Skip to content

Monitoring and Managing Jobs

Open OnDemand provides a browser-based view of jobs submitted to VHPC through Slurm.

The Active Jobs page shows jobs submitted through Open OnDemand as well as jobs submitted from a shell using commands such as sbatch.

Open Active Jobs

From the Open OnDemand navigation bar:

  1. Open Jobs.
  2. Select Active Jobs.

The page displays jobs that are currently queued or running.

The page may also briefly show jobs that are completing.

Find a Job

The Active Jobs table includes columns such as:

  • ID
  • Name
  • User
  • Account
  • Time Used
  • Queue
  • Status
  • Cluster
  • Actions

Use the filters near the top of the page to narrow the list by job, cluster, or search text.

The numeric ID is the Slurm job ID. Include it when requesting help with a specific job.

Understand the Job Status

The most common states shown on this page are:

Status Meaning
Queued The job is waiting for Slurm to start it.
Running The job is currently running on a VHPC compute node.

A queued job is not necessarily experiencing a problem. It may be waiting for CPUs, memory, GPUs, scheduler priority, or another requirement.

View Job Details

Select the + beside a job to expand its details.

The expanded panel may show:

  • Job ID and name
  • Slurm account
  • Partition
  • Current state
  • Pending reason
  • Requested CPU cores
  • Requested memory
  • Time limit
  • Time used
  • Submission time
  • Start and estimated end times
  • Assigned compute node
  • Output file location

The fields shown depend on the current state of the job.

Queued Jobs

A queued job normally shows:

  • State: PENDING
  • A Reason explaining why it has not started
  • Requested resources and time limit
  • No assigned node or start time yet

The pending reason is often the most useful field when deciding whether the job simply needs more time or requires attention.

Running Jobs

A running job normally shows:

  • State: RUNNING
  • The assigned node
  • Elapsed time
  • Start time
  • Estimated end time
  • Requested CPUs, memory, and other resources

The Active Jobs page shows whether the job is running, but it does not necessarily show application progress. Review the job’s output or log files when you need that information.

Common Reasons a Job Remains Queued

A job may remain queued because:

  • The requested CPUs or memory are not yet available
  • A requested GPU is not available
  • The selected partition is busy
  • The job is waiting behind higher-priority work
  • A project or account limit applies
  • A dependency has not been satisfied
  • The request cannot currently fit on an available node

Do not submit repeated copies of the same job merely because the first one is still queued.

If the resource request is larger than necessary, cancelling and resubmitting with a smaller valid request may allow it to start sooner.

Jobs from Interactive Applications

Interactive applications such as Jupyter, MATLAB, GaussView, and the VHPC desktop run as Slurm jobs.

Their jobs appear on the Active Jobs page after the session is submitted.

Use:

  • My Interactive Sessions to connect to and manage the application
  • Active Jobs to inspect the associated Slurm job

Warning

Cancelling the Slurm job immediately ends the interactive application. Unsaved work may be lost.

For more information, see Running Interactive Applications.

Cancel a Job

To cancel a queued or running job:

  1. Locate the correct job.
  2. Verify its job ID and name.
  3. Use the available action in the Actions column.
  4. Confirm the cancellation.

The job may briefly remain visible while Slurm completes cleanup.

Warning

Cancelling a running job terminates its processes. Work not already written to persistent storage may be lost.

Interactive sessions can also be cancelled from My Interactive Sessions, which is usually the preferred method for those applications.

Completed Jobs

The Active Jobs page is intended primarily for currently queued and running jobs.

After a job finishes, it may disappear from the page quickly.

Use Slurm accounting tools when you need information about a completed or older job. From the web-based shell, for example:

sacct -j job-id

Replace job-id with the numeric Slurm job ID.

Detailed Slurm accounting and performance analysis are covered in the VHPC Slurm documentation.

Review Job Output

Batch jobs commonly write output and error messages to files in the directory from which the job was submitted.

The exact file names depend on the Slurm submission script.

You can review those files using:

Output and error files are often the best place to begin when a job fails or produces unexpected results.

Refresh the Page

Job states can change quickly from queued to running or from running to completed.

Refresh the page when the displayed state appears outdated.

A very short job may finish before you see it on Active Jobs.

Job Composer and Project Manager

The Jobs menu also includes Job Composer and Project Manager.

These tools support additional job-submission and project workflows. They are not covered in this introductory guide.

Troubleshooting

My job does not appear

Confirm that:

  • The job was submitted successfully
  • You are viewing Jobs → Active Jobs
  • The job has not already completed
  • The page has been refreshed

A short job may finish before it appears.

Use sacct from the web-based shell if you know the job ID:

sacct -j job-id

My job remains queued

Expand the job details and review the Reason field.

The job may be waiting for resources, scheduler priority, a dependency, or an account or partition limit.

A queued state by itself does not mean the job has failed.

My job failed

Review the job’s output and error files.

Common causes include:

  • Reaching the requested time limit
  • Exceeding the memory allocation
  • Referring to a missing file or directory
  • Using an unavailable command or software module
  • An application-specific error

My job disappeared

The job may have completed, failed, timed out, or been cancelled.

Use Slurm accounting to review its final state:

sacct -j job-id

I cannot cancel a job

Refresh the page and confirm that the job is still active.

It may already be completing or may have ended between page updates.

An interactive application stopped unexpectedly

Check both My Interactive Sessions and Active Jobs.

The associated job may have:

  • Reached its time limit
  • Exceeded its memory allocation
  • Been cancelled
  • Encountered an application error
  • Experienced a compute-node problem

Next Steps

Getting Help

If a job remains queued unexpectedly, fails without a clear explanation, or cannot be cancelled, contact the Kennesaw State University Service Desk.

Method Students Faculty/Staff
Web ServiceNow ServiceNow
E-Mail studenthelpdesk@kennesaw.edu service@kennesaw.edu
Phone 470-578-3555 470-578-6999

When reporting a problem, include:

  • Your NetID
  • The Slurm job ID
  • The approximate date and time of the problem
  • The job name
  • The requested resources
  • The current or final job state
  • Any pending reason or visible error message
  • Relevant output or error file names