User Uptime in Practice
summary
Pinterest implements User Uptime (from Google's Meaningful Availability) as a time-based metric by logging user events from a Python Flask API to a dedicated Kafka topic, processing per-user windows in Flink (Xenon), and exporting aggregates to OpenTSDB with a ~5–10 minute evaluation delay. Endpoints are classified into three criticality levels and only critical flows count; server-side uptime yields ~10 days of downtime per day while clients cause 3–4× more impact. The MVP requires manual corrections for third-party infra and is used in monthly reports and incident triage.
AI-written summary · about
chapters (9)
- 0:00Introduction
- 0:56Availability Metrics
- 2:14Success Rate
- 6:14User Uptime
- 8:20User Uptime Implementation
- 10:20Technology Stack
- 13:07Beyond MVP
- 16:44Roadmap
- 21:24Summary
related talks
29mAchieving 99.99% uptime with Open Source Observability
42mEvolution of Observability Tools at Pinterest
29mDowntime Budgets
32mCustomer Centric Monitoring
41mCustomer Centric Observability at Scale Leveraging AIOps & OpenTelemetry
07mEnd User Experience Monitoring
41mMeasuring Availability the Player Focused Way: How Riot Games Changed Its...
24mThe Metrics Question: Evolving Past MTTX