The live feed
The raw change-event stream, and what happens when a deployment cannot provide one.
The Live section is the change stream itself, unaggregated: each event as it arrives, with its operation type, namespace and document key, newest first.
Where Activity answers “what is busy”, this answers “what just happened” — the difference between a dashboard and a tail. Watching writes land while you trigger something from another app is the quickest way to confirm that the thing you deployed is actually writing where you think it is.
Requirements, stated up front
collection.watch() needs:
- a replica set (a standalone
mongodhas no oplog to watch), and - the change stream privilege for the user you connected as.
If either is missing, Kubo detects it and says so — the section explains that the deployment cannot provide a change stream and that the rest of the screens are running poll-only. It does not retry in a loop or present an error page, because this is a property of the deployment, not a fault.
Reconnects
The driver does not resume a dropped change stream on its own, so the app owns that: the stream is re-established after a drop, the same way the Kubernetes watch lists reconnect. A stream that quietly dies and leaves a screen frozen at the last event is worse than no stream at all.
Back to MongoDB.