A quick fix is to edit all of the mom_priv/config files and add "arch xeon"<br>
at the end of the file then restart the moms. When you submit you'll<br>
have to add "-l arch=xeon".<br>
<br>
Of course, then your users have to do this each time and there's no<br>
reason for the separate queues. Perhaps a better way is to use nodesets.<br>
<a href="http://www.clusterresources.com/products/maui/docs/8.3nodesetoverview.shtml">http://www.clusterresources.com/products/maui/docs/8.3nodesetoverview.shtml</a><br>
<br>
For instance in server_priv/nodes<br>
xeonnode np=2 pXEON<br>
<br>
Then in maui:<br>
FEATUREPARTITIONHEADER p<br>
NODESETATTRIBUTE FEATURE<br>
NODESETLIST pXEON pAMD<br>
<br>
CLASSCFG[ruche] DEFAULT.NODESET=ONEOF:FEATURE:XEON<br>
<br>
This separates your cluster into partitions, so jobs cannot cross into xeon and<br>
amd machines, and also does the limiting of ruche jobs to only xeon nodes.<br>
<br>
Even better, you don't have to edit all the mom config files, just the server and<br>
maui configs.<br>
<br>
-Justin.<br><br><div><span class="gmail_quote">On 11/23/06, <b class="gmail_sendername">Jacques Foury</b> <<a href="mailto:Jacques.Foury@math.u-bordeaux1.fr">Jacques.Foury@math.u-bordeaux1.fr</a>> wrote:</span><blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">
Ryan Thompson a écrit :<br>> Hi Mike,<br>><br>> We currently use an Intel Xeon submit node and all our compute nodes<br>> are amd opterons.<br>><br>> If you use separate queues it should be fine.<br>>
<br>> Cheers<br>> --<br>> Ryan Thompson,<br>> HPC Systems Administrator<br>> <a href="mailto:ryan@zymeworks.com">ryan@zymeworks.com</a><br><br>Hi Ryan and all others.<br><br>Here we have a problem we've not seen yet.
<br><br>I've parted the cluster with 11 Xeon nodes (named ruche) and 14 Opteron<br>nodes (named callas).<br><br>I've made pure opteron queues and pure xeon queues with :<br><br>- in the nodes file<br>xeonname np=2 xeon ruche
<br>...<br>opteronname np=2 opteron callas<br><br>- in the queues definitions<br>set queue ruche resources_default.neednodes = ruche<br>set queue callas resources_default.neednodes = callas<br><br>- in the MAUI config<br>
<br>CLASSCFG[callas] MAXJOB=48<br>CLASSCFG[ruche] MAXJOB=22<br><br>And we thought all was all right !<br><br>BUT if we submit more than 48 jobs in the callas queue, the 49th job is<br>submitted to a xeon node !!!<br><br>How can I ensure this never happens ?
<br><br>(what I found is :<br># checknode xeonname<br><br><br>checking node xeonname<br><br>...<br><br>Features: [xeon]<br>Attributes: [Batch]<br>Classes: [callas 2:2][ruche 2:2][batch 2:2]<br><br><br>how comes the "Features" tell only "xeon" "ruche" but we can see
<br>"callas" in the classes ?????)<br><br><br>Thanks for any help...<br><br><br>--<br><br>Jacques Foury<br>Ingénieur d'études<br>Institut de Mathématiques de Bordeaux<br>Université Bordeaux 1 / CNRS<br>Tel : 05 4000 69 56
<br>Fax : 05 4000 21 23<br><a href="http://www.math.u-bordeaux.fr/maths/cellule">http://www.math.u-bordeaux.fr/maths/cellule</a><br><br>_______________________________________________<br>torqueusers mailing list<br><a href="mailto:torqueusers@supercluster.org">
torqueusers@supercluster.org</a><br><a href="http://www.supercluster.org/mailman/listinfo/torqueusers">http://www.supercluster.org/mailman/listinfo/torqueusers</a><br></blockquote></div><br>