Course contentsShow
← Cloud and Infrastructure
Lesson 1157 of 2,98923. Kubernetes WorkloadsFree lesson

Label Selectors in ReplicaSets

Using matchLabels and matchExpressions to define which Pods a ReplicaSet manages.

Label Selectors in ReplicaSets

What you'll learn: How ReplicaSets use label selectors to identify and manage the Pods they control.

The Matching Problem

A ReplicaSet needs a way to know which Pods it's responsible for. Remember from earlier lessons that Pods have labels—simple key-value pairs attached to them. ReplicaSets use label selectors to create a query that matches those labels, allowing the ReplicaSet to find and manage its Pods.

Think of it like a warehouse manager who needs to count inventory. Instead of memorizing every box, they look for boxes with specific colored tags. The tag color is the "selector," and the boxes are Pods.

Two Selector Types

Kubernetes gives you two ways to define selectors in a ReplicaSet:

matchLabels — Simple equality matching. The ReplicaSet manages Pods where labels exactly match the specified key-value pairs.

selector:
  matchLabels:
    app: frontend
    tier: web

This matches any Pod with both app=frontend AND tier=web labels.

matchExpressions — More flexible, set-based matching with operators like In, NotIn, Exists, and DoesNotExist.

selector:
  matchExpressions:
    - key: app
      operator: In
      values:
        - frontend
        - api
    - key: environment
      operator: NotIn
      values:
        - production

This matches Pods where app is either "frontend" or "api" AND environment is NOT "production."

Why This Matters

The selector creates a boundary of responsibility. The ReplicaSet continuously watches for Pods matching its selector and maintains the desired replica count. If you manually create a Pod with matching labels, the ReplicaSet will adopt it. If a Pod loses its matching labels, the ReplicaSet abandons it.

Key Takeaway:
Label selectors (matchLabels and matchExpressions) define which Pods a ReplicaSet manages by matching Pod labels—creating a dynamic, queryable relationship between controller and workload.