Minutes of GridPP testbed/lcg phone conference.

Thursday 6th May , 14:00


Present:
Rhys Newman (Ox), Santanu Das (Cam), Steve Thorn (Ed), Alessandra (Man), Steve Traylen (RAL), Dave Kant (RAL), Chris Brew (RAL), Barry Saunders (RAL), Jeremy Coles (RAL), Frederic Brouchu (CAM), Mark Nelson (Durham), Laurie Lowe (Birm), Pete Clark (UCL), Pete Gronbech (Ox), Owen Maroney (Imperial), Dave Colling (Imperial), Ben Waugh (UCL), Paul Kyberd (Brunel), Gregori Rybkine (RHUL), Sukhbir Johal (RHUL).


Agenda:

   + Any changes in the current release of LCG and others.
	No change to the LCG distribution.
	Feels like a stable test bed and stable system.
	Problems when a rogue site has a miconfigured replica manager.
	

   + Changes and new developments from sites.
     Scotgrid
       Edinburgh, Glasgow, Durham
	EDIN - Not a great deal of change, almost up with LCG2.
	DURHAM - using sun grid engine - need to do a manual install. Has a crib sheet to help the install. Part way though SE install. Should be posted on the web/distribution once all this pain has been gone though.
	

     London Grid
       Imperial, RHUL, UCL, Brunel, ...
	Imperial - Accpeted into core zone. 50-60 CPUs
	QMUL - Test zone using their EDG array 18nodes.
	UCL - Upgrading EDG, 1 worker node + SE. Not on the test zone yet and no form filled in.
	Brunel - LCFNG server running - trying to get the CE installed.
	RHUL - Process of installing, LCFNG installed+CE+WN but not SE.
	

     Southgrid
       Oxford, RAL-PPD, Cambridge, Bristol, Warwick
	RAL-PPD Running final tests now. Now have 42 CPU's and final testing. Online in 2 days or so.
	CAM Altas software has been installed on worker nodes. Otherwise, all the same, 1 CE 1 SE and 9 WN. LCFNG server crashed - minor problem. New TB storage will be installed soon. In core zone (have been for some time).
	Birm - LCFGng server loaded. Touble with booting client from that (RAM disk??). Will try increasing RAM disk size. Hope to be on line before long. 
	Bristol - Not much here as there are no FTEs or computing resources around for Tier 2. Rhys will be sorting out these issues with Dave Newbold over the next few weeks.
	Warwick - No information.
	Oxford - Finally powered on 10 of the 40 nodes, the rest are racked up and awaiting elextrical testing. Starting with LCFGng today, should be online with 10 nodes by end of next week. The rest should follow by the end of May, pending power/cooling/space.
	
     NorthGrid
       Manchester, Sheffield, Liverpool
	MAN - LCFGng installed, 4 WN and SE but not yet any experiment software (soon). Awaiting 26 nodes on the 17 May.
	Sheff - In test zone.
	Lanc - In test zone.
	LIV - 120 nodes of their 500 are to be dedicated to Tier 2. Will go next week to supervise the process.

   + Installation/Technical Problems
	
	None reported

   + GridPP Provided Frontends.

       Are these suitable - no comment, so assumed yes.
       Which sites actually want them
       How do we procede?

	From Dave Colling:  Puchased though RAL. Will arrive in a matter of weeks depending on RAL funding. 4 machines available for every site which can dedicate 30 nodes + 1 TB. Anticipate 17 sites qualify. Any feedback through the technical coordinators. Dave will circulate a list of those who will get these. Tender process hasn't started yet, but expect tenders from IBM, Clustervision, DELL etc. Tender process will start ASAP, will email coordinators when this starts.


   + Monitoring and Progress
	http://map.gridpp.ac.uk  This is a map of all the sites in the UK - Red not good, Green in the test zone.
	Green Dots is good, but also getting those resources out into the community. Is someone using a monitoring system (e.g. GridICE) to show how many resources are actually available. Some discussion about what is an appropriate measure of "total resources". A green dot merely indicates "at least one job went through". This test job should be more complex to exercise more of the site infrustructure. May introduce a "Green Star" to indicate success of a more complex test job.

	Steve, using gridICE showed - 19TB, 8186 CPUs currently.
	
   + Next Meeting
	
	In about a month. Steve T will convene, either that or meet face to face at GridPP.