[GE users] Sort by sequence number question
macinnis at dal.ca
Tue Jul 17 16:41:47 BST 2007
Hi Andreas and Erik,
On Tue, 17 Jul 2007 Andreas.Haas at Sun.COM wrote:
> On Tue, 17 Jul 2007, Paul MacInnis wrote:
> >> I did this with N1GE 6.1
> >> Could it be that jobs are submitted with -soft option as to specify some
> >> preferece? Or are you using some over-sensitive load thresholds?
> > This is a pretty good duplication of our setup here. However on your
> > "qsub" you use "-q test_*" to select a domain of queues for the
> > scheduler. We don't use -q at all, the scheduler chooses cluster
> > queues as it pleases.
> I used the "-q test_*" just to exclude other queues and the -q option
> has no impact on order of queue consideration.
> > So I would claim that by default the scheduler
> > considers 1 cluster queue at a time to satisfy a request and only
> > looks at additional cluster queues if no instance satisfies.
> Wrong claim. The scheduler considers one queue instance after the
> other. Cluster queues play no role at all with the order of
> consideration. If you don't believe me, I'll be happy to proof
> it with source code :-)
> > Try something that doesn't select queues by name, perhaps a
> > resource request that allows you only on these test queues.
> That is pointless. Queue selection does not play into the order
> of consideration. All it does is filtering. Really ;-)
> That means there must be something different with your setup.
> For that reason I sent you these questions with my last mail:
> (1) Could it be that jobs are submitted with -soft option as to specify some
> (2) Or are you using some over-sensitive load thresholds?
OK, Erik's test and the source code prove me wrong.
The jobs here have no -soft options but they do have load thresholds:
hostlist @1g.hg @2g.hg @4g.hg
load_thresholds load_avg=1.5,mem_used=500M,[@2g.hg=load_avg=1.5, \
hostlist @2g.hg @4g.hg
load_thresholds load_avg=1.5,mem_used=1.5G,[@4g.hg=load_avg=1.5, \
Each 2G and 4G node has a ser.q and a bg.q queue instance, each with same
load_thresholds, but scheduler has a definite preference for the bg.q
instance, inspite of the higher seqno! Perhaps in time an explanation
will appear ...
Thanks for looking at this problem(?).
To unsubscribe, e-mail: users-unsubscribe at gridengine.sunsource.net
For additional commands, e-mail: users-help at gridengine.sunsource.net
More information about the gridengine-users