8 minute read

Cron is the Linux scheduler that runs commands at set times. It is simple, but does have its tricks. This guide gives you copy-paste examples to get you started.

Index

  1. Cron syntax: every 15 minutes and weekdays at 9 AM
  2. Works in terminal, fails in cron
  3. Basic setup: editor, logs, permissions
  4. Stop overlapping cronjob runs with flock
  5. Last day of the month and first Thursday

1. Cron syntax: every 15 minutes and weekdays at 9 AM

A crontab line has 5 time fields, then the command:

┌───────── minute        (0-59)
│ ┌─────── hour          (0-23)
│ │ ┌───── day of month  (1-31)
│ │ │ ┌─── month         (1-12)
│ │ │ │ ┌─ day of week   (0-7, 0 and 7 = Sunday)
│ │ │ │ │
* * * * *  command-to-run

Special characters:

Char Meaning Example
* every value * = every minute
*/n every n-th value */15 = every 15 min
a-b range 1-5 = Mon to Fri
a,b list 1,15 = the 1st and 15th

Example 1: run a script every 15 minutes

*/15 * * * * /home/dev/scripts/sync.sh

Runs at :00, :15, :30 and :45 of every hour.

Example 2: run a script on weekdays at 9 AM

0 9 * * 1-5 /home/dev/scripts/report.sh

Minute 0, hour 9, any day or month, Monday (1) to Friday (5).

Try it in 2 minutes

Add this with crontab -e:

* * * * * date >> /tmp/cron-test.log

Wait two minutes, then check:

cat /tmp/cron-test.log

You should see one timestamp per minute.

Tip: cron uses the server’s timezone. Check it with date. In containers it is usually UTC. To sanity-check an expression, paste it into crontab.guru.


2. Works in terminal, fails in cron

Usual way to develop a chron job is to run a test a command in a terminal and paste it in the crontab. The cronjob may fail:

Why: cron runs jobs in a minimal environment, not your login shell.

Your terminal Cron
Full $PATH from .bashrc / .profile PATH=/usr/bin:/bin only
Your exported variables None of them
Current directory = where you are Working directory depends on the cron implementation
Uses bash Uses /bin/sh by default

Reproducible failing example

Step 1. Add to ~/.bashrc:

export PATH="$HOME/bin:$PATH"
export APP_ENV="production"

Step 2. Create a custom tool, a script, and a data file:

mkdir -p ~/bin ~/data
printf '#!/bin/bash\necho "hello-tool ran"\n' > ~/bin/hello-tool
chmod +x ~/bin/hello-tool
echo "some input" > ~/data/input.txt

Step 3. Create ~/report.sh:

#!/bin/bash
hello-tool
echo "APP_ENV=$APP_ENV"
cat data/input.txt
chmod +x ~/report.sh

Run it from ~ in your terminal:

hello-tool ran
APP_ENV=production
some input

Step 4. Schedule it, replacing dev with your username:

* * * * * /home/dev/report.sh >> /home/dev/report.log 2>&1

Step 5. Check the log after a minute (cat ~/report.log). Example output:

/home/dev/report.sh: line 2: hello-tool: command not found
APP_ENV=
cat: data/input.txt: No such file or directory

(This is typical output. Exact wording varies slightly by distro.)

What went wrong

Log line Cause
hello-tool: command not found ~/bin was added to PATH in .bashrc, which cron never reads
APP_ENV= (empty) variable exported in .bashrc, not visible to cron
data/input.txt: No such file relative path; cron’s working directory is not where you tested

The fix

Make the script self-contained (~/report.sh). Important is to include env variables (on top).

#!/bin/bash
export PATH="$HOME/bin:/usr/local/bin:/usr/bin:/bin"
export APP_ENV="production"
cd "$(dirname "$0")" || exit 1   # run from the script's own folder

hello-tool
echo "APP_ENV=$APP_ENV"
cat data/input.txt

Or set variables at the top of the crontab (they apply to all jobs below):

PATH=/home/dev/bin:/usr/local/bin:/usr/bin:/bin
APP_ENV=production

* * * * * cd /home/dev && /home/dev/report.sh >> /home/dev/report.log 2>&1

Debugging trick: see cron’s real environment

* * * * * env > /tmp/cron-env.txt

Then compare:

diff <(sort /tmp/cron-env.txt) <(env | sort)

Checklist: use absolute paths, set PATH explicitly, never rely on .bashrc, and always redirect output with >> file 2>&1.


3. Basic setup: editor, logs and permissions

Edit your crontab

crontab -e     # edit
crontab -l     # list
crontab -r     # remove ALL (no confirmation!)

Before testing an entry, make sure cron is running:

# Debian / Ubuntu
sudo systemctl enable --now cron
systemctl status cron

# RHEL / Rocky / Fedora
sudo systemctl enable --now crond
systemctl status crond

After saving, confirm the entry with crontab -l. crontab -r removes only the current user’s crontab and asks for no confirmation, so use it carefully.

If a crontab is saved with invalid syntax, the cron daemon will usually ignore it rather than running only the valid-looking entries. Remove or fix the invalid line, then confirm the complete file again with crontab -l.

How to exit the editor

Editor Save and quit Quit without saving
vim Esc, then :wq, Enter Esc, then :q!, Enter
nano Ctrl+O, Enter, then Ctrl+X Ctrl+X, then N

Prefer nano? Pick it for one session:

EDITOR=nano crontab -e

Or run select-editor on Debian/Ubuntu to set it permanently.

Where do the logs of cronjob go?

Cron logs that a job started. It does not log the job’s own output.

