Hello and welcome to the Horizontal Pod Autoscaling issue again!
There were soooo many great HPA articles that we had to do another. This time we are focusing on external and custom metrics, how to use them - why use them and more. There are a couple more goodies in here like Canary Analysis and how to be Overcommitted, the right way.
Read and enjoy!
Canary Analysis: What is it even? We will let you read more here but the tl;dr is Does the canary pass your core KPIs or not? (In keeping with the theme of this issue Flagger does consume HPA objects so don’t worry)
Horizontal Pod Autoscaling on Istio Metrics the Banzai way!
Applications use resources based on their workload (mostly) - Why not scale on the workload? That's where custom metrics come in to play. Included is an awesome example of how to implement a provider for the custom metrics API!
How to connect your prometheus to your Horizontal Pod Autoscaler in a nutshell. Thanks @weaveworks.
Maybe you use Stackdriver for HPA? Maybe you want an example of how to use external metrics for HPA? Checkout @JessicaGreben's slick article on how.
Included here because not a lot of people talk about the PriorityClass resource which is important when your over committing clusters (especially when you have horizontal autoscaling in play).
The aha moment of HPA - love it!