Pains and gains of Kubernetes for Developers, story from experience
summary
Swisscom runs production Kubernetes clusters since 2017 with 300+ worker nodes and thousands of pods and discovers scaling pain points like a DNS script hammering requests, so they shard workloads across clusters. A second service ingests about 100,000 events/sec into Apache Kafka and Zookeeper managed by the Strimzi operator, exposes functionality via Go microservices packaged with Helm, and automates CI/CD with GitLab pipelines, cert-manager, nginx ingress, Prometheus, ELK, Velero and Kafka Connect to S3 while aiming for a cf push-like developer UX.
AI-written summary · about
chapters (3)
related talks
27mRunning Kubernetes in Production A Million Ways to Crash Your Cluster
33mKubernetes: Crossing the Chasm
27mA Safari of Kubernetes and its Natural Habitat
35mKubernetes, Lessons Learnt
42mMastering Kubernetes: Challenges, Strategies, and Best Practices
45mLearning How to Pronounce Kubernetes to Production in 3 Months!
35mKubernetes journey
41mKubernetes — A Year in Production