I have a cluster (GKE) with 1 node of type N1 Standard (1 CPU). My deployment requests 1 replica of a container that requests .5 of 1 CPU.
When I deploy this, I expect that the pod should get created (enough resources available in the cluster, 1 CPU available and only .5 requested). But the pod goes in pending state and waits for cluster to scale up the nodes. The pod comes up with cluster scaled up to 3 nodes. Why 3 nodes when only 1 was sufficient?
This is the same example used by Nigel in the Cluster Autoscaling demo lesson.
Am I missing something?