It's easy to miss this, but the big point is doing a rolling update when running at serious scale:
> What is exciting is that while successfully handling 1 million HTTP requests per second with uninterrupted availability, we have Kubernetes perform a zero-downtime rolling upgrade of the service to a new version of the software while we're still serving 1 million requests per second.
Yep! The IPTables work, service endpoints based on labels and the built in rolling-update functionality is what makes this possible. No agents or service discovery required!
> What is exciting is that while successfully handling 1 million HTTP requests per second with uninterrupted availability, we have Kubernetes perform a zero-downtime rolling upgrade of the service to a new version of the software while we're still serving 1 million requests per second.