The Gotchas of Zero-Downtime Traffic on Kubernetes /w Leigh Capili
Are you noticing your customers receive 503's every now-and-then? Does it seem to happen a lot when you're updating your app or even just rotating your k8s cluster nodes?
Maybe you used to have this problem; then you added some strange settings and it's mostly working now, but you have no idea why.
What most people need from kubernetes regarding traffic is a very repeatable but under-documented combination of esoteric, non-default options.
We'll explain these options, by first walking through the basic needs of shifting application traffic and then applying that knowledge to the states of compute, rollout, and canonical networking we see with Kubernetes.
Expect tidbits about CRI, CNI, Ingress, and k8s design trade-offs :)
By the end of this session, we should all understand how to keep modern applications serving successful requests for a myriad of typical edge-cases.
Leigh Capili is Developer Experience Engineer @ https://weave.works
He enjoys being helpful and currently contributes to kubernetes.
Leigh longboards with his partner, Tanya, and their dog, Pepsi, is a very good boy.
You can find him @ https://github.com/stealthybox and https://twitter.com/capileigh