A0: Computing Resources Warmup

In this assignment, you will set up accounts at the Notre Dame CRC and Purdue Anvil, then submit one small batch job at each site. The jobs themselves are deliberately simple. The point is to get account and scheduler problems out of the way before we begin larger assignments.

Complete the first-time setup and read the general instructions before starting. Start both account requests early because provisioning may take a few days.

Part A: Notre Dame CRC

1. Request a CRC account

Follow Request a CRC account and wait for CRC to confirm that your account is ready.

2. Log in to the HTCondor front end

Review the network and VPN requirements in Using the Notre Dame CRC. Then replace NETID with your Notre Dame NetID and log in to the HTCondor front end:

ssh NETID@condorfe.crc.nd.edu

The “front end” machine is a normal Linux machine, capable of running ordinary applications. (Experiment with the w and ps and top commands to view the large number of people and processes running.)

It’s ok to execute short-running programs here (seconds to minutes), but it should not be used to run programs that are long-running, memory intensive, or highly multi-threaded. Doing so will disrupt the many other people logged in.

The purpose of the front-end machines is to submit jobs to the cluster, where they will be dispatched to available machines, and then can perform intensive tasks without interference.

3. Submit an HTCondor job

So, let’s submit a simple job to run on a different machine in the cluster.

Get in the habit of creating a fresh directory for each assignment/task/activity:

mkdir -p warmup
cd warmup

Here, create a file called crc-hello.submit that contains this HTCondor submit file:

universe = vanilla
executable = /bin/hostname
output = crc-hello.out
error = crc-hello.err
log = crc-hello.log

request_cpus = 1
request_memory = 128MB
request_disk = 10MB

should_transfer_files = yes
when_to_transfer_output = on_exit

queue 1

This is a set of instructions to run one application (/bin/hostname) on the cluster, instead of on the front end machine. Submit the job like this:

condor_submit crc-hello.submit

Remember the job ID printed by condor_submit, such as 12345.0. To see whether it is still in the queue, replace NETID with your NetID:

condor_q NETID

This job may finish before you see it in condor_q. Once it is gone from the queue, examine the three files it produced:

cat crc-hello.out
cat crc-hello.err
tail -n 12 crc-hello.log

The job succeeded if crc-hello.out contains the name of an execute machine, crc-hello.err is empty, and the end of crc-hello.log reports that the job terminated with exit code 0.

Explore Further

Try these commands and explore various options to view the state of the cluster and its users.

condor_q
condor_status
condor_userprio
condor_matrix

Part B: ACCESS and Purdue Anvil

1. Log in to Anvil

Follow Use Purdue Anvil to open an Anvil shell. Confirm that mybalance lists the course CPU account cis261613 before continuing.

2. Submit a Slurm job

Use Slurm to run the test job on a compute node.

Again, get in the habit of creating a directory for each new activity:

mkdir -p warmup
cd warmup

Create a file called anvil-hello.slurm containing instructions to run a job on the cluster:

#!/bin/bash
#SBATCH --account=cis261613
#SBATCH --partition=shared
#SBATCH --nodes=1
#SBATCH --ntasks=1
#SBATCH --cpus-per-task=1
#SBATCH --mem=128M
#SBATCH --time=00:01:00
#SBATCH --job-name=hpds-warmup
#SBATCH --output=anvil-%j.out

echo "system=anvil"
echo "job_id=${SLURM_JOB_ID}"
echo "user=$(whoami)"
echo "compute_host=$(hostname)"
date -u '+utc_time=%Y-%m-%dT%H:%M:%SZ'

Submit the job like this:

sbatch anvil-hello.slurm

sbatch will print a numerical job ID. Remember it. You can check the job while it is waiting or running with:

squeue --me

After the job leaves the queue, replace JOBID below with your job ID:

cat anvil-JOBID.out
seff JOBID

The job succeeded if anvil-JOBID.out names a compute host and seff reports State: COMPLETED with exit code 0.

Turning In

Create your repository as described in the first time setup.

In that repository, create a warmup directory, and include the following:

crc-hello.submit, crc-hello.out, crc-hello.err, and crc-hello.log from the CRC job.
anvil-hello.slurm and anvil-JOBID.out from the Anvil job.
A README.md that records both job IDs, describes the outcomes, and explains what you learned from Explore Further.

Make sure you have committed and pushed all your work.

Submit your work in Canvas by pasting the repository URL into the assignment submission page.