London SC4 Activity

CMS Activity

The CMS activity is currently focusing on the CSA2006 that will start in September. The current activities are

• Setup PhEDEx at Imperial to manage the transfer for each London Site.
• Setup each Site with CMSSW and a trivial catalog to be able to receive production jobs from ProdAgent.

The status of sites in the UK can be seen here. Note there are problems accessing DPM from CMSSW (versions 081/082/083) via rfio. It is not yet clear whether this is fixed in CMSSW090.

Brunel

• CMSSW081 installed via a job - needs testing/verifying. Required to install the libGLU and libcurl on all the wn nodes.
• Phedex transfers of 2GB loadtest files timing out due to low bandwith - transferring 100MB test data in progress.
• Also needs rpm-build and libf2c on worker nodes
• CMSSW083 & CMSSW090 installed via job - need testing/verifying.
• /opt/d-cache/srm/sbin/srm edited on worker nodes, new line 161: SRM_CONFIG_PARENT=`dirname $SRM_CONFIG`
• Brunel is validated (2006-09-15) and running production jobs

19/09/2006

• Duncan observed that the merging was not working. Probably because the dpm hack was not in place. I have done it.

QMUL

• CMSSW081 manually installed by Olivier.
• 11/08/06: Nicola tested running ProdAgent jobs. Is validated
• Phedex transfers going on but se very fragile. FTS channels set to 1 file in parallel.

18/09/06

• Made a reservation using QOS on cn100 for the cmssgm.

SRCFG[test] TASKCOUNT=1 RESOURCES=PROCS:1
SRCFG[test] USERLIST=cmssgm
SRCFG[test] HOSTLIST=cn100
SRCFG[test] QOSLIST=test1
QOSCFG[test1]    QFLAGS=USERESERVED
USERCFG[cmssgm]  QDEF=test1

• Installed apt on cn100

yum install apt installed apt-0.5.15cnc6-4.SL

• Made the storage.xml and site-local-config.xml configuration files.

19/09/06

• I checked the cmssw installation fired yesterday.

CMSSW_0_8_3,CMSSW_0_8_4,CMSSW_1_0_0 have been sucessfully installed.

• Installed the DPM hack for RFIO found here

Imperial

IC-HEP

• apt-get installed in the VO_CMS_SW_DIR area. maui configured to send cmssgm jobs on only one machine
• CMSSW081 manually installed
• 11/08/2006: ProdAgent running.

IC-LESC

• Could not install the cmssw081 since the rpm database is 64 bit and apt-get does not understand that

RHUL

• Maui configured to send cmssgm job to sft reserved node (usually node75/node76).
• Following rpm's added to this node: curl, rpm-build, apt.
• CMSSW090 seems installed via job - need testing/verifying.
• Problems directing cmssgm job to reserved nodes - above rpms extended to all nodes.
• CMSSW083 installed via job - need testing/verifying.
• 18/09/06: Need to put the right permissions on dpm and also create the relevant groups

dpns-entergrpmap --group "cms/Role=production"

dpns-entergrpmap --group "cms/Role=lcgadmin"

dpns-setacl -m "g:cms/Role=lcgadmin:rwx,m:rwx" /dpm/pp.rhul.ac.uk/home/cms/

dpns-setacl -m "g:cms/Role=production:rwx,m:rwx" /dpm/pp.rhul.ac.uk/home/cms/

dpns-setacl -m "d:g:cms/Role=lcgadmin:rwx,d:m:rwx" /dpm/pp.rhul.ac.uk/home/cms/

dpns-setacl -m "d:g:cms/Role=production:rwx,d:m:rwx" /dpm/pp.rhul.ac.uk/home/cms/

• Then we need to create the /store directory for the production data

dpns-mkdir /dpm/pp.rhul.ac.uk/home/cms/store/

• /opt/d-cache/srm/sbin/srm edited on worker nodes, new line 161: SRM_CONFIG_PARENT=`dirname $SRM_CONFIG`