Load Testing Kubernetes: How to Optimize Your Cluster Resource Allocation in Production [I]
summary
Case study moves a high‑throughput link‑count service to Node + DynamoDB, starts at 4 replicas and hits OOMKilled after copying an old deployment with tiny limits. Uses cAdvisor / heapster / DataDog to observe CPU/memory, runs a ramp-up test with loader.io to find a breaking point (~800–900 RPS), increases memory from 50M to 250M and CPU iteratively (hits scheduling when set too high), then runs a duration test just under the break to set stable limits and capture 500s, spikes, and cron‑job noise.
AI-written summary · about
chapters (9)
- 0:00Intro
- 0:40Case Study
- 3:51Optimal Limits
- 4:19Under and Over Allocation
- 4:58Scaling Up
- 6:09Kubernetes Monitoring
- 8:15Ramp Up Test
- 9:19Demo
- 14:13Load Testing
related talks
48mScale and Performance Testing of Kubernetes: Answers for Specific Applications
47mOptimizing Kubernetes Resource Requests/Limits for Cost-Efficiency and Latency
02h 26mKubernetes Performance Tuning Workshop
15mDemo: Resource(Cost) optimization on planet scale Kubernetes clusters.
27mRunning Kubernetes in Production A Million Ways to Crash Your Cluster
44mPerformance and Scalability Tuning Kubernetes for OpenShift and Docker
29mScale Your Batch / Big Data / AI Workloads Beyond the K8s...
21mLet your Kubernetes environment flourish