The shop from Kubernetes: Desired State runs as a Deployment of 4 pods on a cluster of 3 nodes. Each node is a VM with 4 CPU cores (illustrative size). Today 120 requests per second are shared by the 4 pods, so each handles 30. A campaign is planned and traffic may double within minutes.
Three jobs follow. Give each pod the CPU and memory it needs, and no more than its share. Add capacity when load rises and shed it when load falls. Ship new versions without taking the shop down. This lesson takes them in that order.
💡Good to know
Capacity, elasticity and safe change: three separate problems with three separate settings.