Kubernetes: Zero downtime update while handling one million req per sec

#1

Post by brendandburns » Wed, Nov 11, 2015, 9:49 PM UTC

Kubernetes: Zero downtime update while handling one million req per sec
blog.kubernetes.io

Re: Kubernetes: Zero downtime update while handling one million req per sec

#2

Post by boulos » Wed, Nov 11, 2015, 10:17 PM UTC

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.

Re: Kubernetes: Zero downtime update while handling one million req per sec

#3

Post by TheIronYuppie » Wed, Nov 11, 2015, 11:23 PM UTC

boulos wrote: post #2

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.

Disclaimer: I work at Google on Kubernetes.

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!