UMD HEP T3 Computing ClusterWelcome
New user? Start here
Fellow T3 admin? Start here
News
- September 30, 2009: CRAB_2_6_3 is installed and linked by /scratch/crab/current. It hasn't been verified to be bug free at UMD, please notify Malina Kirn if the client crashes (some job failures are normal).
- September 10, 2009: CRAB_2_6_1 is installed and linked by /scratch/crab/current. CRAB_2_6_2 is also installed, but not linked by /scratch/crab/current as it appears to be buggy.
- August 29, 2009: You can again ssh to hepcms.umd.edu, which will now send you to one of our two interactive nodes.
- August 27, 2009: Our DBS registration name has been fixed. You can revert to 'proper' CRAB syntax of:
se_white_list = T3_US_UMD
ce_white_list = umd.edu - August 26, 2009: We are no longer registered with our 'proper' SE name of T3_US_UMD in DBS. We are still registered with our FQDN of hepcms-0.umd.edu. To submit CRAB jobs to our cluster, set:
se_white_list = umd.edu, UMD.EDU, T3_US_UMD
ce_white_list = umd.edu, UMD.EDU, T3_US_UMD
We'll be fixing our DBS registration such that the proper syntax will work again soon. - August 24, 2009: CMSSW_2_2_12 & 2_2_13 are now installed.
- August 17, 2009: We can now service all CRAB jobs. To force CRAB jobs to run at our site, set:
se_white_list = T3_US_UMD
ce_white_list = umd.edu
Note that we now support 'standard' SE output syntax in crab.cfg. e.g., when copy_data=1, you can use:
storage_element = T3_US_UMD
user_remote_dir = subdir
Further details are in the user guide. - August 16, 2009:
- The cluster is back up and can service CRAB jobs. However, it cannot service glite CRAB jobs (in crab.cfg, set scheduler=condor_g). Also, its storage element output is currently down. CRAB jobs can be submitted to the cluster which run over data hosted at the cluster, but CRAB jobs which try to send output back to the T3_US_UMD SE via the copy_data configuration option in crab.cfg will fail. You can choose to send output to a different SE if you prefer.
- The CMSSW installation directory has changed to /sharesoft/cmssw. Edit your ~/.cshrc & ~/.bashrc files and change:
setenv VO_CMS_SW_DIR /software/cmssw
to
setenv VO_CMS_SW_DIR /sharesoft/cmssw
Your existing CMSSW release areas will no longer work and must be reinstalled. - The OSG installation directory has changed to /sharesoft/osg. Edit your ~/.cshrc & ~/.bashrc files and change:
source /share/apps/osg/setup.csh
to
source /sharesoft/osg/ce/setup.csh
- August 15-16, 2009: The cluster is down for scheduled upgrades.
Status
Note that most status pages do not produce output that is easily readable. If you are a user looking for current cluster load, the best monitor for your needs is probably Ganglia. If you want the status of the batch job submission system, use the Condor monitor.
- Ganglia
- Condor
- Squid Frontier server (primarily intended for cluster admins)
- RDC temperature monitor (primarily intended for cluster admins)
- Dell OpenManage (primarily intended for cluster admins)
- Grid monitors:
- Grid policy
Site map
How To -User guide: Learn how to get an account, connect to the cluster, transfer files, run CMSSW, submit condor jobs and more.
How To -Admin guide: A step-by-step guide with instructions specific to install and configure all software needed by our cluster. Includes a list of critical files, instructions for recovery from failure, and solutions to encountered errors.
Help -For users: A list of emails and links to get further information. Includes links for CMSSW, Root, Condor, CRAB, DBS, PhEDEx, and CVS.
Help -For admins: A list of emails or listservs to get further help. Includes addresses for Rocks, OSG T3 sites and CMS grid tools.
Configuration: A description of the cluster, including the functions of the various nodes, the hardware, disk partitions and network configuration. Intended for admins and advanced users.
Log: A work log of all changes done to the cluster. Intended for admins, users should consult the news above for any relevant announcements.