Leverage LXD/LXC with Kubernetes
summary
Experiment runs Kubernetes worker nodes inside LXD/LXC containers to raise free‑tier density on a 1‑core, 8GB host, starting workers in under a minute versus 5–10 minutes for VMs. Implementation runs kubelet and kube‑proxy directly in LXC with Docker for pods; idle workers use about 100MB average per worker (peaks ~140MB), guestbook deploys peak ~1.5GB. Problems: privileged Docker is unsafe for multi‑tenant free tier and Kubernetes OOM tuning required patching.
AI-written summary · about
chapters (19)
- 0:00Introduction
- 0:28IBM Cloud Kubernetes
- 0:57Key capabilities of Kubernetes
- 2:24Deploying Kubernetes in Bluemix
- 4:20Free Kubernetes Tier
- 5:25Single VM
- 6:09Free Tier
- 6:47Density
- 8:10Experiments
- 9:24First Experiment
- 10:35LXD
- 12:57LXE
- 15:53Cuban Ad Workers
- 18:05Screenshots
- 20:09Demo
- 31:25Memory
- 33:40Issues
- 34:22Summary
- 37:30Architecture differences
related talks
37mManaging Containers at Scale with CoreOS and Kubernetes
14mCut Costs, Not Performance: Implement Kubernetes Cost Optimization Strategies
02h 26mKubernetes Performance Tuning Workshop
44mQuick Prototyping with LXC and Puppet
43mKubernetes cost optimisation - How to actually do it
36mRunning Containers on a Resource Constrained Embedded Device
43mBuilding Docker Images on Kubernetes: A Long Journey!
01h 39mLab: Linux Container Internals