Bugzilla – Bug 72
Generic Consumable Resources Do Not Work
Last modified: 2011-07-26 17:12:21 MDT
You need to
before you can comment on or make changes to this bug.
should be able to do the following:
In /var/spool/torque/mom_priv/config on a node:
$usecp *:/home /home
$usecp *:/clusterfs/foo /clusterfs/foo
Now this should be a consumable resource for that node according to the
above-cited documentation. However, it does not appear in the pbsnodes output:
state = free
np = 8
properties = dualcore,photo,nfs
ntype = cluster
status = opsys=linux,uname=Linux n0010 2.6.18-164.6.1.el5 #1 SMP Tue Nov 3
16:12:36 EST 2009 x86_64,sessions=? 0,nsessions=?
Nor can I request it using qsub:
$ qsub -I -l nodes=1:ppn=1:software=nfs_bw -q batch
qsub: Job exceeds queue resource limits MSG=cannot locate feasible nodes
There was a fix made on May 3 that corrected the consumable resource so that it
shows up in pbsnodes. It is available in any of the 2.4.9 snapshots dated later
than May 3. Can you please try a later 2.4.9 snapshot and let us know if it
works okay for you.
This fix probably also corrected the qsub issue since I can issue "qsub -I -l
software=nfs_bw -q batch" and it works just fine.
Created an attachment (id=42) [details]
Patch to fix gres handling in pbsnodes/qsub
This is the patch pulled from SVN changeset 3603 by Al Taufer
2.4.9 does seem to fix both the qsub and the pbsnodes issues. I've yet to have
it actually execute the job, but I'm testing that now. So far, so good though.
I am marking this bug as resolved since I have not heard anything informing me
that there is still a problem. If there still is, please let us know and we
can reopen the issue.