SRE Lab · error budgets

SLO Burn Simulator

Pick an SLO and how fast you're burning error budget. See how long the budget lasts, which standard multi-window burn-rate alerts would fire and when, and get Prometheus alert rules to match.

Availability SLO SLO window Burn rate How many times faster than "just on budget" you're failing. 14.4×

Incident length How long the bad period lasts, in minutes.

Burn rate × allowed error rate = actual error rate. At 99.9%, the allowed error rate is 0.1%, so a 14.4× burn means 1.44% of requests failing.

Error budget—

Error rate at this burn—

Budget gone in—

This incident uses—

Budget remaining over time

Which alerts fire, and when

Multi-window, multi-burn-rate alerts, the pattern from Google's SRE Workbook. Each alert needs both a long window (so it's sure) and a short window (so it clears quickly once you've fixed things) above its threshold.

Alert	Threshold	Windows	At this burn	Budget used when it fires

Prometheus alert rules

Assumes an http_requests_total counter with a code label. Swap in your own SLI.