[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Condor-users] Forcing a job classad from config file?

I'm just using the default without the runtime (StateTimer) calculation.
 PREEMPTION_REQUIREMENTS = RemoteUserPrio > SubmittorPrio * 1.2
All the other policy stuff is set at condor_config "TESTINGMODE" defaults.

Note that I am also testing with some policy rules from the Bologna Batch System
paper. Instead of "Long" or "Short" jobs my focus is more on "Normal" vs.
"Priority" users. The priority user jobs can either be self-declared in the
submit file or described in classad expressions. In this scenario the priority
user gets access to an idle Virtual Machine. The normal user job still runs
on the host but at background, or nice-level, priority. Not sure if we will do
this yet, but note that each startd machine could be configured for a different
"priority" set of users to partition resources that way.

Note also that my users consist of a _very_ small, well-behaved, group that play
well with each other. So I'm not looking for bullet-proof, just simple and workable.

And another disclaimer - I have not solved any real production problems yet. I'm
still getting familiar with Condor and testing various ideas to run by my boss and


From: "Ian Chesal" <ICHESAL@xxxxxxxxxx>

The problem I'm trying to solve in the current production
environment is that User-A would submit thousands of vanilla
jobs (one or more clusters).
Runtime for each job is typically under 1 hour. User-B
submits a few jobs and never gets access to any machines. The
greater insult is that User-A then submits more jobs and they
run before the User-B jobs.
I don't necessarily want preemption to
kill/checkpoint/restart the User-A jobs, just to insert a
wedge so User-B can get access to some resources within a
reasonable period of time. I stumbled on MaxJobRetirementTime
from reading this mailing list - not finding it in the
version 6.6.7 manual, began exploring 6.7.3. It does EXACTLY
what I need - simple, clean, and straight-forward when used
with a simple PREEMPTION_REQUIREMENTS expression based on
priority, and a shorter (than 1 day) PRIORITY_HALFLIFE.


I'd be very interested in hearing how you're changing
PREEMPTION_REQUIREMENTS. I'm doing something very similar and have run
into a few issues. See the threads with subjects "Default User priority
factor", "When do machine RANK settings apply?" and "about scheduling
algorithm in condor and condor-g" for some background on what I am
trying to do. Does that look familiar?


- Ian C.

Condor-users mailing list