If you want to schedule a Linux job that will happen only once,
cron is overkill. The
at family of commands is what you need! And if you want to run processes only when your system has free resources, you can use
How to Schedule Linux Jobs
cron daemon maintains a list of jobs it runs at specific times. These tasks and programs run in the background at the scheduled times. This offers you great flexibility for scheduling tasks that need to be repeated. Whether you need to run a task once every hour, at a specific time each day, or once a month or year, you can set it up in
However, this doesn’t help if you want to schedule a task to run just once. Sure, you can set it up in
cron, but then you have to remember to go back and remove the crontab entry after the task executes, which is inconvenient.
With Linux, if you’re struggling with a problem, it’s almost a guarantee someone else has struggled with it, too. Fortunately, because Unix-like operating systems have been around so long, there’s also an excellent chance someone has created a solution to your problem.
For the problem outlined above, they have, and it’s called
Installing the at Command
We had to install
at on Ubuntu 18.04 and Manjaro 18.1.0 (it was already installed on Fedora 31).
at on Ubuntu, use this command:
sudo apt-get install at
After the installation completes, you can start the
at daemon with this command:
sudo systemctl enable --now atd.service
On Manjaro, you install
at with this command:
sudo pacman -Sy at
After the installation completes, type this command to start the
sudo systemctl enable --now atd.service
On any distribution, you can type this command to make sure the
atd daemon is running:
ps -e | grep atd
How to Use the at Command Interactively
at, you have to assign it a date and time to run. There’s a great deal of flexibility in the way you can write these, which we cover later in this article.
However, even though we’re going to use
at interactively, you have to provide the date and time upfront. If you don’t include anything on the command line, or you type something that isn’t a date and time,
at responds with “Garbled time,” as shown below:
Dates and times can be explicit or relative. For example, let’s say you want to have a command execute one minute from now.
at knows what “now” means, so you can use
now and add one minute to it, like so:
at now + 1 minute
at prints out a message and an
at prompt, and waits for you to type the commands you want to schedule. First, though, consider the message, as shown below:
It tells you
at launches an instance of the
sh shell and will run the commands inside that. Your commands won’t be executed in the Bash shell, which is compatible with the
sh shell but has a richer feature set.
If your commands or scripts try to use a function or facility that Bash provides, but
sh doesn’t, they’ll fail.
It’s easy to test whether your commands or scripts will run in
sh. Use the
sh command to start an
The command prompt changes to a dollar sign (
$), and you can now run your commands and verify that they operate correctly.
To return to the Bash shell, type the
You won’t see any standard output or error messages from the commands. This is because the
sh shell launches as a background task and runs without any sort of screen interface.
Any output from the commands—good or bad—is emailed to you. It’s sent via the internal mail system to whomever runs the
at command. This means you have to set up and configure that internal email system.
Many (most) Linux systems don’t have an internal email system as there’s rarely a need for one. Those that do typically use a system like sendmail or postfix. If your system doesn’t have an internal email system, you can have scripts write to files or redirect output to files to add logging.
If the command doesn’t generate any standard output or error messages, you won’t get an email, anyway. Many Linux commands indicate success via silence, so in most cases, you won’t get an email.
Now, it’s time to type a command in
at. For this example, we’ll use a small script file called
sweep.sh that deletes the
*.o files. Type the path to the command, as shown below, and then press Enter.
Another command prompt appears, and you can add as many commands as you like. It’s usually more convenient to have your commands in a single script and simply call that script from within
Press Ctrl+D to tell
at you’re finished adding commands.
at shows <EOT>, which means end of transmission. You’re told the job number and when the job is scheduled to run, as shown below:
After the job executes, type the following to check your internal mail:
If there’s no mail, you have to assume success. Of course, in this case, you can check and see if the
*.o files were deleted to confirm the command worked.
Type the following to run the whole thing again:
at now + 1 minute
After one minute, type the following to recheck your mail:
Hey, we’ve got mail! To read message number one, press 1, and then hit Enter.
We received an email from
at because the commands in the script generated error messages. In this example, there were no files to delete because when we ran the script previously, it removed them.
Press D+Enter to delete the email and Q+Enter to quit the mail program.