# Debian / Ubuntu
grep CRON /var/log/syslog
journalctl -u cron

# RHEL / Rocky / Fedora
sudo tail /var/log/cron
journalctl -u crond

Sample line:

Nov 19 09:00:01 server CRON[1234]: (dev) CMD (/home/dev/scripts/report.sh)

For the script’s own output, redirect it yourself to display logs in a file:

0 9 * * 1-5 /home/dev/scripts/report.sh >> /home/dev/report.log 2>&1

the >> will always overwrite the contnet of the file. if you like to concatenate use single >.

Permissions

chmod +x /home/dev/scripts/report.sh
ls -l /home/dev/scripts/report.sh   # look for the "x" bits

Output:

drwxr-xr-x@ 3 ubuntu  staff    96 14 Nov 21:54 report.sh

Also check:

  • The script has a shebang on line 1: #!/bin/bash
  • The user running the job can write to the log file’s folder
  • For beginner examples, prefer a log inside the user’s home directory, such as /home/dev/report.log

4. Stop overlapping cronjob runs with flock

Flock is able to manipulate locks on files and thus allowing/preventing the run of a job. Locks are used as a semaphor to signal when job can run.

Comon problem: a job runs every minute but sometimes takes longer to finish. Cron starts the script and they pile up.

Solution: flock takes a lock file. If the lock is held, the new run exits immediately.

Lock the file for the duration of the run:

* * * * * flock -n /tmp/slow-job.lock /home/dev/scripts/slow.sh >> /home/dev/slow.log 2>&1
  • -n = don’t wait; exit at once if the lock is taken
  • The lock is released automatically when the process ends, even after a crash (unlike PID files)

Try it

Create ~/scripts/slow.sh:

#!/bin/bash
echo "$(date +%T) start"
sleep 90
echo "$(date +%T) done"
chmod +x ~/scripts/slow.sh

How install the cronjob mentioned above using crontab -e.

Without flock, cat /home/dev/slow.log shows overlapping starts:

10:00:01 start
10:01:01 start      <- second copy while the first is still running
10:01:31 done
10:02:01 start
10:02:31 done

With flock, the minute-1 run is skipped because the lock is still held:

10:00:01 start
10:01:31 done
10:02:01 start
10:03:31 done

Test the lock by hand (no cron needed)

# Terminal 1
flock -n /tmp/demo.lock sleep 30

# Terminal 2, while terminal 1 is running
flock -n /tmp/demo.lock echo "got lock"; echo "exit code: $?"
# -> exit code: 1  (lock busy, command did not run)

Comand variants:

flock -w 10 /tmp/job.lock ./job.sh   # wait up to 10s, then give up
flock -n -E 0 /tmp/job.lock ./job.sh # exit 0 (not 1) when skipped, quiet for monitoring

flock is part of util-linux (preinstalled on most Linux distros). On macOS it is not available by default, use brew.


5. Complex cronjob schedules: last day of the month, first Thursday

Standard 5-field cron cannot express these. The simple, practical fix: schedule a broader window, then check the date in the command.

Two things to know first

1. % must be escaped as \% in crontab, because cron treats a bare % as a newline.

2. The day-of-month / day-of-week “OR” problem. If you restrict both fields, cron runs when either matches:

# Looks like "first Thursday", but runs on every day 1-7
# OR every Thursday
0 9 1-7 * 4 /path/job.sh

Last day of the month

Run on days 28-31, then proceed only if tomorrow is the 1st:

0 23 28-31 * * [ "$(date -d tomorrow +\%d)" = "01" ] && /home/dev/scripts/month-end.sh

Test the check in your terminal:

# Linux (GNU date)
date -d tomorrow +%d

# macOS / BSD
date -v+1d +%d

First Thursday of the month

Run every Thursday, then proceed only if the day is 1-7:

0 9 * * 4 [ "$(date +\%d)" -le 7 ] && /home/dev/scripts/first-thursday.sh

Test the check:

[ "$(date +%d)" -le 7 ] && echo "first week" || echo "not first week"

Last Friday of the month (bonus)

Run every Friday, proceed only if the next Friday is in a different month:

0 9 * * 5 [ "$(date -d '+7 days' +\%m)" != "$(date +\%m)" ] && /home/dev/scripts/last-friday.sh

Cleaner option: move the check into the script

Crontab lines get ugly. Keep the schedule simple and put the guard first in the script:

#!/bin/bash
# month-end.sh: only run on the last day of the month
[ "$(date -d tomorrow +%d)" = "01" ] || exit 0

echo "Running month-end job"

On macOS/BSD, change the guard to:

[ "$(date -v+1d +%d)" = "01" ] || exit 0
0 23 28-31 * * /home/dev/scripts/month-end.sh

Limits of this approach

  • date -d is GNU date (Linux). macOS/BSD use date -v+1d.

Alternative: systemd timers

If you need a clean expression, systemd supports it natively:

# first Thursday at 09:00
OnCalendar=Thu *-*-01..07 09:00
# last day of the month at 23:00
OnCalendar=*-*-~01 23:00

Cheat sheet

*/15 * * * *   every 15 minutes
0 9 * * 1-5    weekdays at 9:00
0 0 1 * *      midnight on the 1st of every month
@reboot        once at startup

Debug checklist when a cron job fails

  1. Is cron running? (systemctl status cron or systemctl status crond)
  2. Is the job listed for the current user? (crontab -l)
  3. Is it executable (chmod +x) and does it have a shebang?
  4. Are all paths absolute, and is PATH set?
  5. Is output redirected (>> file 2>&1)?
  6. Did cron start it? (grep CRON /var/log/syslog)
  7. Is the timezone what you expect? (date, timedatectl)

Commands in the post target Linux (Debian/Ubuntu and RHEL-family). Where they differ, it is noted.

Tags:

Updated: