Storage accounting consistency checklist

From GridPP Wiki
Jump to: navigation, search

This page will track the consistency of the Storage Accounting system with the actual level of resources at GridPP sites. Issues and how they were resolved are recorded.

Summary Table

Green is good (meaning the condition is satisfied), red is the opposite. See the detailed site report below for an explanation.

Site Name Is site publishing distinct resources per VO? Are actual site resources consistent with Storage Accounting?
ScotGRID-Edinburgh
UKI-SCOTGRID-DURHAM
UKI-SCOTGRID-GLASGOW
UKI-NORTHGRID-LANCS-HEP
UKI-NORTHGRID-LIV-HEP
UKI-NORTHGRID-MAN-HEP
UKI-NORTHGRID-SHEF-HEP
UKI-SOUTHGRID-BHAM-HEP
UKI-SOUTHGRID-BRIS-HEP
UKI-SOUTHGRID-CAM-HEP
UKI-SOUTHGRID-OX-HEP
UKI-SOUTHGRID-RALPP
UKI-LT2-Brunel
UKI-LT2-IC-HEP
UKI-LT2-IC-LeSC
UKI-LT2-QMUL
UKI-LT2-RHUL
UKI-LT2-UCL-CENTRAL
UKI-LT2-UCL-HEP
RAL-LCG2


ScotGrid

ScotGRID-Edinburgh

Edinburgh is reporting the correct used storage. The dCache is configued such that all of the LHC VOs have their own storage resources (i.e. pool groups) and is using the dCache.org provided GIP plugin so that storage is not being n-tiply counted.

UKI-SCOTGRID-DURHAM

Durham are using the new GridPP provided DPM GIP plugin and so are reporting the correct used storage per VO.

UKI-SCOTGRID-GLASGOW

Glasgow are using the new GridPP provided DPM GIP plugin and so are reporting the correct used storage per VO.

NorthGrid

UKI-NORTHGRID-LANCS-HEP

Lancaster are running dCache. They have split up their disk such that Atlas has s large quantity of dedicated space. They have moved back to using the dcache.org GIP plugin. The small VOs are sharing space, but the quantity of disk used by them is only of the order of 10's of GB compared to Atlas's 15TB of used space.

UKI-NORTHGRID-LIV-HEP

There is currently (26-02-2007) a problem with Liverpool and they are not publishing any information to the BDII. From looking at historical information it appears that they do not have separate resources allocated to the (main) VOs. Rather, they are all sharing the same set of disk. They need to set up the relevant pools and pool groups in their dCache. It would help if their dCache webpage was publicly visible.

UKI-NORTHGRID-MAN-HEP

Manchester is correctly publishing the storage in the BDII. It is the way gstat or other aggregating WEB clients query ldap that don't read correctly.

UKI-NORTHGRID-SHEF-HEP

Sheffield have now deployed the DPM GIP plugin, meaning that they are correctly accounting for used storage per VO.

SouthGrid

UKI-SOUTHGRID-BHAM-HEP

Birmingham are using the new GridPP provided DPM GIP plugin and so are reporting the correct used storage per VO.

UKI-SOUTHGRID-BRIS-HEP

Bristol are using the new GridPP provided DPM GIP plugin and so are reporting the correct used storage per VO.

UKI-SOUTHGRID-CAM-HEP

Cambridge have NOT (26-02-2006) deployed the new GridPP provided DPM GIP plugin.

UKI-SOUTHGRID-OX-HEP

Oxford are using the new GridPP provided DPM GIP plugin and so are reporting the correct used storage per VO.

UKI-SOUTHGRID-RALPP

The dCache is correctly reporting disk used per VO. There is a problem with the accounting display mechanism which causes incorrect information to be plotted for the ralpp VO.

LondonT2

UKI-LT2-Brunel

Brunel are using the new GridPP provided DPM GIP plugin and so are reporting the correct used storage per VO.

UKI-LT2-IC-HEP

Their dCache is split up such that CMS has a large amount of dedicated resources. This is being correctly published to the BDII using the dcache.org provided GIP plugin.

UKI-LT2-IC-LeSC

Nothing is publihsed for IC-LeSC since they share their SE with IC-HEP. As each SE can only have a single GlueSEUniqueID, it does not make sense to publish the same used space as for IC-HEP since this would lead to double counting when looking at the overall view of the LT2. Really, the storage accounting is picking out the space information on each SE, not on each site (it just happens that in the majority of cases each site has at least 1 SEUniqueID).

UKI-LT2-QMUL

QMUL are using the new GridPP provided DPM GIP plugin and so are reporting the correct used storage per VO.

UKI-LT2-RHUL

RHUL are using the new GridPP provided DPM GIP plugin and so are reporting the correct used storage per VO.

UKI-LT2-UCL-CENTRAL

Still to deploy the GridPP provided DPM GIP plugin. Olivier has opened a ticket against the site. He does not have root access to the machines so is unable to do it himself.

UKI-LT2-UCL-HEP

UCL-HEP are using the new GridPP provided DPM GIP plugin and so are reporting the correct used storage per VO. They do not currently have much used storage (26-02-2006).

Tier-1

RAL-LCG2

See RAL Tier1 CASTOR Accounting