There are two ways of accessing Purdue TeraGrid systems:
If you have any problems following the above steps, or if you can not log on after following these steps, contact the TeraGrid Helpdesk.
Once your account is enabled you can submit jobs via Condor-G to tg-condor.purdue.teragrid.org/jobmanager-condor or simply login to tg-login.purdue.teragrid.org and use condor_submit.
You must define the ClassAd TGProject in your submit file so that your usage can be matched with your Teragrid allocation.
+TGProject = "your project number"
Failing to include this ClassAd will prevent Condor from running your job!
For remote submissions, use the "project" RSL attribute for your Project number:
globus-job-run tg-gatekeeper.purdue.teragrid.org/jobmanager-condor -x '&(project=TG-XYZ12345)' /bin/hostname
With Condor-G, use the "globusrl" expression in your submit file:
globusrsl = (project=TG-XYZ12345)
If your condor job requires CTSS, add the following Requirements ClassAd:
Requirements = (HasCTSS =?= True)
This will ensure that only a node with CTSS available to it will be chosen to run your job. If your job requires outbound internet connectivity (ie, outside of purdue.edu):
Requirements = (CanReachInternet == True)
If your job requires a different architecture (i.e. Windows):
Requirements = ( Arch=="Intel") && ( OpSys=="WINNT51" ) Requirements = ( Arch=="IA64") && ( OpSys=="LINUX" ) or Requirements = ( Arch=="SUN4U") && ( OpSys=="SOLARIS" )
You can add arbitrary condor ClassAds to your Condor-G globusrsl. The following example will request a node that has CTSS installed, and outbound internet connectivity:
globusrsl = (condorsubmit=(Requirements 'HasCTSS == TRUE && CanReachInternet ==TRUE') )
TeraDRE has made a significant leap in usability with the addition of WebStart job submission. You can get started using web submission from the official TeraDRE homepage http://teradre.rcac.purdue.edu TeraDRE WebStart currently supports Maya, POVray, Blender, & Gelato. Other software packages are in the works. In addition, TeraDRE will soon be released as open source to further aid development in the graphics community.
You should run long jobs in the Standard Universe, otherwise it will continue to be stopped and restarted (from scratch) and never finish.
Another option is to develop some application-level checkpointing capability in your jobs.
A very useful program for finding out which machines and architectures are out there, is the program condor_all. It should be noted, that even though it is located in the "official" Condor directory - /opt/condor/bin, it is a locally (Purdue) developed tool. It is very handy for finding out how many of a certain machine architecture that are available. This is very useful for the submit description file.
Another useful command to see how many machines is there, and how many are available, is condor_pool. Example:
-bash-3.00$ condor_pool -t
POOL
-------
egret.rcac.purdue.edu (Total=590,Unused=20)
broker.ics.purdue.edu (Total=619,Unused=65)
condor.calumet.purdue.edu (Total=348,Unused=347)
emu.rcac.purdue.edu (Total=2127,Unused=118)
flamingo.rcac.purdue.edu (Total=2256,Unused=613)
Use the command condor_q.
You can run various Globus tests to narrow the possible error down, and determine if it is something in the connection which is wrong, rather than your script. Look here for some example tests:
http://wwwdev.rcac.purdue.edu/userinfo/howtos/teragrid/howto_teragrid.cfm#testing
First, note that this is only for Purdue Affiliates!
Second, this step is not really necessary, since you just need a certificate from one of the TeraGrid sites and you can generate one yourself from NCSA when you have gotten your TeraGrid account.
To get a Purdue certificate you must show up in person and speak with Matthew Lundmark on 6th floor of Young Hall. You must bring some sort of ID with you. It is also a good idea to bring some media to copy the certificate to - like a USB drive.
He will then setup a Purdue Certificate for you and copy it to your USB drive. The certificate consists of the files: usercert.pem and userkey.pem. There will also be a file describing how to install the certificate.
Installing the certificate:
The TeraGrid User portal can be reached from the menu bar on any of the User Info pages on http://www.teragrid.org or by linking to http://portal.teragrid.org.
Then, on the Portal home page, enter your portal username and password in the corresponding fields in the login form, and click the "Login" button. You should have received your portal username and password in your start package.
If you don't see the Login form immediately after you load the portal in your web browser, click the Login link in the upper right-hand corner of the page.
When you have logged in to the Portal, there will be 6 tabs in the top menu. They are the different services available from the portal. Under each tab is a set of navigation links for services within each main service category. Below is taken from the official TeraGrid page.
If you need help, then click the question mark icon in the upper left corner of the title bar. It will display the available help documentation.
Many of the pages can be customized. If that is the case for a certain page, then there will be a pencil shown in the upper left corner of the title bar. Click the icon to see/change your personal preferences.
When your TeraGrid account was created, a Distinguished Name (DN) was created on your behalf using the NCSA certificate authority. This DN is stored in your profile in our account database and automatically propagated to all sites when your account is created. A proxy certificate with this same DN can be retrieved from the TeraGrid MyProxy certificate repository (myproxy.teragrid.org) on any TG resource by using the myproxy-get-delegation command and supplying your portal username with the -l option and your portal password as the MyProxy passphrase.