April 2009 log
Updated gLite-UI, fixed gLite submission problem to UMD, installed graphics tools on phedex node.
April 27, 2009
MK -- Frontier problems
- Encountered some problems with Frontier server. Suspect problem on CERN's end, but restarted Frontier Squid service.
April 20, 2009
MK -- Updated certs
- Updated user certs on phedex and rsv accounts.
- Updated site certs in OSG.
April 17, 2009
MK -- Fixed GlueCESEBindGroupSEUniqueID/gLite submission issue
- Changing everything to all caps did not cause GIP issue to go away. Changed config.ini, OIM, and SiteDB back to lowercase and emailed hypernews. Burt tracked the source of the GIP error to the setting advertise_gsiftp in our config.ini value. It was reporting multiple values of GlueSEUniqueID associated with hepcms-0.umd.edu and it's assumed to be single-valued. After turning off advertise_gsiftp in config.ini (stopping, reconfiguring, starting), gip_info showed only one value being reported and MyOSG GIP tests worked. glite-wms-job-list-match worked with the requirement Member("hepcms-0.umd.edu" , other.GlueCESEBindGroupSEUniqueID) & CRAB w/gLite scheduler worked! Yay!
April 16, 2009
MK -- Installed OpenGL tools on PN & edited hostname in several places
- Installed several OpenGL devel packages needed for Frog on the phedex node. Frog now compiles, asked Isaac to continue. gdb also installed on PN for debugging Fireworks. Found out about Rocks allowing assignment of group id such as workstation-common in this Rocks thread. This may solve my constant problems with not having the needed packages for interactive users on the WNs. Plan to do this when we convert to separate interactive and worker nodes. May need to do this on the PhEDEx node for Isaac to get Fireworks and Frog working.
- Tracked problem in gLite submissions to glite-wms-job-list-match being unable to match to the requirement Member("hepcms-0.umd.edu" , other.GlueCESEBindGroupSEUniqueID). Confirmed by MyOSG GIP tests giving the error "GLUE Entity GlueCESEBindSEUniqueID does not exist". Initially suspected problem with uppercase hostname in Rocks and lowercase hostname everywhere else. Stopped OSG services, edited config.ini for uppercase, started OSG services, checked new uppercase values in BDII. Also edited OIM & SiteDB registrations to uppercase. Suspected this wouldn't help with datasets in DBS showing us lowercase, but might help with GIP & job-list-match.
April 11, 2009
MK -- Installed gLite-UI 3.1.28-0
- Removed lcg-CA package on all nodes and created a soft link from /etc/grid-security/certificates to the appropriate directory in the OSG area.
- Installed new gLite-UI package via shell script on all WNs, as well as updating the Rocks kickstart file to install 28 instead of 14. Binaries in the gLite-UI area were out of date and CRAB was having problems running with them. Was able to source the OSG environment again to fix the problem, but opted to just upgrade gLite-UI. The name of the gLite-UI environment file changed from 14 to 28. Either way, users were going to have to change something in their behavior, so opted to do the upgrade at the same time. Preliminary indications are that the newer gLite-UI package does not resolve the issue with gLite submission to UMD, though I didn't expect it to.
UMD HEP T3 Computing Cluster