Purdue CMS
CMS/OSG Resources at Purdue Print
There are quite a few resources that are available, but it's quite easy to get lost with what's where. To make life easier, this doc will be broken up into three sections: those users who are accessing the Purdue resources through grid based submissions, local accounts, and things that are common to both. If you have a local account on our resource, and a grid certificate, then you can also use the information tailored for grid users.

Common Information

Site Information

ServiceHosts
Globus Gatekeeper (pre-ws)

osg.rcac.purdue.edu (fork, pbs, mis)
lepton.rcac.purdue.edu (fork, pbs, mis)
grid.physics.purdue.edu (fork, condor, pbs, mis)
itb.rcac.purdue.edu (fork, pbs, mis)

Globus Gatekeeper (ws)itb.rcac.purdue.edu:9443 (fork, pbs)
dCache Serverdcache.rcac.purdue.edu
Phedex/Pubdb Serverphedex.rcac.purdue.edu
GRIS Server (GIP support)osg.rcac.purdue.edu port 2135
mds-vo-name=local,o=grid

Queue Information

PBS QueueInfo
cms-production@hamletx86 CMS queues dedicated to cms production jobs. (1)
cms-grid@lear
cms-analysis@lear

Intel x86-64 CMS queues. The purpose is for cms analysis and jobrobot.
preemptx@hamletOur x86 preemptable queue. All other VO jobs go here by default. Jobs here don't count against the number of allocated nodes, but are subject to being requeued during heavy load.
preempti@lear The Intel x86-64 preemptable queue. The same condition applies for the 32bit counterpart.
 

Note 1: See the appropriate grid/user specific sections to see how the queues are automatically selected.

Note 2: We are currently in the process of streamlining the process of job submission without having to designate a machine name for the queue. Accessible to regular users only is a program: cmsqsub which will dynamically select which machine to perform job submission to based on load. Be warned that one cluster is 32 bit, and the other is 64 bit which shouldn't impact any running binaries, but will impact building or execution.

Additional notes: All sites support srm and gridftp. To access the datasets, it is strongly encouraged that you use srm, and connect to dcache.rcac.purdue.edu. The contents of /cms are the datasets that we are subscribed to, and /VO/cms is for VO specific storage to be used throughout. For those who are participating in the CSA06, the data lives in /store/data/CSA06. Recently, we installed $OSG_APP/cmssoft/cms_site_catalogue.sh to assist users with the generation of the orcarc fragments used to obtain a dataset that we are currently subscribed to.

VO Specific app goodies

$APP or $OSG_APP will contain the location of all VO specific apps. For CMS, the apps are located in $OSG_APP/cmssoft. To use the CMS apps, source the file $OSG_APP/cmssoft/cmsset_default.sh APP_VERSION where APP_VERSION is the application and version of the package you wish to use.

Grid Users

When submitting a job to Purdue's gatekeeper, you may use any of the three servers mentioned above, but the preferred gatekeeper is osg.rcac.purdue.edu, and the preferred jobmanager is pbs. If the grid user established their proxy with a voms server (by using voms-proxy-init), then the jobmanager's default queue will be changed to reflect the role. Otherwise, if you are a cms user submitting a grid job, then your default queue will go into cms-grid. If not, then your job will go to the preemptable queue.

Local Users

To make use of the Open Science Grid's VDT software, the first file that will need to get sourced is /opt/osg/setup.sh or /opt/osg/setup.csh. This will provide access to all the grid tools and utilities.

The second file to source is $GLOBUS_LOCATION/etc/grid3-info.conf which is used to provide the common environment variables such as $OSG_APP_DIR for the location of the cms applications, $GRID3_DATA_DIR for the location of node accessable work directory, $OSG_TMP_DIR for global scratch, and $OSG_TMP_WN_DIR for temporary worker node scratch.

If you would like to submit a job, you can use qsub or cmsqsub to submit directly to the pbs queue. You can also submit a grid job by using Globus, or condor_submit for condor-g.


If you have any questions or problems, don't hesitate to let us know by emailing This email address is being protected from spam bots, you need Javascript enabled to view it .
 
< Prev

 

 

Utilization
1134 jobs running
Total: 1728 job slots
Jobs running: 1134
Jobs queued: 0
dCache
52% utilization
Total: 441 TB
52% Used, 48% Free

This site, and the work it describes, is primarily funded by a grant from the National Science Foundation (NSF).