Hi,
We have been running slurm-14.03.0-0pre5 quite successfully on our 700 node cluster since the
beginning of February, using slurmdbd and msyql for accounting. We are using Scientific Linux 6.4.
I've experienced very long waits for sreport AccountUtilizationByUser to complete - in fact I've had
to rebuild it with a larger SLURMDBD_TIMEOUT to get output at all. The following command took 33
minutes to return (after increasing the timeout value from 15 mins to 1 hour):
sreport -t hours cluster AccountUtilizationByUser start=2014-02-01T00:00:00
end=2014-02-26T23:59:59
The final report contained information on about 500 accounts and users, and the number of jobs in
this 26 day period was around 90000.
Raw sacct data by comparison returns output from the same period in a few
seconds.
My question is am I probably doing something incorrect to see sreport performance like this, with
this many jobs and users over this sort of period?
Many thanks for your help.
Best regards,
Stuart
--
Dr. Stuart Rankin
Senior System Administrator
High Performance Computing Service
University of Cambridge
Email: [email protected]
Tel: (+)44 1223 763517
