Phoenix LiveView · Kubernetes · BYO everything

This page is a WebSocket.

A Phoenix LiveView app running as a clustered, multi-replica deployment. Nothing below is baked in at render time — every number is pushed from the server once a second over the connection this page is holding open. Watch them move.

live from the pod serving you 10.244.3.43
Viewers, cluster-wide 3 counted through Presence, across every pod
Clustered peers 1 10.244.13.39
Uptime, this pod 11h 31m resets on every deploy
BEAM processes 421 one per connected LiveView, plus the runtime
Schedulers 1 pinned to the CPU request, not the node
Memory 76MB idle sockets cost memory, not CPU

Things to poke at

All of these are LiveViews. None of them ship client-side application JavaScript.

Try to break it

Open this page twice

The viewer count goes up in both tabs. If the two tabs land on different pods, the count still agrees — that is Presence replicating over the cluster, not a per-pod counter.

Leave a tab open

The connection is held for as long as you want it. There is no polling; the numbers arrive because the server pushes them.

Come back after a deploy

Uptime resets and the pod name changes, but the page reconnects on its own. Old pods drain for up to 60 seconds before they close anything.