What happens to AI traffic when a machine fails?
Cluster Live
Each tile is a node: the model it serves, its load and its speed, refreshed every second. Stop a worker and watch the gateway move its traffic to a fallback.
Until the home cluster is connected, this view runs on a simulator and is labelled as simulated.
- The outcome: health checks, retries and failover you can see, mapped to their equivalents in a GPU data center.
Built and run by Deepak Kumar Jha, Technical Lead and Solution Architect in Dubai. Talk to Deepak · LinkedIn · GitHub
Questions people ask
What does this page show?
How an AI inference cluster stays up when a machine fails: health checks notice the failure, the gateway moves traffic to a fallback worker, and retries are counted.
Is the cluster live?
Not yet. Until the home cluster is connected, the page runs a simulator and labels its values as simulated.
Who built this?
Deepak Kumar Jha, a Technical Lead and Solution Architect based in Dubai, UAE, with 14 years of building and running production systems. He designed, built, deployed and operates every part of the lab himself: the Angular console, the NestJS APIs, the model serving and the gateway.