cron & systemd timer

核心 · Key Idea

In one line: there are two mainstream ways to schedule scripts: classic cron (short and flexible) and modern systemd timer (observable, journal-logged, can depend on other units). Production should prefer timer.

cron speedrun#

# m  h  dom mon dow  command
0    3   *   *   *   /usr/local/bin/backup.sh
*/5  *   *   *   *   curl -fsSL https://example.com/healthz > /dev/null
0    9   *   *   1-5 /opt/scripts/report.sh   # Mon–Fri 09:00
@reboot                /opt/scripts/init-once.sh

crontab -e          # current user
crontab -l
sudo crontab -e -u root

systemd timer speedrun#

/etc/systemd/system/backup.service:

[Unit]
Description=Daily backup
 
[Service]
Type=oneshot
ExecStart=/usr/local/bin/backup.sh

/etc/systemd/system/backup.timer:

[Unit]
Description=Run backup daily 03:00
 
[Timer]
OnCalendar=*-*-* 03:00:00
Persistent=true
RandomizedDelaySec=10min
 
[Install]
WantedBy=timers.target

sudo systemctl enable --now backup.timer
systemctl list-timers
journalctl -u backup.service

Analogy#

打个比方 · Analogy

cron is a sticky note on the fridge — reminds you once at the time, but whether you did it or not, nobody knows. systemd timer is a calendar event with read receipt — fires a service, full logs + retries on failure + dependency tracking.

Key concepts#

Five cron fieldsmin hour dom mon dow

wildcard * / range 1-5 / step */15 / list 1,15.

Environmentcron's classic gotcha

cron does **not** source your `~/.bashrc`; PATH is minimal. Use absolute paths or `source` something.

OnCalendarTimer time expression

More natural than cron: `Mon..Fri 09:00`, `hourly`, `daily`.

Persistent=trueCatch up

Run missed jobs after the machine boots. cron can't do this.

RandomizedDelaySecJitter

Avoids every node hammering origin at the same second.

How it works#

Practical notes#

cron output is mailed — without mail configured it piles up in /var/spool/mail/. Redirect to a log file: >> /var/log/backup.log 2>&1.
Timers log to journal by default: journalctl -u backup.
Avoid the top of the hour: 0 0 * * * makes the whole fleet hit origin at once. Use 7 0 * * * or timer RandomizedDelaySec.
cron expression debugging: crontab.guru shows next runs.
Jobs must be idempotent — scheduled tasks must survive multiple runs (retries / catch-up).
@reboot runs once when cron restarts, not every boot. Use systemd timer + OnBootSec=1min for boot-time scheduling.

Easy confusions#

cron

Decades old, one-line config.
You handle logging and catch-up yourself.

systemd timer

Journal integration, dependency / resource limits.
More verbose, **better in production**.