Monitoring in a Kubernetes World
summary
Kubernetes adds at least four infrastructure layers (master node, worker nodes, pods, containers) so monitoring must answer where not just what — monitor the Kubernetes control plane and API server, track orchestrator health as a heartbeat, and centralize logs to stable storage. Expect extreme scale (example: 50 TB of traces per hour; one org with ~120 clusters × ~5,000 nodes) and ephemeral events (AWS Lambda cold starts ~400–700 ms vs warm ~38 ms). Use pre-instrumented distributed tracing libraries and keep tooling counts low.
AI-written summary · about
chapters (11)
- 0:00Introduction
- 1:04Abstraction complexity
- 4:19Kubernetes orchestration
- 7:58Monitoring challenges
- 11:53Orchestrator health
- 13:43Scaling environments
- 16:52Tooling strategies
- 19:23Data granularity
- 21:51Short-lived containers
- 23:36New monitoring rules
- 25:16Stable storage advice
related talks
34mOptimizing operations & container management
11mKubernetes Observability
30mOperating within Normal Parameters: Monitoring Kubernetes
34mMonitoring Cloud Infrastructure
34mPanel: Improving and Managing Kubernetes at Scale
42mDistributed Monitoring: How to understand the Chaos
05mLightning Talk: The Top 5 Kubernetes Metrics to Monitor
29mKubernetes-Defined Monitoring [I]