김태오

The local dual-cluster comparison became the most important dashboard phase of the project.Before this, I had an experimental controller and a live dashboard. That was useful, but it was still too easy to fool myself. I needed a baseline. Not a perfect cloud baseline, but at least a local setup where the same external stimulus hit two Kubernetes clusters: one running the experimental Agent A/B/C..

After the six-layer stack existed, the next problem was making it feel alive. Synthetic traces were useful, but I wanted the dashboard to react to Kubernetes state, not just replayed rows.This phase had two jobs. First, I needed trace and synthetic inputs to keep exercising the orchestrator when no cluster was running. Second, I needed a live Kubernetes loop that could collect a snapshot, run th..