On Mon, Apr 20, 2009 at 4:16 PM, Tomas Kouba <[log in to unmask]> wrote:
> Dear rollout,
>
> is there a suggested way how to have various types of worker nodes in
> one CE and in one
> batch system?
Hi Thomas,
Currently there are two options in both cases though you can and probably
should have one pbs_server.
1) Create two CE nodes representing the two clusters.
2) Publish two clusters and subclusters on the same CE.
http://goc.grid.sinica.edu.tw/gocwiki/How_to_publish_different_memory_limits_for_different_queues_on_the_same_CE
or wait till one day the new richer configuration is released.
https://twiki.cern.ch/twiki/bin/view/EGEE/WNWorkingGroup
which addresses exactly this.
> We are currently moving from PBSPro to torque at our site and we install
> all new nodes
> with 64bit SL5, gLite 3.2 and torque mom.
> Older machines are 32bit SL4, gLite 3.1 and pbspro mom.
> We would like to move to 64bit sl5 where possible, but some nodes have
> CPUs without
> 64bit.
This last bit is interesting. You really still have Pentium IVs or
equivalent that
you have to run 32 bit... I ask since I asked in a forum though possibly not
a big enough one if this we needed a SL5 32 bit WN support... The answer was
a no. There are no plans to currently to provide such a solution. If
sites need that
then raise with justification to your ROC managers or something.
> My question is: what is the "right way" the install this mixture for WLCG?
>
> I see quite few options: one/multiple CEs, one/multiple torque servers,
> different set of queues for 32bit/64bit, keep 32bit on sl4/wait for
> 32bit on sl5.
>
> Thanks for any advice,
>
> --
> Tomas Kouba
> Institute of Physics, Academy of sciences of the Czech Republic
>
--
Steve Traylen
|