Submitting jobs¶
There are two ways you can run your jobs, namely submitting a Slurm script and running a job interactively. Either way, jobs are submitted to CARC by the command sbatch. For more information on available options type man sbatch.
Submitting the Slurm Script to the Batch Scheduler¶
In order to run our simple Slurm script, we will need to submit it to the batch scheduler using the command sbatch followed by the name of the script we would like to run. For more information please see our page on writing a Slurm batch script.
In the following example, we submit our simple hello.sbatch script to the batch scheduler using sbatch. Note that it returns the job identifier when the job is successfully submitted. You can use this job identifier to query the status of your job from your shell.
For example:
Interactive Slurm Jobs¶
Normally a job is submitted for execution on a cluster or supercomputer using the command sbatch script.sbatch. CARC recommends that all jobs are submitted this way as job submission fails if there are errors in resources requested. However, at times, such as when debugging, it can be useful to run a job interactively. To run a job in this way type salloc followed by resources requested, and the batch manager will log you into a node where you can directly run your code.
Here's our simple helloworld_parallel.sbatch script, which prints a hello message from every task it's given:
export THIS_HOST=$(hostname)
echo "Job $SLURM_JOB_ID running on $THIS_HOST"
echo "Hello World from host $THIS_HOST"
To actually run it across multiple tasks, it needs to be launched with mpirun. Just running it with plain bash would only ever execute it once, regardless of how many tasks you requested with salloc.
Here's an interactive session running it across all 8 tasks of a node:
Job 1035500 running on easley031
Hello World from host easley031
Job 1035500 running on easley031
Hello World from host easley031
Job 1035500 running on easley031
Hello World from host easley031
Job 1035500 running on easley031
Hello World from host easley031
Job 1035500 running on easley031
Hello World from host easley031
Job 1035500 running on easley031
Hello World from host easley031
Job 1035500 running on easley031
Hello World from host easley031
Job 1035500 running on easley031
Hello World from host easley031
Three commands were executed here. The first,
asked the batch manager to provide one node of easley with all 8 of that node's cores for use. It is good practice to request all available processors on a node to avoid multiple users being assigned to the same node. The walltime was specified as 5 minutes, since this was a simple code that would execute quickly. The second command,
loaded the openMPI software module, giving us the MPI libraries and the mpirun launcher needed to actually run the script across all 8 tasks. The third command,
ran the helloworld_parallel.sbatch script across all 8 requested tasks, using an absolute path since mpirun doesn't necessarily preserve your current directory across every task.
This quickbyte was validated on 7/30/2026
Migrated from UNM-CARC QuickBytes (last source update 2026-07-30). Spotted a problem? Open an issue or pull request.
Machine-readable versions of this page: Markdown twin · raw source on GitHub · llms.txt · llms-full.txt (whole site). See For AI agents.