CRI-O development: behind the scenes
summary
CRI-O development promises to never break Kubernetes and enforces that by running extensive CI: a CRI test binary, Kubernetes node e2e suites, Kata container and VM workload checks, plus performance runs and artifact production. The project runs 1500 tests locally and about ~2000 tests per PR, catches regressions (including a persistent memory leak caught in performance runs), parallelizes CI to reduced runtime from 4h to <2h, and wires Travis and custom CI to produce runnable artifacts.
AI-written summary · about
chapters (22)
- 0:00Intro
- 0:55CRIO
- 2:24Promise
- 3:11Features
- 4:54Stability
- 5:55Testing
- 6:58Content
- 7:26Maps
- 8:00Unit tests
- 8:42Node endtoend tests
- 9:09Manual endtoend tests
- 9:50CRI validation
- 10:11upstream test suites
- 10:29dere CI
- 10:41virtual machine
- 11:10performance tests
- 11:43memory leak
- 12:34performance testing
- 12:52integration testing
- 15:10default
- 15:44takeaways
- 17:44Github page
