Fix for Service load balancers should include unschedulable nodes
#k8s
#solved
Solution Summary
Kubernetes service load balancers may prematurely drop traffic during node cordoning operations. The fix, detailed in PR #90823, updates the cloud controller logic so load balancer target groups include unschedulable nodes. This decoupling allows active connections to terminate gracefully rather than failing immediately upon node drainage, ensuring high availability.
The Problem
Detailed fix for Kubernetes issue PR #90823
Why does this happen?
Technical root cause identified in the networking layer.
Code Example
https://github.com/kubernetes/kubernetes/pull/90823Step-by-Step Fix
Follow the resolution implemented in PR #90823 on GitHub.