Is structured logging really the be-all and end-all?
summary
Structured logging often adds heavy indexing cost so the proposed alternative keeps raw log lines in cheap storage while indexing just a few fields: time, app, server, pod and a small set of labels. Logs and metrics share labels from Kubernetes/Prometheus so a dashboard query can open log results next to metrics filtered by the same selectors. The system indexes only the metadata to index only metadata, reducing operational complexity and letting teams log freely without exploding search costs.
AI-written summary · about
chapters (9)
- 0:00Introduction
- 0:50History of logging
- 2:18What is structured logging
- 3:09Problems with structured logging
- 5:44Alternative approach
- 6:43Sheep
- 9:20Common Case
- 12:00Ingre Fauna
- 13:43Conclusion
related talks
18mRevolutionizing Kubernetes Logging: Structured Logs for Enhanced Monitoring
10mStructured Logging and You
31mMy Logs Cost More Than My Services
19mFirst Principles of Observability Systems
39mFast log management for your infrastructure
12mOpen-source observability is the way forward, and here’s why
16mShrinking the Observability Bill
20mFast log management for your apps