Keynote: Reliability and Performance Testing for Kubernetes Operators
summary
Operators hit limits because the reconciliation loop turns every event into work, causing large caches, concurrency and rate‑limit bottlenecks; testing must cover both load vs stress testing and client/server rate settings, not just CPU/memory. The team evaluates clusterloader2 to generate custom CR templates, emulate external dependencies and produce reproducible metrics; the tool works but has a steep learning curve and forces tradeoffs between realism and controllability.
AI-written summary · about
related talks
35mKubernetes, Lessons Learnt
12mKeynote: From Outage To Observability: Lessons From a Kubernetes...
30mLive Experiments with K8s Applications: Pitfalls and Ho...
41mScaling Reliability Engineering with Tools
20mKeynote: Finding the Joy in Chaos Engineering
50mFast-track Kubernetes and platform rebuilds with observability
34mReading 100+ Kubernetes Post-Mortems - lessons learned
20mYou can't have a cluster*!$& without a cluster!