Re: [Condor-users] condor rocks

[Erik Aronesty <erik@xxxxxxx>]

>  Let us know what config changes couldn't be reconfiged. We're always looking to make fewer and fewer changes require a restart.

- If you change the number of cpus available to a slot, a restart is
needed -  or if you switch from static to dynamic.   This is
reasonably well documented.

- Would be nice if there was a way to have a machine decide to wait to
free up cpus for a multicpu job if that job is old or high priority or
something.  What I do now is force a couple machines to only accept
jobs with requestcpus > 1.   That way the singles don't ever starve
out the multiples.

- I have this script running in cron...   this is necessary because
Vacated jobs will get stranded forever if they didn't originally
request enough memory:

condor_q -f "%s\t" ClusterId -f "%s\t" ImageSize -f "%s\n"
RequestMemory | perl -ne '($p, $i, $m) = split("\t");
$i=1+int($i/1024); print "condor_qedit $p.0 RequestMemory $i\n" if $i
> $m;' | sh

Probably should just be something you can configure... it's a bit
obtuse.