Engineering Decisions · Middle · Question 12 of 14
On-call engineers get dozens of alerts a night, and most need no action. Users, meanwhile, sometimes notice outages first. What do you alert on?
- 1Burn rate of a user-facing SLO
- 2Every error line in the application logs
- 3Each host going above 80% CPU
Show the answer
Correct answer: Burn rate of a user-facing SLO
Explanation
An SLO alert fires when users are actually affected and the error budget is melting, so a page means real work. CPU alerts fit capacity planning as a dashboard or a ticket, but high CPU often harms nobody. Alerting on every error line fits a brand-new service for a short time, and at scale it recreates the noise you're trying to remove.