Observability in Kata containers 2.0
summary
Kata Containers 2 ships built-in observability: Prometheus-format metrics from runtime, VMM and guest agent, plus a new kata-agent-ctl for in-VM debugging. The agent is rewritten in Rust, shrinking memory from ~11 MB to ~300 KB. Metrics expose about 70+ time series per sandbox, gzip to ~7 KB per scrape, with end-to-end collection under ~20 ms and RPC agent queries ~3 ms, so metric collection runs with minimal overhead and no extra collectors.
AI-written summary · about
chapters (11)
- 0:00Intro
- 0:55Agenda
- 4:49Features
- 6:16Why isolation is important
- 13:54Cloud Native - key attributes for infrastructure
- 15:18What's Observability
- 17:36Observability in Kata containers 2.0
- 20:36Metrics in Kata containers 1.x
- 21:39Target for metrics in Kata containers 2.0
- 24:32Metrics pipeline
- 27:00Performance and overhead
related talks
22mKata Containers: Security and Containers Without Compromise
32mUnderstanding Observability in Kubernetes
25mFrom Cloud To AI: How Kata Containers 4.0 Reinvents the Sandbox for the Agent Era
43mA Guide to Observability for Your (Java) Apps in K8s
25mSecuring Untrusted Workloads with Kata Containers on Kubernetes - David Angot & Alex Price
35mKata Containers: Getting Involved with the Kata Community - Anne Bertucio & Samuel Ortiz
39mMore Than Secure: Containerd + KataContainers as Kubernetes Runtime
31mObservability: Dust Off Prometheus with VictoriaMetrics