Tutorials

• Follow instructions in the following page.

https://www.rc.colorado.edu/crcdocs/start

• SSH to the research computing machine.

ssh -Y userId@login.rc.colorado.edu

• Load Torque modules

use Torque

• Print all nodes

pbsnodes -a

• Node resource options

man pbs_resources

• sample script (helloworld.sh), submit via qsub

#!/bin/bash
# Lines starting with #PBS are treated by bash as comments, but interpreted by qsub
# as arguments.  For more details about usage of these arguments see "man qsub"

#
# Name the job.

#PBS -N helloworld

#
# Set a walltime for the job. The time format is HH:MM:SS

# Run for 30 seconds:
#PBS -l walltime=0:00:30

#
# Select one node, and only 1 processors per node
#you can see what properties are available
# with the "pbsnodes -a" command which will list all nodes and their properties

#PBS -l nodes=1:ppn=1

# # Join the Output and Errors in one file. you can also set the # path for this output.
# (see "man qsub" for details.)

#PBS -j oe

#
# Source the Dotkit init so you can pull in extra software via the "reuse" command:

. /curc/tools/utils/dkinit

#
# cd to the jobs working directory, which you can set above with a #PBS # directive,
# (see "man qsub" for details)

cd /home/shumin

#
# Execute the program.

use Java

java HelloWorld > hello.txt

# This script needs to be submitted via qsub to run on the cluster.

• qsub

qsub $your_script -q $queue_name

See https://www.rc.colorado.edu/crcdocs/queues for description of queues. "crc-serial" is the blade queue. There are 16 total blade nodes, each node w/ 24 virtual cores and 96GB memory.

• More advanced job queuing:

https://www.rc.colorado.edu/crcdocs/advanced