The address that keeps changing
The API is configured with the current addresses of the three worker instances. It worked when those were long-lived machines.
Now they're containers that get replaced on every deploy, rescheduled when a node drains, and given new addresses each time. The list goes stale within hours and nobody notices until a call fails.
Give the API one address that stays true.
Components — tap one, then tap a slot on the diagram
!The API is calling addresses that no longer exist. The list went stale hours ago.
Boundaries, outermost first: Replaced constantly: Worker, Worker, Worker Outside every boundary: API (holds a list of addresses; FAILED: stale list), an empty slot for the what the caller asks for Connections: API calls what the caller asks for — one name (step 1) what the caller asks for calls Worker (step 2) what the caller asks for calls Worker what the caller asks for calls Worker
Worker
Worker
APIholds a list of addressesstale list
Worker