Understanding Kubernetes Pods, ReplicaSets, Deployments, and StatefulSets

1. Core Component Principle — Pod Fundamentals

1.1 What is a pod

Pod can be seen as a container that holds Docker containers, essentially a container for containers.

Pod is a virtualized group with its own IP address and hostname, using namespaces for isolation, like an independent sandbox.

Pod acts like an independent host that can encapsulate one or multiple related containers, with inter‑container communication via localhost.

1.2 What are pods used for

Typically, a pod manages a set of related services; the diagram shows multiple containers (C) inside a pod.

How to understand a group of related services? A request goes to Nginx, which forwards to a web container, which accesses a database, and the response returns to the user; this chain defines related services.

1.3 Implementing a web service cluster

Simply replicate multiple pod copies; scaling up or down is done by controlling pod count.

1.4 Underlying pod network and storage

Pods have their own IP and ports; using IP:port for inter‑container communication incurs overhead. To improve performance, Kubernetes uses a pause container that shares network and storage.

Before creating containers, a pause container is created to share network and storage.

All service containers share the pause storage, delegating data storage to pause.

The pause container acts as the network interface, allowing localhost communication with high performance.

2. ReplicaSet Controller

2.1 Basic understanding

Manages the number of pod replicas to match the desired count (e.g., replicas = 3).

If a pod fails, the ReplicaSet immediately creates a new one to maintain the count.

2.2 Difference between ReplicaSet and ReplicationController

Both control replicas, but their label selector capabilities differ: ReplicaSet supports both single and composite selectors, while ReplicationController only supports single selectors.

Labels (key=value) allow the controller to identify and manage the correct pods.

3. Deployment Object

3.1 Rolling updates

Deployments enable rolling updates by creating a new pod version (v2) to replace the old one (v1) without downtime.

3.2 Deployment model

Deployment works with a ReplicaSet; the steps for a rolling update are:

Deployment creates a new ReplicaSet.

The new ReplicaSet creates new pods.

During update, old pods are removed while the old ReplicaSet remains, allowing rollback.

4. StatefulSet for Stateful Services

4.1 Motivation

Containers are transient; data can be lost if a pod restarts. Deployments are suited for stateless services, while StatefulSet addresses stateful workloads by preserving hostnames and using PersistentVolumeClaims.

4.2 Understanding stateful vs stateless services

Stateful services require real‑time data storage and cannot rejoin the cluster seamlessly after removal.

Stateless services have no persistent data and can be removed and re‑added without impact.

4.3 Deployment model

StatefulSet uses PVCs to store data; when a pod is recreated, the same hostname ensures it can locate its previous data.