at Command: One-Time Job Scheduling on Linux

2026-06-05 Reading time: About 8 min Difficulty: Intermediate

What You'll Learn

• How to schedule a one-time job with at
• How to list and cancel pending jobs with atq / atrm
• When to reach for at instead of cron

What Is the at Command?

Conclusion: at runs a command once at a specified time. It is built for single, non-repeating tasks.

Where cron handles recurring jobs ("every day", "every week"), at handles one-off scheduled execution — "once tonight at 11 PM" or "once, 5 minutes from now".

A scheduled job is watched by atd (the at daemon), which runs it when the time arrives. The job survives logging out or closing your shell.

Is at Ready to Use? Checking and Installing atd

Conclusion: Having the at binary is not enough — jobs only run if atd is active. Check the daemon first.

On many Ubuntu / Debian systems at is not installed by default. Install and verify it:

# Debian / Ubuntu
$ sudo apt install at

# RHEL / CentOS / Rocky
$ sudo dnf install at

Check the daemon status:

$ systemctl status atd

● atd.service - Deferred execution scheduler
     Loaded: loaded (/lib/systemd/system/atd.service; enabled; preset: enabled)
     Active: active (running) since ...

If it is not active (running), start and enable it:

$ sudo systemctl enable --now atd

How Do You Schedule a Job?

Conclusion: Run at TIME, type commands at the prompt, and finish with Ctrl+D. For scripts, pass them with -f.

Interactive entry

$ at now + 5 minutes
warning: commands will be executed using /bin/sh
at> echo "hello" >> /tmp/at-test.log
at> <EOT>
job 3 at Fri Jun  5 23:10:00 2026

Type commands at the at> prompt, then press Ctrl+D (shown as <EOT>). On success you get a job number and the scheduled time.

From a file (recommended)

Interactive input is error-prone. In practice, pass a script with -f:

$ at -f /path/to/job.sh 23:00

# Pipe a one-liner in
$ echo 'tar czf /backup/data.tgz /var/www' | at 02:00 tomorrow

How Do You Write the Time Spec?

Conclusion: You can use HH:MM, now + N units, midnight / noon / teatime, and explicit dates. Relative specs are the most practical.

at accepts flexible time expressions. Common patterns:

# Example: rotate logs at 3 AM tomorrow
$ echo '/usr/local/bin/rotate-logs.sh' | at 3:00 tomorrow

How Do You List and Cancel Jobs?

Conclusion: List with atq, remove with atrm JOBID, and inspect with at -c JOBID. These three cover job management.

List pending jobs (atq)

$ atq

3	Fri Jun  5 23:10:00 2026 a user
5	Sat Jun  6 02:00:00 2026 a user

Columns are job number / scheduled time / queue / owner. a is the normal queue; b is the batch queue.

Remove a job (atrm)

$ atrm 3

Delete by number. You can remove several at once:

$ atrm 3 5

Inspect a job (at -c)

To see exactly what is scheduled before it runs, use -c:

$ at -c 5

It prints the saved environment-restoration block and the actual command body.

Where Does the Output Go?

Conclusion: A job's stdout and stderr are sent to the user by local mail. Without a mailer, the output is lost — redirect it to a file.

Output from an at job does not go to your terminal; it goes to local mail for the submitting user. Read it with mail or under /var/mail/<user>.

On a server with no MTA (mail delivery), the output is simply discarded. To keep results reliably, redirect inside the job:

$ echo '/usr/local/bin/backup.sh > /var/log/backup.log 2>&1' | at 02:00

When Should You Use at Instead of cron?

Conclusion: Use cron / systemd timer for recurring work and at for one-time work. "Just tonight" or "once after maintenance" fits at best.

Writing a one-off task into cron and forgetting to remove it — so it fires again the next day — is a common mishap. When you know it is single-use, at needs no cleanup and is safer.

For recurring jobs in general, see cron Basics and systemd Timer vs cron.

Access Control and Caveats

Conclusion: Access to at is controlled by /etc/at.allow and /etc/at.deny. If at.allow exists, it takes precedence.

• If /etc/at.allow exists: only listed users may use at
• If at.allow is absent but /etc/at.deny exists: everyone except listed users may use it
• If neither exists: many distros allow only root (behavior varies by distro)

# Allow a specific user only
$ echo 'deploy' | sudo tee -a /etc/at.allow

Summary

• at schedules a one-time job; recurring work belongs to cron
• Jobs only run if atd is active — check the daemon first when nothing happens
• Prefer relative specs like now + N minutes; a past time rolls to the next day
• Manage jobs with atq (list) / atrm (remove) / at -c (inspect)
• Output goes to mail — redirect production jobs to a log file