How will experiment software be installed and tested on LCG-2 sites?

It is recognised that

The release cycle for experiment software may be far too rapid to require a separate release of LCG-2 tagged software for each development of experiment code
Individual site managers are unlikely to have a high level of expertise in all the experiment software that is likely to be installed on grid sites

A different software installation method has been developed for this. Each VO has designated Experiment Software Managers that are responsible for deploying and maintaining the software at individual sites. All the sys admin must do is provide a suitably mounted disk space and set appropriate environment variables. The methodology is described here or in PDF format. A detailed recipe for the sys admin requirements is below.

The easiest solution for this is to define an area on the SE say

/experiment-software/alice /experiment-software/atlas ...
that is automounted on the WNs.

Users locate this area since
VO_ATLAS_SW_DIR=/experiment-software/atlas
is defined in the the jobs profile.

This is how to do it with LCFG. If you are doing it by hand that is just an excercise for the reader.

In you site-cfg.h you should have the following. #define EXPERIMENT /experiment-software #define WN_AREA_ALICE EXPERIMENT/alice #define WN_AREA_ATLAS EXPERIMENT/atlas #define WN_AREA_CMS EXPERIMENT/cms #define WN_AREA_LHCB EXPERIMENT/lhcb #define WN_AREA_DTEAM EXPERIMENT/dteam

The first line is an extra value to add.

On the SE you must now create these directories. You do that by adding the following to its profile. #ifdef SE_VO_ALICE EXTRA(dirperm.ents) alicesw dirperm.path_alicesw EXPERIMENT/alice dirperm.owner_alicesw alicesgm:alice dirperm.type_alicesw d #endif #ifdef SE_VO_ATLAS EXTRA(dirperm.ents) atlassw dirperm.path_atlassw EXPERIMENT/atlas dirperm.owner_atlassw atlassgm:atlas dirperm.type_atlassw d #endif #ifdef SE_VO_CMS EXTRA(dirperm.ents) cmssw dirperm.path_cmssw EXPERIMENT/cms dirperm.owner_cmssw cmssgm:cms dirperm.type_cmssw d #endif #ifdef SE_VO_LHCB EXTRA(dirperm.ents) lhcbsw dirperm.path_lhcbsw EXPERIMENT/lhcb dirperm.owner_lhcbsw lhcbsgm:lhcb dirperm.type_lhcbsw d #endif #ifdef SE_VO_DTEAM EXTRA(dirperm.ents) dteamsw dirperm.path_dteamsw EXPERIMENT/dteam dirperm.owner_dteamsw dteamsgm:dteam dirperm.type_dteamsw d #endif

Now you must nfs export this area by again adding your SE's profile

    EXTRA(profile.components)        nfs
    profile.version_nfs              1
    boot.level_nfs                   2
    EXTRA(boot.services)             nfs
    client.reconfig_nfs              start

    /* Export the File System */
    EXTRA(nfs.exports)        experiment
    nfs.fs_experiment         EXPERIMENT
    nfs.options_experiment    SITE_NFS_ACL_FROM_SE

Reboot the SE after recompiling the profiles.

We now want to automount this onto each of the WNs by adding the following their your profiles, putting this in a common file and including it of course makes sense.

    EXTRA(profile.components)       autofs
    profile.version_autofs          1
    EXTRA(boot.services)             autofs
    boot.level_autofs               2
    client.reconfig_autofs          start

    EXTRA(dirperm.ents)      experiment
    dirperm.path_experiment  EXPERIMENT
    dirperm.owner_experiment root:root
    dirperm.type_experiment  d

    EXTRA(autofs.maps)        experiment
    autofs.path_experiment    EXPERIMENT

    #ifdef SE_VO_ALICE
    EXTRA(autofs.map_experiment) alice
    autofs.path_experiment_alice SE_HOSTNAME:EXPERIMENT/alice
    autofs.key_experiment_alice  alice
    autofs.options_experiment_alice -rw,intr,noquota,nosuid,noatime,hard
    #endif

    #ifdef SE_VO_ATLAS
    EXTRA(autofs.map_experiment) atlas
    autofs.path_experiment_atlas SE_HOSTNAME:EXPERIMENT/atlas
    autofs.key_experiment_atlas  atlas
    autofs.options_experiment_atlas -rw,intr,noquota,nosuid,noatime,hard
    #endif

    #ifdef SE_VO_CMS
    EXTRA(autofs.map_experiment) cms
    autofs.path_experiment_cms SE_HOSTNAME:EXPERIMENT/cms
    autofs.key_experiment_cms  cms
    autofs.options_experiment_cms -rw,intr,noquota,nosuid,noatime,hard
    #endif

    #ifdef SE_VO_LHCB
    EXTRA(autofs.map_experiment) lhcb
    autofs.path_experiment_lhcb SE_HOSTNAME:EXPERIMENT/lhcb
    autofs.key_experiment_lhcb  lhcb
    autofs.options_experiment_lhcb -rw,intr,noquota,nosuid,noatime,hard
    #endif

    #ifdef SE_VO_DTEAM
    EXTRA(autofs.map_experiment) dteam
    autofs.path_experiment_dteam SE_HOSTNAME:EXPERIMENT/dteam
    autofs.key_experiment_dteam  dteam
    autofs.options_experiment_dteam -rw,intr,noquota,nosuid,noatime,hard
    #endif

Unfortuantly LCG in their wisdom decided to remove the autofs lcfg object from WNs. So you must add edg-lcfg-autofs-1.0.3-edg1 to the rpm list for WNs.
There are two steps in this process:
1. Download and install edg-lcfg-autofs-defaults-s1-1.0.3-edg1.noarch.rpm on your LCFGng server, with rpm -ihv edg-lcfg-autofs-defaults-s1-1.0.3-edg1.noarch.rpm.
2. Download edg-lcfg-autofs-1.0.3-edg1 then:
  - Move it to /opt/local/linux/7.3/RPMS/LCFGng
  - cd to /opt/local/linux/7.3/RPMS/LCFGng and execute gmake.
  - Follow the instructions here for adding the edg-lcfg-autofs to the RPM list for your Worker Nodes.
Recompile the profiles for the WNs and reboot them.
Afterwards on a WN check you can
cd /experiment-software/atlas
for example.
If all is well the area will be automounted.

LCG FAQS

Last modified Mon 23 August 2004 . View page history

Switch to HTTPS . Website Help . Print View . Built with GridSite 1.4.3

Sections

About GridPP

Documents

Activities

Management

Help

Information

How will experiment software be installed and tested on LCG-2 sites?