raw Usage was only accumulated from the longest running job and
for the owner of that job.
Downgrading to 14.03.10, however, did the trick. Now everything apparently
works as it should, and Raw Usage is accumulated from all running jobs for
all users. I guess this means there is something broken in 14.11, but I
have no idea what. The logs never showed anything useful (that I could
interpret, that is).
Anyway, just in case someone encounters a similar problem, this serves as
a documented work-around.
Cheers,
Mikael J.
http://www.iki.fi/~mpjohans/
On Thu, 27 Nov 2014, Mikael Johansson wrote:
Hello again,
A bit more detail, in case it helps. It seems that the Raw Usage is only
updated based on the longest running job on the system (or perhaps only for
the user who owns the oldest job). As the "oldest job" owner changed, the Raw
Usage started accumulating only for the user with the new "oldest job".
Still, any suggestion very much appreciated.
Cheers,
Mikael J.
On Thu, 27 Nov 2014, Mikael Johansson wrote:
Hello All,
OK, so yesterday we upgraded to 14.11.0. Everything went rather smoothly,
except for one thing: Accounting of Raw Usage seems not to be working
properly.
It works partly, yesterday it seemed to be working for all users, for the
moment only the Raw Usage for one user is updated, while all others stay
at zero (after resetting Raw Usage with sacctmgr).
It seems like an odd problem, as SLURM apparently does succeed in updating
Raw Usage in principle, but not for all users. I cannot find anything
special in the logs.
Are there perhaps any special settings one should take care of after
upgrading so many versions? The config files (slurm.conf, slurmdb.conf)
are still the same as for version 2.2.7, and I've attached them as .txt
files to this.
All help appreciated. Cheers,
Mikael J.
http://www.iki.fi/~mpjohans/
