Mail de Sophie (2006-03-27)

Here is (the start of) an LFC service description page : https://twiki.cern.ch/twiki/bin/view/LCG/LfcWlcg

Comments very welcome !

I copied it from the BDII wiki page, so don't be surprised if the BDII appears...

And these pages :

https://twiki.cern.ch/twiki/bin/view/LCG/DataManagementTop

https://twiki.cern.ch/twiki/bin/view/LCG/DataManagementDocumentation

Mail de Sophie (2006-03-16)

A description of all the LFC and DPM commands/methods, as well as all the man pages are now available on the web :

• LFC : https://twiki.cern.ch/twiki/bin/view/LCG/LfcApi

• DPM : https://twiki.cern.ch/twiki/bin/view/LCG/DpmApi

Mail de Sophie (2006-03-13)

Here are some more details about the LFC service :

LFC servers

• lfc001 (lcg_lfc_lhcb_test) : LFC test server for LHCb

• lfc002 / lfc-atlas-test (lcg_lfc_global_w) : central production LFC for ATLAS

• lfc003 / lfc-cms-test (lcg_lfc_global_w) : central production LFC for CMS

• lfc004 / lfc-atlas (lcg_lfc_local_w) : local production LFC for ATLAS.

• lfc005 / lfc-dteam-test (lcg_lfc_dteam_cert_w) : DTEAM certification Oracle LFC (= test machine for release certification).

• lfc006 / lfc-shared (lcg_lfc_shared_w) : central production LFC for DTEAM, UNOSAT and GEANT4.

• lfc007 / lfc-alice (lcg_lfc_local_w) : local production LFC for ALICE.

• lfc008 / lfc-atlas (lcg_lfc_local_w) : local production LFC for ATLAS.

• lfc009 / lfc-cms (lcg_lfc_local_w) : local production LFC for CMS.

• lfc010 / lfc-lhcb (lcg_lfc_lhcb_w) : central READ-WRITE production LFC for LHCb.

• lfc011 / lfc-lhcb-ro (lcg_lfc_lhcb_r) : central READ-ONLY production LFC for LHCb.

We set up 2 local production ATLAS servers, because of a very high ATLAS load at some point.

The aliases are not handled via LANDB, but by a specific procedure (see James/me for details).

OS / RPMS

The OS and the RPMs are managed by FIO.

There is their Quattor profile "profile_lfc003" (for instance) :

http://lemonweb.cern.ch/lemon-status/tpl_view.php?profile=profile_lfc003&host=lfc003

The specific LFC RPMs are in the "pro_software_lfc_slc3":

http://lemonweb.cern.ch/lemon-status/tpl_view.php?profile=pro_software_lfc_slc3

Monitoring

The machines are monitored by Lemon :

http://lemonweb.cern.ch/lemon-status/info.php?cluster=lfc

There can be 3 different LFC-specific alarms :

• LFC_DAEMON_WRONG : after 3 attempts, the "lfcdaemon" service still didn't start

• LFC_DLI_WRONG : after 3 attempts, the "lfc-dli" service still didn't start

• LFC_DB_ERROR : Oracle error detected (i.e. presence of "ORA:" string in /var/log/lfc/log)

The errors are reported to the Service Challenge 2nd level support mailing list : hep-service-sc-level2@cernNOSPAMPLEASE.ch

I guess this will change when the CERN-SC is included in CERN-PROD or CERN-CIC, but until then, should I add you to it ?

LFC Service Mailing List

Currently, the service and third level support mailing list is the same : hep-service-lfc@cernNOSPAMPLEASE.ch

Until the procedure is clearer, should I add you to it ?

Documentation

All the documentation available can be found from :

https://twiki.cern.ch/twiki/bin/view/LCG/DataManagementTop

TO DO

The LFC servers should be upgraded to the last certified version :

• lfc001, lfc004, lfc006, lfc007, lfc008, lfc009, lfc010, lfc011 : LFC 1.4.1 to 1.5.4

• lfc005 : LFC 1.4.5 to 1.5.4

No configuration change needed.

Only RPM upgrade and restart of the service.

The time has to be agreed with the Service Challenge coordination team beforehand.

Mail de James Casey (2006-03-14)

I've updated the edg-mkgridmap.conf on all nodes in the LFC cluster to use VOMS (they were just using LDAP and were missing new people). For a temp measure until the new cluster with YAIM is created, I've put in a better grid-mapfile.

David, Yvan : I'm going to give you root access on all the clusters (i.e fts, gridfts, lfc, gridlfc).