Cron every minute — `* * * * *` explained

Each field of the cron expression is independent. When all five are wildcards, the job matches every minute of every hour of every day:

• Minute (*) — every value 0-59.
• Hour (*) — every value 0-23.
• Day of month (*) — every value 1-31.
• Month (*) — every value 1-12.
• Day of week (*) — every value 0-7.

At every clock minute, the scheduler matches all five fields and fires the job. There's no delay between runs (other than the time to fire), and no startup grace period beyond what the runtime takes.

Linux crontab

* * * * * /path/to/script.sh
# or with a lock to prevent overlap:
* * * * * flock -n /tmp/myjob.lock /path/to/script.sh

Vercel cron jobs

{
  "crons": [
    { "path": "/api/cron/every-minute", "schedule": "* * * * *" }
  ]
}

Hobby plan limits cron to once-per-day; Pro plan allows once per minute.

AWS EventBridge

cron(* * * * ? *)
# AWS uses 6 fields — ? in day-of-week, * in year.
# Or use the simpler rate expression:
rate(1 minute)

Kubernetes CronJob

apiVersion: batch/v1
kind: CronJob
metadata:
  name: every-minute
spec:
  schedule: "* * * * *"
  concurrencyPolicy: Forbid
  jobTemplate:
    spec:
      template:
        spec:
          containers:
            - name: worker
              image: myapp:latest
          restartPolicy: OnFailure

node-cron

import cron from 'node-cron';

cron.schedule('* * * * *', () => {
  console.log('Running every minute');
});

• * * * * * — every minute, every hour, every day.
• * 9-17 * * 1-5 — every minute during business hours on weekdays. ~480 runs per weekday.
• * * 1 * * — every minute on the 1st of every month.
• */2 * * * * — every 2 minutes (less aggressive alternative).
• */5 * * * * — every 5 minutes (recommended for most polling jobs).

• Long-running jobs — if the work takes more than 60 seconds, runs will overlap. Use a lock file (flock) or a scheduler with concurrencyPolicy: Forbid.
• Cold starts — serverless platforms (AWS Lambda, Vercel) can spend several seconds spinning up. The effective cadence is “every minute, plus a few seconds”, not exactly every 60 seconds.
• Cost — 43,200 invocations/month adds up on paid platforms. Always check whether you really need per-minute cadence.
• Logs — every-minute jobs spam the log stream. Add log levels and rate-limit at the application layer.