Every 5 minutes cron: what */5 * * * * means and how to run it safely

How the */5 step works

The slash is a step: */5 in the minute field means start at the lowest allowed value (0) and take every fifth one, giving minutes 0, 5, 10, 15, 20, 25, 30, 35, 40, 45, 50 and 55. The other four fields are asterisks, so that set repeats every hour of every day, for 288 runs per day.

Because the step is anchored to minute 0, not to when you saved the crontab, the runs land on the same wall-clock minutes on every server. The parser on this page reads it as: Every 5 minutes. The next run times in the panel above are computed in your browser, in your own timezone, so they are not printed in this text.

Example crontab entry

*/5 * * * * flock -n /run/lock/sync-feeds.lock /opt/app/bin/sync-feeds >> /var/log/sync-feeds.log 2>&1 runs a sync every five minutes and refuses to start a second copy while the first is still working. Without flock, a run that takes six minutes overlaps the next one, and two writers on the same data is how duplicates and lock contention start.

If many servers share the same crontab, they all fire on the same minute. Offsetting the start, for example 2-59/5 on half the fleet, spreads load on the database or API they all call.

systemd, Kubernetes and GitHub Actions

systemd writes the same schedule as OnCalendar=*:0/5. In Kubernetes, schedule: "*/5 * * * *" with concurrencyPolicy: Forbid and a startingDeadlineSeconds of a minute or two gives the same no-overlap behaviour as flock.

Five minutes is the shortest interval GitHub Actions allows for on: schedule, and GitHub documents that scheduled runs can be delayed or dropped during high load, most often at the top of the hour. Treat it as roughly every five minutes, in UTC, not as a precise timer.

Related intervals

Steps only divide evenly when they divide 60. Every 15 minutes and every 30 minutes behave the same way as this one; a step such as */7 produces 0, 7, 14 … 56 and then restarts at 0, leaving a four-minute gap every hour. If you need sub-minute polling, look at every minute plus a loop, or a systemd timer with a seconds field.