On-call, silences and SLOs

On-call, silences and SLOs

On-call schedules

Define rotations so the right person is notified first, with escalation if they don't ack. Critical alerts follow the schedule; lower severities can go to a channel only.

Silences

Mute known/expected noise for a window — a deploy, a maintenance task — without disabling the rule. The rule keeps evaluating; you just won't be paged for the muted scope.

• Scope a silence to a service, environment, or specific alert.
• Always set an expiry so silences don't outlive the reason.

SLOs and error budgets

Define a Service-Level Objective for a user-facing metric and VerOps tracks the error budget — how much failure you can absorb before breaching.

SLO: "99.9% of checkout requests succeed over 30 days"
  -> error budget = 0.1% of requests
  -> alert when budget burn is fast (e.g. 2% of monthly budget in 1 hour)

Burn-rate alerts page you when you're spending the budget too fast — focusing attention on what actually affects users instead of every transient blip.