Scheduling deploys applications by assigning pods to suitable nodes. The scheduler selects a node and verifies that sufficient resources are available; Deployments and ReplicaSets maintain workload replicas. section 1.4.2 section 2.1.2
Use these mechanisms to control pod placement:
- Apply node labels and select them with nodeSelector or label selectors. section 4.1 section 4.3
- Use nodeName for direct placement or node affinity for required rules and preferred placement. section 4.3
- Use pod affinity to colocate pods within a node or topology domain such as a zone or rack. section 4.3
- Declare CPU and memory requests and limits so pods reserve and constrain resource usage. section 4.3
Schedule a pod onto an SSD node
Inspect labels with `kubectl get no --show-labels`, label the worker using `kubectl label no kind-worker disktype=ssd`, and add `nodeSelector: {disktype: ssd}` to the pod YAML. Apply it, then verify placement with `kubectl get po -o wide`; the pod should show `kind-worker` in the NODE column. section 4.1
Figure 4.1 The scheduler schedules the pods to run on the worker node after verifying resource availability.
Visualizes the scheduler verifying node resources before assigning pods to a worker node. section 4.1
