Workload scheduling constraints

When you deploy workloads on Kubernetes with Karpenter, you can specify a variety of scheduling constraints in your pod specs to control where and how those pods will run. These constraints ensure that your pods are scheduled on nodes that meet certain requirements, such as specific hardware specs, specific zones, or co-location (or separation) from other pods. As you’ve learned in a previous section, Karpenter honors these constraints to decide what nodes to provision for these pending pods.

Scheduling happens through layered constraints: the cloud provider defines broad limits (the universe of all available instance types, zones, etc.), then you define NodePool constraints (nodes that Karpenter can create), and finally, you define your pod’s scheduling constraints. Keep in mind that a pod’s constraints must fall within the requirements set by the NodePool, else pods will remain unschedulable. The following sections describe...