Large-Memory Jobs
Priorities
Queues on Grommet

Jobs with large memory requirements may need special handling to run efficiently. In particular, jobs requiring more than about 512 MB of physical memory to run will need to raise the process memory use limit.

To see the current resource limits in effect for your shell (memoryuse, datasize, stacksize, etc.) type: limit

To see the maximum resource limits available, type: limit -h

To raise the current limits of your shell to the maximum limits, type: unlimit
After typing unlimit, then submit your job. The child processes should inherit the limits from the parent process.

You can place the unlimit command in your .cshrc or .profile file to ensure that the limits are always raised.

These commands work for all shells on grommet (tcsh, csh, bsh, ksh).

All jobs on grommet expected to take more than 1 hour should run "weightless" using the npri -w command. Running background jobs weightless improves interactive performance on grommet (everything from executing shell commands to running visualization software over the network). Please be considerate of other users and follow the priority policies.

NOTE: Running a script that submits a series of short jobs (each under 1 hour, but overall creating a long job) without using npri -w is a violation of the priority policy. The policy allowing short jobs to run at a higher priority is meant to help people debug programs or new calculation methods efficiently. An automated series of short jobs is actually a long job. Running such a series without reducing the priority of the job violates the spirit, if not the letter of this policy.

• Jobs running with incorrect priorities may have their priorities altered and the user will receive a reminder about the priority policies.
• If a user repeatedly runs jobs with incorrect priorities, those jobs may be killed.

Please submit all of your jobs using the following commands (csh/tcsh and sh/bsh/ksh).

npri -w command_to_run_job.

NOTE: You cannot change the priority of a running job using npri (only the systems administrator can - blame SGI). If your forget to submit your job using npri, you have two options to correct it:

• use ps -lfu username or top to get the process ID number of your job (pid) and kill and restart the job with the proper priority, OR
• if you can't kill it (or it is very undesirable to kill it, e.g. because you didn't notice right away), send email to pmcmahon@udel.edu with information about the job and the priority will be fixed.

All About Grommet...	Still to come...
Grommet Home Page Participating Faculty Funding Sources Research from Grommet Description of the Facility Software Available on Grommet	Interesting Links for Grommet Users and others

This page is maintained by Patrick McMahon.
Last Updated: 7 July 2008
The URL for this page is: http://www.udel.edu/chem/grommet/help