Post

LearnKube
@learnkube.com

This week on the Learn Kubernetes Weekly: 🧠 Workload-Aware Scheduling ⚙️ Automaxprocs Uses One CPU 🔀 Why gRPC Routes to One Pod 🔐 Static Kubernetes OIDC Issuer 🧩 ZSvirt Open-Sourced ⭐️ NeuBird Read it now: https://kube.today/issues/204

6 reposts7 likes

devtocash.bsky.social
@devtocash.bsky.social

The gRPC one bites every team once: a ClusterIP balances per TCP connection, and gRPC keeps one HTTP/2 connection open, so every RPC rides to whichever pod won the first handshake. Fixes: headless Service + dns:/// resolver with round_robin, or MaxConnectionAge server-side to force reconnects.