What happens when, Kubernetes edition

github.com

168 points·hyzyla··8 comments

I'd be really interested to see one of these trace all the way through a typical micro service application.

I.E. Exactly where does my initial http request go, how does the front-end make it out, how does the front-end reach the back-end, how does that data make it out

How are the DNS names being resolved, that sort of stuff.

I've been using Kubernetes (or openshift) for a while and I still don't fully understand all of the mechanics related to serving client requests end to end

I wonder what would be necessary to get this in Jaeger

This couldn't come at a better time for me. Being able to step through the "magic" of Kubernetes is very instructional. thanks!

This is a great way explain kubernetes. I'm a veteran user and I'm finding it delightful to read. Consider selling an e-book :)

For me this article was a base for understanding other articles about k8s. I agree with you that the author should create an e-book from this.

Seems to be missing any mention of how this whole process is resilient to killing the apiserver or kubelet at any point in the process (it is resilient to that, right...?).

The API server is stateless and should be loadbalanced. The kubelet should be set to restart on failure and is not resilient. It keeps trying to connect to your API server and update its status if it can't it will keep retrying. The controller manager will remove the node after X minutes of no response.

I should mention that the kubelet is plenty resilient for what it does. If at some point kubelet can connect again it will rejoin the cluster.

Reply on news.ycombinator.com