Kubernetes Pod

A Kubernetes Pod groups one or more closely related containers for scheduling and execution on a node. Containers in a Pod share networking and can share declared storage volumes.

Restarting a container inside a Pod differs from replacing the Pod. A replacement has a new identity and may run on another node. Neither action proves that the previous attempt made no external write.

Treat the Pod as an execution attempt boundary, not a durable business record. Keep required output and replay evidence in storage with the intended lifecycle, and diagnose Pod conditions and container status together.

Containers in one Pod share an IP address and port space, so they can communicate over localhost and must avoid conflicting port bindings. They do not automatically share their separate image filesystems; share files through an explicitly mounted volume. Put containers together when they need this close lifecycle and locality, rather than putting every pipeline stage into one Pod.

See Running Data Workloads on Kubernetes for manifest examples.

Further reading

• Kubernetes: Pods