Composition of containers - A design pattern for applications running on K8s
summary
Demonstrates splitting huge application images (some >20 GB) into separately versioned containers so fixes and scans target only the small image that changed, cutting build and push time and avoiding full-image redeploys; earlier workarounds used init containers, sidecars copying into shared volumes or risky CSI drivers, but Kubernetes adds process namespace sharing (v1.17) which lets one container access another’s filesystem and PID namespace; implemented for Python virtualenv libraries and yields seconds-scale startup and independent security patching in practice.
AI-written summary · about
chapters (7)
- 0:00Start
- 2:12Usage Patterns
- 11:24Leader and Election Pattern
- 12:02Init Container
- 13:12Definition of the Composition
- 20:53Sidecar Container
- 26:10How Do You Use that Kind of Composition in Production Platform
related talks
26mContainerization crossroads: Experiences-Experiments-Expectations
05mGetting Started Getting Started with Containers
01h 00mContainer Usage Patterns
43mContainer images anti-patterns
31mRussian Doll: Extending Containers with Nested Processes
05mBare Metal to Containers to K8s: How We Got Here
30mContainers Without the Magic
48mContainerize everything!