Frequently asked questions

UK grid deployment support hosts an FAQ to answer questions relating to the deployment of LCG-2 on UK sites in general. This page provides additional information not hosted on the main GridPP support FAQ, usually as it will relate to London Tier 2 specific issues.
  • What mailing lists should I be a member of?
  • The main FAQ gives the mailing lists you should join as part of LCG and GridPP.
    As part of the London Tier 2, you should join:
    • The London Tier 2 Mailing list
    • This is intended to be a comprehensive list of all active and interested participants in the London Tier 2 centre.
    • LT2 Technical and Support Group
    • If you are actively involved in the deployment or maintenance of a grid site as part of the London Tier 2 you should join this list.

    If you have any problems with these lists, please contact the London Tier 2 Technical Co-ordinator

  • How do I install a CE to send jobs to an existing PBS server on a different machine?
  • In the CE configuration file add the line:
    #define PBS_MASTER %pbsservername%

    where %pbsservername% is the name of the existing pbs server, before
    #include "pbs-server-cfg.h"

    This should miraculously take care of all the jobmanager and info provider problems! There will then be lots of problems with the installation of LCG and experiment software required on the WN's. A recipe is being tested for this and more details will be forthcoming.
  • Who are the contacts for LHC experiments use of the Grid?
  • This link gives a list of the support contacts for experiment software, but these contacts are more from a user point of view. The Experiment Software Managers, who are responsibile for the actual deployment of the experiment software on LCG-2 sites can be discovered from the VO server using:
    ldapsearch -LLL -h grid-vo.nikhef.nl -p 389 -x -b "o=$VO,dc=eu-datagrid,dc=org" "(ou=lcgadmin)" member
    where $VO must be replaced by the name of the VO.

    At the time of writing, the ESM's are:

    1. ALICE
    2. Giuseppe Lo Re
      Piergiorgio Cerello
      Stefano Bagnasco
      Daniele Mura
    3. ATLAS
    4. Alessandro De Salvo
      Miguel Branco
    5. CMS
    6. Marco Corvo
      Alessandra Fanfani
      Andrea Sciaba
    7. LHCb
    8. Joel Closier
      Andrei Tsaregorodtsev
  • How do I change the home directories of all my pool accounts?
  • Edit the following files:
    1. Users-cfg.h
    2. s%/home/%/grid/home/%g
      
    3. user-edguser-cfg.h
    4. s%/home/%/grid/home/%g
      
    This will change all the pool accounts to a different home directory on all the nodes. Unfortunately the lcfg object does not deal with this cleanly. If you have already created the /home directories, LCFG will not create new ones.

    Comment out the

    #include Users-cfg.h 
    
    in the node configuration files, remake the profiles and restart the lcfg service on the client. This deletes the users from the client. The uncomment the Users-cfg.h from the client nodes, remake the profiles and restart the lcfg objects again. This time, LCFG believes these are new accounts so creates the /grid/home directories. You will have to manually delete, or move, the old /home directories, though, as LCFG does not clean this up when the users are removed.

    Also, the LCFG auth object doesn't seem to know about the "-p" option when creating the home directories. You will need to do an mkdir -p /grid/home before restarting the lcfg service.

  • How do I change the G/UID's for the pool accounts?
  • GID's are easy. Edit the file Users-cfg.h and change the lines:
    auth.groupgid_alice            1395
    auth.groupgid_atlas            1307
    auth.groupgid_cms              1399
    auth.groupgid_lhcb             1470
    auth.groupgid_dteam            2688
    
    to values you like. NB. if using NFS mounts these gid's must be the same on all machines.

    UID's are simple, but tedious if you are very restricted in ranges. The basic principle is the same, you need to change all the lines in Users-cfg.h that are like:

    auth.useruid_alicesgm              18941
    auth.useruid_alice001              10417
    ...
    auth.useruid_alice050              18055
    
    If you know, for example, that the alice VO needs to be in the range 20000-30000, you could do this by a wild card substitution:
    s/\(auth.useruid_alice??? *\) ?/\1 2/g
    
    If your ranges must be more tightly defined you may find you need to manually edit 50 lines per VO.
  • How do I change the group name for a VO?
  • Suppose your site is integrating with an existing PBS farm, for which the group name of a VO, such as "atlas" is already taken. Edit the following files:
    1. Users-cfg.h
    2. change
      EXTRA(auth.groups)             atlas
      
      to
      EXTRA(auth.groups)             atlasgrid
      
      Then do a global search and replace:
      s/\(auth.usergroup_atlas???  *)atlas/\1atlasgrid/g
      
    3. flatfiles-dirs-SECLASSIC-cfg.h
    4. change
      dirperm.owner_atlas          root:atlas
      
      to
      dirperm.owner_atlas          root:atlasgrid
      
    5. voswmgr-dirs-cfg.h
    6. change
      dirperm.owner_atlassgm                   atlassgm:atlas
      
      to
      dirperm.owner_atlassgm                   atlassgm:atlasgrid
      
    7. user-edguser-cfg.h
    8. change
      EXTRA(auth.usersuppgroups_edguser) atlas
      
      to
      EXTRA(auth.usersuppgroups_edguser) atlasgrid
      
  • Where are the slides from the LCG2 Administrator Course?
  • Here
  • The LCFG based setup to create password free ssh from WN to CE didn't work.
  • Try this manual check.

The London Tier 2


Last modified Thu 13 January 2005 . View page history
Switch to HTTPS . Website Help . Print View . Built with GridSite 1.4.3
For more about GridPP please contact Neasan O'Neill