Observing Applications in Production
summary
Claim: production systems need end-to-end observability, not just topic-level monitoring. Concrete practices: use cloud vendor tooling (AWS/GCP/Azure) or New Relic/Dynatrace for dashboards and alerts, log aggregation and tracing with propagated request IDs, and proactive synthetic users that hit full workflows. Watch for surprising failure modes: a two‑year recurring outage becomes a two‑hour fix once one customer’s malformed data and a jammed message queue are traced. Avoid alert fatigue, add health checks, monitor CPU/memory/throughput, and audit via event sourcing or immutable DBs.
AI-written summary · about
chapters (12)
- 0:00Introduction to observability
- 3:10Monitoring vs observability
- 8:07Failure is inevitable in systems
- 10:14Cost of poor performance
- 17:19Key metrics to monitor
- 25:42Granularity in error reporting
- 33:08Actionable error handling
- 36:37Event correlation in distributed systems
- 43:40Case study: diagnosing system failure
- 46:06Synthetic users for monitoring
- 53:32Designing for end-to-end tracing
- 57:20Complexity of distributed systems
related talks
30mBest practices for application observability at scale
42mEffective App Monitoring in Production
13mForget Production: Observability a Non-Production Necessity
40mDemystifying Observability and Monitoring
38mEffective App Monitoring in Production
45mDeveloping Observable Applications for Production
28mProduction Testing through Monitoring
30mA Well-Maintained Production Application