Intermediate
Open
Pro
Setting an SLO and Alerting on Burn Rate
You own a checkout API doing 50 million requests per 30-day month. The product team wants "99.99% availability" in the SLA. Historically the service runs at about 99.95% success, and your two upstream dependencies each have 99.9% SLAs.
- Is a 99.99% SLA achievable? What SLO and SLA would you propose, and why should they differ?
- For your proposed SLO, compute the monthly error budget in requests and in minutes of full outage.
- Design the alerting: what SLI, what burn-rate thresholds and windows, and what should page versus ticket? Explain why you would not simply page on "error rate > 1%".
Share this question