On Tue, 26 Jul 2016 at 1:35pm, Benjamin Redling wrote
On 2016-07-25 22:46, Joshua Baker-LePain wrote:
I think that my initial question was too complex/detailed. Let me ask a
more open-ended one. Do folks have any strategies they'd like to share
on partition setups that favor paying customers while also allowing for
usage of spare resources by non-paying users? Thanks!
On Fri, 15 Jul 2016 at 3:56pm, Joshua Baker-LePain wrote
o Our "lab" queue is for contributing users. Jobs in this queue run
un-niced, and each lab has a number of slots in this queue equal to
their share of the cluster.
o Our "long" queue is for all users. Jobs in this queue run "nice -19".
Really? You use the _lowest niceness_ == highest priority for all users?
So the "un-niced" (nice 0, medium nicety) jobs by the contributing users
get less cpu time?
o We also have a "short" queue for quick jobs. These jobs run at "nice
-10" and are limited to 30 minutes.
Your short running jobs have a _lower_ (_negative_ nice == not nice)
priority than your long running processes?
The short running jobs are limited in time _and_ risk getting starved
due to _less nice long running processes_?
Maybe I misunderstood what you are trying to accomplish, but currently
your usage of "nice" sounds wrong to me.
It's not our usage, it's my explanation -- sorry about that. As you
suspected, lab.q jobs are at nice 0, short.q jobs at nice 10, and long.q
jobs at nice 19. All positive values.
--
Joshua Baker-LePain
QB3 Shared Cluster Sysadmin
UCSF
