slos-and-triggers

A burn rate of 1.0 means budget is consumed evenly over the window; 10.0 means exhaustion in 3 days instead of 30. Budget remaining above 50% is healthy, 10-50% means slow down changes, under 10% means freeze non-critical deploys, and negative means investigate immediately.

Use SLOs to measure reliability against commitments over time and track error budget consumption. Use triggers to catch immediate operational issues like error spikes or latency threshold crossings. SLOs answer "are we meeting commitments"; triggers answer "is something broken right now."

Zero percent compliance usually indicates a misconfigured SLI rather than a real outage. Common causes include the wrong column, inverted boolean logic in the calculated field, or no events matching the SLI. Check the SLI definition before treating it as an incident.

Configure at minimum two alerts: a fast-burn alert with exhaustion time around 4 hours that pages on-call via PagerDuty, and a slow-burn alert tracking budget rate over 24 hours that notifies the team via Slack. An exhaustion time of 0 hours means the budget is fully depleted.

Multi-service SLOs support up to 10 datasets sharing one error budget, but the SLI must be an environment-level calculated field that classifies each event independently. Events cannot be correlated across services, and all included datasets must contain the fields used in the SLI.