HTCondor

From NJIT-ARCS HPC Wiki
Jump to: navigation, search

Using HTCondor

The purpose of this page is to guide users to the specifics of using HTCondor at NJIT. For an overview of HTCondor and detailed usage instructions, see the user manual.

HTCondor Environment

To include the HTCondor commands in the user's path, at the prompt enter --

module load condor

HTCondor Nodes

Compute Nodes

All of the 66 oslN.njit.edu machines are HTCondor compute nodes. These machines have 8 cores and 16GM RAM. The machines are divided into 5 slots.

Slots[1-4]: 1 Core, 2GB RAM
Slot5: 4 Cores, 8 GB RAM

If a job needs to run on slot 5, a request for RAM > 4GB would need to be included in the submit script. For example -

request_memory = 5000

Submit Node

All jobs are to be submiited from afsaccess1.njit.edu.

HTCondor and AFS

The user manual has a section called "AFS and HTCondor for Users" with specific instructions on using HTCcondor and AFS.

To allow read and write access to the HTCondor processes, the AFS group 'condor_pool' needs the correct permissions. Users should create a separate directory for use with HTCondor. For example --

mkdir condor
fs sa condor condor_pool write

Alternatively, files can be copied to /scratch on afsaccess1.njit.edu and jobs submitted from there.