Skip to content OpenLabs

Kubernetes Visual Handbook phase 1 · A workload becomes real

A Pod is not its container

A container can restart without a Pod being replaced. A Pod can be replaced without an application being redeployed. Those two identities explain the resource tree, the restart counter and the controller's job.

A restart counter goes up while the Pod’s UID stays put. Delete that Pod, and its replacement has a different UID. The container and the Pod have different lifetimes. Kubernetes schedules each Pod once; a replacement is another object, even if it does the same job. Pod lifetime

The owner matters

A standalone Pod has no controller responsible for replacing it. A Deployment uses ReplicaSets to maintain a desired replica count. Changing replicas changes that count; changing the Pod template starts a rollout. The lab explicitly selects Recreate, which removes old Pods before starting the new revision. It does not simulate the default RollingUpdate strategy. Deployments

Placement is not consumption

Requests reserve schedulable capacity. Limits constrain consumption. A low CPU limit can throttle a container; exceeding a memory limit can lead to an OOM kill. Neither setting itself scales replicas. Resource management

Read the two status fields

Pending and Running are Pod phases. ImagePullBackOff, CrashLoopBackOff and Terminating are useful display statuses, not additional phases. Read the waiting or termination reason alongside the phase. Pod status

Why can a scheduled Pod still be Pending?

Binding chooses a node; it does not prove the container has started. An image-pull failure can leave a bound Pod waiting. In the lab, its node badge and reserved capacity remain while the display reports ImagePullBackOff. The reason explains more than the phase alone. Container states

Primary sources

Cited above by name. Where a source has no link it is because none is stable enough to print — the citation is what you search for.

  1. Kubernetes Authors (current). Pod Lifecycle. Kubernetes documentation.

    The distinction between Pod identity, container restarts, phases and display status.

  2. Kubernetes Authors (current). Resource Management for Pods and Containers. Kubernetes documentation.

    Requests determine placement; limits constrain resource consumption.

  3. Kubernetes Authors (current). Deployments. Kubernetes documentation.

    Replica counts, Pod templates and the explicitly selected Recreate strategy.