As part of the WNWorkingGroup a review of the methods that the experiments use to install software must be made. In particular we want to move from a per CE hostname publishing of tags to a per GlueSubCluster.

Review of How VOs are Installing Software Tags

ALICE

Alice do not use any software tags at all, they will be unaffected by anything.

Geant4

Geant4 do apparently use lcg-VOManageTags, it is unclear how they use it.

CMS

Software installation is done for a list of software at a list of hosts (CEs?). A perl script generates JDL to match the sites. The software tags are then added again manually async to the installation and testing phase. With respect to moving to the per GlueSubCluster installation and publishing it would be wise adjust their perl script.

CMS use lcg-tags.

LHCb

SAM jobs run at all sites and install and validate software if it is missing. They subequently add or remove software tag depending on the validation status. The validation tests are running continuously via SAM. The tags are not used for actual job steering but used as a summary table to check where software is installed effectively for operations purposes.

With respect to moving to per GlueSubCluster publishing this may present some problems for LHCb. They submit jobs and would have to resolve the GlueSubCluster somehow. This takes us back to a single file in $LCG_VO_SOFTWARE_LHCB/glite-subCluster.conf that they can look up on to find possibly a list a GlueSubClusters to look at.

They use lcg-ManageVOTag to do the actual publication.

ATLAS

Use their framework LJSFi. This has a list of GlueCEs from the BDII some of which are marked as auto install, if software not there. For other sites they request software installation.

While some changes would need to be made for installing tags to GlueSubClusters Alessandro De Salvo that the changes would make their life easier especially for those sites that are now publishing multiple architecture GlueSubClusters on the same CE. This is exactly what we plan to make easy for all sites.

Atlas use lcg-tags

Comments from this.

lcg-tags vs lcg-ManageVOTag
The fact that we have two commands to do identical things does not make sense one should be dropped. For the LHC VOs the split is even.
Per GlueSubCluster rather than per CE tag publishing
It is okay for all except LHCb. LHCb would have to add an extra step to resolve the SubCluster they are on. Resolving the SubCluster is easy so long as the proposed glite-subCluster.conf file is installed by the admins. On the +'ve side LHCb don't use the tags for job steering but only for information, i.e it is not mission critical to them so they can adapt in time.

-- SteveTraylen - 07 May 2008

Edit | Attach | Watch | Print version | History: r1 | Backlinks | Raw View | WYSIWYG | More topic actions
Topic revision: r1 - 2008-05-07 - SteveTraylen
 
    • Cern Search Icon Cern Search
    • TWiki Search Icon TWiki Search
    • Google Search Icon Google Search

    EGEE All webs login

This site is powered by the TWiki collaboration platform Powered by Perl This site is powered by the TWiki collaboration platformCopyright & by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding TWiki? Ask a support question or Send feedback