Revision poster: how a Pod gets its node.

The top row is the pipeline each Pod passes through. The bottom row lists what you control: QoS through requests and limits, and six placement controls.
- 1Filter removes nodes that cannot run the Pod; a failed filter is disqualifying.
- 2Score ranks what remains; the best node is bound.
- 3Requests drive placement, limits drive enforcement.
- 4Guaranteed is evicted last, BestEffort first.
- 5Taints repel, tolerations permit, affinity attracts.
- 6Topology spread keeps replicas balanced across zones.
- 7Priority can preempt; a PDB protects voluntary disruption.
Self-check: cover the poster and answer
- A Pod is Pending with no FailedScheduling event and no Pod object at all: where do you look?
- Why can three replicas still all die with one node?
Requests decide placement, limits decide enforcement.