-----Original Message-----
From: Tim Barrass [mailto:[log in to unmask]]
Sent: Tue 9/27/2005 3:44 PM
To: Coles, J (Jeremy)
Subject: RAL, CMS, SC3 (was about VO Boxes ...
Hi Andrew, all,
[ Thanks to Jeremy for fowarding this. Could you include me on cc for
now? ]
Sorry; after a change of email address I am rather crippled wrt
TB-SUPPORT, am trying to resolve this now. This is an email I've been
trying to send in response to the RAL, CMS, SC3 part of this thread.
>>>> Unfortunately RAL seems to be the only CMS Tier1 not fully working
>>>> in
>>>> SC3 today.
>>> If this is the case I don't think we (tier1) are aware of that
>>> though I may
>>> have missed something.
>>
>> Hi John, all,
>>
>> I'm not sure how this fits in to the rest of this discussion, so
>> thought I'd answer separately. As a general point, no CMS Tier 1s are
>> really active in SC3 as our operations revealed some serious problems
>> with Castor at CERN; currently only FNAL is _allowed_ to transfer
>> data, and that only a trickle.
>>
>> As for job submission, all site admins have done so far is send jobs
>> to their own sites to test them. PIC is the first to start large
>> scale testing with remote submission, and that's starting today.
>>
>> RAL has certainly been slow compared to the other CMS T1s. On my part
>> this is because the preparation is pretty much serial in nature. We
>> have, however, also faced problems with gsi access to dCache and grid
>> installation of CMS sw that have not been faced by other sites, and
>> which have delayed us significantly. I believe that thanks to Derek
>> and Steve most of these issues are now resolved (as of a week ago) or
>> in resolution.
>>
>> I'm now involved in the (still moderately lengthy) process of
>> tweaking the CMS services to take part in SC. I have most of the
>> structure needed to transfer files in place; we need to publish the
>> data as well if grid jobs are to run on it.
This covers activity wrt SC3. The same hurdles prevented real
involvement in MC production for CMS. We were able to run jobs at RAL
but unable to get the data out of RAL. This is because we need to
"publish the data" (exactly as above) if we're going to distribute it.
Dave N has asked me to provide some feedback on any way we can improve
our interaction with RAL, and I'm trying to compose that now.
Thanks,
Tim
|