The Pager Goes Off
Once detected, an alert is routed to an on-call engineer via an alerting platform like PagerDuty or Opsgenie. These tools manage on-call schedules, escalation policies, and notification channels (SMS, phone call, push notification).
The system is screaming for help. Response is the act of hearing that scream and saying, "I'm on it." This is called Acknowledging (or "Acking") the alert — a deliberate action in the alerting tool that signals: a human has seen this and is taking ownership.
Acknowledging is important because it stops the escalation chain. Without an ack, PagerDuty will escalate to the next person in the rotation after a configured timeout (e.g., 5 minutes). This prevents alerts from silently dying if the first engineer is unavailable.
Metric: Time to Acknowledge (TTA)
TTA is the elapsed time between the alert firing and an engineer acknowledging it.
TTA = Time of Acknowledgement − Time of Alert
- Goal: For SEV1 incidents, TTA should be under 5 minutes. For SEV2, under 15 minutes.
- Warning signs of high TTA:
- Alert fatigue: Engineers are receiving so many low-quality alerts that they've started ignoring pages. If your team is getting paged 50 times a day for things that auto-resolve, they will eventually stop responding quickly to the one that matters.
- Wrong team paged: The alert routing is misconfigured and the page goes to a team that doesn't own the affected service.
- On-call schedule gaps: Nobody is actually on call during the incident window.
A common SRE interview question is: "How do you reduce alert fatigue?" The answer has two parts. First, raise the signal-to-noise ratio: every alert that fires should be actionable and require a human response — if an alert auto-resolves without action, it should be deleted or downgraded to a log entry. Second, invest in runbooks: if an engineer knows exactly what to do when a specific alert fires, they're less likely to dread it and more likely to respond quickly.