We currently derive dynamically it from maui, but maui has been timing out at times when asked so it may appear and disappear at random intervals. I am planning to get it created semi statically from our maui quattor configuration soon.
Derek
On 28 Jan 2011, at 16:23, John Gordon wrote:
> RAL-LCG2 was publishing shares when I looked the other week.
>
>
> -----Original Message-----
> From: J Coles
> Sent: Jan 28, 2011 3:04 PM
> To: [log in to unmask]
> Subject: Publishing VO shares
>
>
>
>
> Dear All
>
> You may be aware that the WLCG Management Board is currently validating information being published in gstat (for GridPP http://gstat-prod.cern.ch/gstat/summary/GRID/GRIDPP/). At the moment very few UK sites publish VO shares but we are now encouraged to do this (our ticket is https://gus.fzk.de/ws/ticket_info.php?ticket=66564 and for those who feel this is not necessary see a relevant discussion here https://gus.fzk.de/ws/ticket_info.php?ticket=66469). These shares are NOT used for scheduling, they are purely for project management and oversight (i.e. to see what is available to a VO).
>
> Some of us discussed the UK figures in the deployment meeting on Tuesday but we need to recheck details in the GridPP (Steve Lloyd) allocation spreadsheet before confirming the values we would expect to be published. If you happen to know already for your site please go ahead and publish now. Sites that are explicitly mentioned in the ticket as needing to do this are:
>
> EFDA-JET
> UKI-LT2-Brunel
> UKI-LT2-IC-HEP
> UKI-LT2-QMUL
> UKI-LT2-UCL-HEP
> UKI-NORTHGRID-LANCS-HEP
> UKI-NORTHGRID-LIV-HEP
> UKI-SCOTGRID-DURHAM
> UKI-SCOTGRID-ECDF
> UKI-SCOTGRID-GLASGOW
> UKI-SOUTHGRID-BHAM-HEP
> UKI-SOUTHGRID-BRIS-HEP
> UKI-SOUTHGRID-CAM-HEP
> UKI-SOUTHGRID-OX-HEP
> UKI-SOUTHGRID-RALPP
> RAL-LCG2
>
> We will discuss the request and response in more detail at Tuesday's dteam and sites meeting (the main reason I am sending this message today is as background for those discussions).
>
> For now here is what Duncan had to say back in October when we last discussed this area:
>
> "I think that the reason that some Tier-2s are showing entries for
> HEPSPEC06 for given VO here
>
> http://gstat-wlcg.cern.ch/apps/capacities/comparision/
>
> and some are not is that some sites may have filled out the fairshare
> for each VO, e.g. for RHUL the site-info.def entry:
>
> CE_CAPABILITY="CPUScalingReferenceSI00=1975 Share=atlas:97 Share=cms:1
> Share=lhcb:1 Share=others:1"
>
> results in this in the information system:
>
> RHUL
>
> GlueCECapability Share=atlas:97
> GlueCECapability Share=cms:1
> GlueCECapability Share=lhcb:1
> GlueCECapability Share=others:1
>
> and looking at the ATLAS share, 97% of the total site HEPSPEC06 of 3160
> is 3065.2. Looking at the above web-site the HEPSPEC06 column for London
> is 3065 (I don't think the other London sites publish the shares).
> Similarly, for NorthGrid
>
> Manchester: 18482 HEPSPEC06
>
> GlueCECapability Share=atlas:40
> GlueCECapability Share=cms:5
> GlueCECapability Share=lhcb:40
>
> Sheffield: 3900 HEPSPEC06
>
> GlueCECapability Share=atlas:80
> GlueCECapability Share=cms:1
> GlueCECapability Share=lhcb:20
>
> and so for ATLAS the NorthGrid total will be
>
> 0.4 * 18482 + 0.8 * 3900 = 10512.8
>
> NorthGrid have 10513 listed as their ATLAS capacity. So I think in order
> for this to be properly populated all sites should enter the proportions
> they give to each of the LHC VO's."
>
> regards,
> Jeremy
|