[slurm-users] Nastygramming (was: Prevent users from updating their jobs)

[Loris Bennett]

Hi Samuel,

"Fulcomer, Samuel" <samuel_fulcomer at brown.edu> writes:

[snip (5 lines)]

> Our practice has been to periodically look at running jobs to see if
> they are using (or have used, in the case of bigmem) less than their
> requested resources, and send them a nastygram telling them to stop
> doing that.

[snip (28 lines)]

We already do a fair bit of nastygramming (although I like to feel it's
more infogramming, or, at the worst annoyagramming) to let users know
about things like their disk usage, the impending expiry of their
university account, or the fact that they haven't logged in for a
certain period.

I have also started generating histograms of CPU and memory efficiency
from seff data along with a short textual report indicating whether the
user needs to look more carefully at the resources he or she is
requesting.

I would be interested in the following:

1. Do you use some kind of framework to automate the actual sending of
   the nastygrams?

2. What metrics do you use for deciding whether a nastygram regarding
   resource usage needs to be sent?

Cheers,

Loris

-- 
Dr. Loris Bennett (Herr/Mr)
ZEDAT, Freie Universität Berlin         Email loris.bennett at fu-berlin.de