Detection
Most incidents are detected by our monitoring before customers notice them. We run continuous probes against all ingress endpoints, track error rates, CPU and memory saturation, and disk I/O. Anomaly detection alerts are configured with thresholds tuned for each cluster's normal baseline.
Response
When an alert fires:
- An on-call engineer is paged immediately
- Initial triage happens within 5 minutes
- You receive a notification via your support channel with an initial assessment
- A P1 ticket is created in your dashboard automatically
Communication during the incident
We send updates at minimum every 30 minutes for active P1 and P2 incidents. You don't need to follow up — we'll come to you. If you have additional context (a recent deploy, a spike in traffic, a code change), share it via the ticket — it helps significantly.
Resolution and post-mortem
Once the incident is resolved, the ticket is updated with a summary of what happened. For P1 incidents, we deliver a full post-mortem document within 48 hours, covering:
- Timeline of events
- Root cause analysis
- What we did to resolve it
- Changes made to prevent recurrence
Post-mortems are written without blame, focused on systemic improvements.
What you can do
During a live incident, the most helpful things you can do are:
- Stay available on your support channel for questions
- Share any application-level logs or error context you have access to
- Hold off on deploying changes unless specifically asked by the engineer on the incident