How do I run my programs on the production cluster?


First you need to have an account on the cluster, if you do not have an account yet contact support@stat.ucla.edu.

The production cluster (PC) allows you to run your programs/jobs only by submitting them to the scheduler.

The instructions on how to use the scheduler are as follows:

Login to compute.stat.ucla.edu

Type in:
% cluster
This will cd you into your home directory on the PC.
The PC has it's own fileserver, so any files that your program uses will have to be copied into this directory.

For (s)ftp copy your files to:
/Network/Servers/node-00/Volumes/craid/Home/username

When you submit your job to the scheduler you have to: You can submit jobs to the scheduler in the following ways:

A successfully submitted job will return a job id, something like:
username:1077215811:34

The files this job creates are:
username:1077215811:34.err (contains error messages, check this file if things go wrong)
username:1077215811:34.log (environment variables and job related info)
username:1077215811:34.out (any output your program produces is stored in this file (basically everything that normally would go to STDOUT)

To check a job's status while running
% checkjob jobid

To check the availability of nodes:
% checknode

Or for more details:
% checknode -V

To cancel a job:
% canceljob jobid

To check the status of the scheduler:
%showq

 


If you have any questions regarding this topic e-mail to support@stat.ucla.edu

Date Created: 2004-04-12 10:54:33 Date Last Modified: 2004-05-18 15:06:55


UCLA Department of Statistics
Last updated: 13-Jul-2009
Access count is: 39758, since 06-Apr-2004
Maintained by: Web Staff [webstaff@stat.ucla.edu]