> -----Original Message-----
> From: Maarten Litmaath [mailto:[log in to unmask]]
> Sent: 12 September 2012 19:38
> To: Ewan MacMahon
>
> > For testing purposes I think it might
> > make sense if both Stephen Jones and I install that update on our EMI
> > 2 worker nodes when it's available, but leave one cluster running the
> > current EMI 2 WN release, plus that fix, and if we upgrade the other
> > cluster to the beta EMI 2 WN release, plus that fix, then we get the
> > interesting VOs to test on both systems.
> >
> > How does that sound to everyone?
>
> Pretty good! I'll have fries with that. :-)
>
> What would be the name(s) of the test queue(s)?
The Oxford ones are behind the CE 't2ce02.physics.ox.ac.uk', and
since they're fully in production they can ben looked up with
lcg-infosites and all the usual mechanisms. We have four queues,
one for just ops and dteam:
t2ce02.physics.ox.ac.uk:8443/cream-pbs-expressfive
and three of different lengths for 'real' VOs:
t2ce02.physics.ox.ac.uk:8443/cream-pbs-longfive
t2ce02.physics.ox.ac.uk:8443/cream-pbs-mediumfive
t2ce02.physics.ox.ac.uk:8443/cream-pbs-shortfive
which differ in the amount off wall and CPU time allowed. All
four queues run on the same nodes, but there's a reserved slot
for dteam and ops jobs, so as long as they're short and there
aren't too many, they should run quickly.
As of now, the nodes behind t2ce02 are still running the current
EMI 2 WN release on SL5, and without the fix for the BDII lookup
bug. Once we have that fix available, I'll get them upgraded to
the EMI 2 beta.
Ewan
|