Date and Time Formats
You have a lot of flexibility when it comes to the time formats you can use with
at. Here are a few examples:
- Run at 11:00 a.m.:
at 11:00 AM
- Run at 11:00 a.m. tomorrow:
at 11:00 AM tomorrow
- Run at 11:00 a.m. on this day next week:
at 11:00 AM next week
- Run at this time, on this day, next week:
at next week
- Run at 11:00 a.m. next Friday:
at 11:00 AM next fri
- Run at this time next Friday:
at next fri
- Run at 11:00 a.m. on this date, next month:
at 11:00 AM next month
- Run at 11:00 a.m. on a specific date:
at 11:00 AM 3/15/2020
- Run 30 minutes from now:
at now + 30 minutes
- Run two hours from now:
at now + 2 hours
- Run at this time tomorrow:
- Run at this time on Thursday:
- Run at 12:00 a.m.:
- Run at 12:00 p.m.:
- If you’re a Brit, you can even schedule a command to run at teatime (4 p.m.):
Looking at the Job Queue
You can type the
atq command to see the queue of scheduled jobs, as shown below.
For each command in the queue,
atq displays the following information:
- Job ID
- Scheduled date
- Scheduled time
- Queue the job is in. The queues are labeled “a,” “b,” and so on. Normal tasks you schedule with
atgo into queue “a,” while tasks you schedule with
batch(covered later in this article) go into queue “b.”
- The person who scheduled the job.
Using at on the Command Line
You don’t have to use
at interactively; you can also use it on the command. This makes it easier to use inside scripts.
You can pipe commands into
at, like this:
echo "sh ~/sweep.sh" | at 08:45 AM
The job is accepted and scheduled by
at, and the job number and execution date are reported just as before.
RELATED: How to Use Pipes on Linux
Using at with Files of Commands
You can also store a sequence of commands in a file, and then pass it to
at. This can be a plain text file of commands—it doesn’t have to be an executable script.
You can use the
-f (file) option in the following way to pass a filename to
at now + 5 minutes -f clean.txt
You can achieve the same result if you redirect the file into
at now + 5 minutes < clean.txt
Removing Scheduled Jobs from the Queue
To remove a scheduled job from the queue, you can use the
atrm command. If you want to see the queue first to find the number of the job you want to remove, you can use
atq . Then, use that job number with
atrm, as shown below:
How to See a Detailed View of Jobs
As we mentioned previously, you can schedule jobs far into the future. Sometimes, you might forget what a job is going to do. The
atq command shows you the jobs in the queue, but not what they’re going to do. If you want to see a detailed view of a job, you can use the
-c (cat) option.
First, we’ll use
atq to find the job number:
Now, we’ll use job number 13 with the
at -c 13
Here’s a breakdown of the information we get back about the job:
- First line: This tells us the commands will run under the
- Second line: We see the commands will run with both a user and group ID of 1000. These are the values for the person who ran the
- Third line: The person who receives any emails
- Fourth line: The User Mask is 22. This is the mask used to set the default permissions for any files created in this
shsession. The mask is subtracted from 666, which gives us 644 (the octal equivalent of
- Remaining data: The majority are environment variables.
- Results of a test. A test checks to make sure the execution directory can be accessed. If it cannot, an error is raised, and the job execution is abandoned.
- The commands to be executed. These are listed, and the contents of the scripts that are scheduled are displayed. Note that although the script in our example above was written to run under Bash, it will still be executed in an
The batch Command
batch command operates similarly to the
at command, but with three significant differences:
- You can only use the
- Rather than scheduling jobs to execute at a specific time, you add them to the queue, and the
batchcommand executes them when the system’s average load is lower than 1.5.
- Due to the above, you never specify a date and time with the
When you use the
batch command, you call it by name with no command line parameters like so:
Next, add tasks just as you would with the
Controlling Access to the at Command
at.deny files control who can use the
at family of commands. These are located within the
/etc directory. By default, only the
at.deny file exists, and it’s created when
at is installed.
Here’s how these work:
at.deny: Lists applications and entities that cannot use
atto schedule jobs.
at.allow: Lists who can use
atto schedule jobs. If the
at.allowfile doesn’t exist,
atonly uses the
By default, anyone can use
at. If you want to restrict who can use it, use the
at.allow file to list those who can. This is easier than adding everyone who cannot use
at to the
Here’s what the
at.deny file looks like:
sudo less /etc/at.deny
The file lists components of the operating system that cannot use
at. Many of these are prevented from doing so for security reasons, so you don’t want to remove any from the file.
Now, we’ll edit the
at.allow file. We’re going to add
mary, but no one else will be allowed to use
First, we type the following:
sudo gedit /etc/at.allow
In the editor, we add the two names, as shown below, and then save the file.
If anyone else tries to use
at, he’ll be told he doesn’t have permission. For example, let’s say a user named
eric types the following:
He would be refused, as shown below.
eric is not in the
at.deny file. As soon as you put anyone in the
at.allow file, any- and everyone else is denied permission to use
Great for One-Offs
As you can see, both
batch are ideal for tasks you only need to run once. Again, as a quick review:
- When you need to do something that isn’t a regular process, schedule it with
- If you want to run a task only when the system load is low enough, use