Intro + DeepDive: SIG Scalability
Talkat KubeCon + CloudNativeCon North America 2021 · on Friday, 2021-10-15· at 17:25· Room 402 AB + Online·
summary
SIG Scalability defines SLIs/SLOs, publishes limits like 30 pods per node and 10,000 services, and runs automated performance suites using cluster-loader and Cubemark to validate a 5,000 nodes envelope plus a 100-node diagnostic scale. Test infra collects Prometheus metrics, profiles and visualizes runs in PerfDash/Grafana to spot regressions — e.g., routine count jumped to ~1,000,000 causing 99th‑percentile pod startup slowdowns — and lands fixes like efficient watch resumption and immutable secrets to cut API server load.
AI-written summary · about
chapters (15)
- 0:00Intro
- 0:30What do we do?
- 3:34Kas Scalability Envelope
- 4:38Defining happiness of cluster
- 5:47Scalability SLO
- 6:18API Call Latency SLO
- 7:45Scalability definition
- 8:19Scalability Limits
- 10:06ClusterLoader2
- 11:32Kubemark - Cluster Simulation
- 15:17Observability & Debuggability - Peridash
- 17:59Observability & Debuggability - Profiling
- 22:10Scalability Tests infrastructure
- 23:14Notable 1.22 Scalability Regressions
- 25:52Recent Scalability Improvements
