How we run on-call without burning people out
Short rotations, real handoffs, and a rule that every page must lead to a fix or a deleted alert.
Beacon Engineering1 min read
On-call is where good engineering cultures go to die. We've tried to design ours so it doesn't.
One week, then you're off
Rotations are seven days long and you get the following week with no meetings. Nobody is on call two weeks in a row, ever.
Every page gets a verdict
If you get paged, one of two things happens by the end of the week: the root cause is fixed, or the alert is deleted. Alerts that page without an action attached are noise, and noise is the thing that burns people out.
Handoffs are written
Every rotation ends with a short doc: what fired, what you did, what's still open. The next person reads it before their shift starts. It takes fifteen minutes and saves hours.
