Usage

Proper usage of cluster

Policies

1. master node is to be used for editing jobs and non-intensive data preparation with unix tools such as awk or grep
2. nodes are to be used only via the PBS batch processing system, using qsub
3. only queues serial (and ?x) may be used by user-created PBS batch scripts
4. qsub -I may be used to start an interactive session on a node, primarily to test jobs that will be submitted via qsub
5. there exist various helper scripts (see below) such as qR which submit jobs via qsub and are therefore encouraged
6. certain queues are for special purposes are are not to be used except via via qsub or one of the helper scripts listed below
7. ssh to nodes for the purposes of subverting the batch system to run jobs is forbidden; repeat offenders may be denied system access privileges for a period decided by the IT committee
8. users have limits on disk space and node usage (exceptions to these limits will be considered by the IT committee)
9. appeals to overturn cluster committee decisions may be submitted for consideration at regular faculty meetings
10. requests to install new software on the cluster should be made by email to Helen or Joey and entered in the VIPBG IT ticket system

Submitting jobs

A helper script is available for submitting jobs is available for many of the common languages.

R

qR your_r_script.R

MX

qmx your_mx_script

SAS

qsas your_sas_script

Open MX

qRh your_openmx_script.R

All other open resources

qsub your_own_script

To find out more information about how to use PBSPro and submit jobs, check out the user manual here.

Checking jobs

Use qstat -q to check queues on cluster, which will print something like this:

server: light

Queue            Memory CPU Time Walltime Node   Run   Que   Lm  State
---------------- ------ -------- -------- ---- ----- ----- ----  -----
workq              --      --       --     --      4     0   --   E R
serial             --      --       --     --     34     0   --   E R
mxq                --      --       --     --      0     0   --   E R
express            --      --       --     --      0     0   --   E R
openmx             --      --       --     --      0     0   --   E R
slowq              --      --       --     --      0     0   --   E R
                                               ----- -----
                                                  38     0

The Run column indicates there are 34 jobs currently running on the serial queue.

Nodes and queues configuration

Nodes 2-20 queues on cluster:

Queue	Purpose
serial	default
workq	running converge
mxq	running old version mx jobs
openmx	running openmx and parallel jobs
express	for short and quick jobs not wait in the queue
slowq	slow and test jobs on old nodes