Kubernetes StatefulSets

Saurabh AdhauSaurabh Adhau
5 min read

Introduction

Kubernetes provides various mechanisms to manage and orchestrate containerized applications, with StatefulSets being a crucial component for applications requiring stable, unique network identifiers and persistent storage. This article explores what StatefulSets are, their advantages, suitable scenarios, practical examples, and how to scale StatefulSets effectively.

What are StatefulSets?

StatefulSets are a Kubernetes resource designed to manage stateful applications. Unlike Deployments, which are intended for stateless applications, StatefulSets provide guarantees about the ordering and uniqueness of pods. Each pod in a StatefulSet has a stable, unique network identity that persists across rescheduling.

Key Features of StatefulSets

  1. Stable Network Identity: Each pod gets a unique, stable hostname.

  2. Stable Storage: Each pod is associated with a persistent volume, retaining data across rescheduling.

  3. Ordered Deployment and Scaling: Pods are created, updated, and deleted in a specific order.

  4. Ordered, Graceful Rolling Updates: Pods are updated in sequence, ensuring application stability during upgrades.

Pros of Using StatefulSets

  • Data Persistence: Ensures that data stored by the application is not lost during pod rescheduling.

  • Stable Network Identity: Each pod retains the same hostname, facilitating stable network identities.

  • Ordered Operations: Guarantees the order of pod creation, update, and deletion, which is critical for applications like databases.

  • Easy Scaling: Simplifies scaling stateful applications while maintaining data integrity.

Scenarios for Using StatefulSets

StatefulSets are ideal for applications that require one or more of the following:

  • Persistent Storage: Applications like databases (MySQL, PostgreSQL, MongoDB) that need data persistence.

  • Stable Network Identities: Services requiring consistent network identities, such as Zookeeper, Kafka, or Redis.

  • Ordered Deployment: Applications needing ordered startup and shutdown processes.

Practical Examples of StatefulSets

Example 1: Deploying a Stateful MySQL Cluster

Step 1: Create a Headless Service

A headless service is required to manage the network identities of the pods in the StatefulSet.

apiVersion: v1
kind: Service
metadata:
  name: mysql
spec:
  ports:
    - port: 3306
  clusterIP: None
  selector:
    app: mysql

Create the service using:

kubectl apply -f mysql-service.yaml

Step 2: Create a StatefulSet

Define the StatefulSet to deploy a MySQL cluster.

apiVersion: apps/v1
kind: StatefulSet
metadata:
  name: mysql
spec:
  serviceName: "mysql"
  replicas: 3
  selector:
    matchLabels:
      app: mysql
  template:
    metadata:
      labels:
        app: mysql
    spec:
      containers:
      - name: mysql
        image: mysql:5.7
        ports:
        - containerPort: 3306
        env:
        - name: MYSQL_ROOT_PASSWORD
          value: "rootpassword"
        volumeMounts:
        - name: mysql-persistent-storage
          mountPath: /var/lib/mysql
  volumeClaimTemplates:
  - metadata:
      name: mysql-persistent-storage
    spec:
      accessModes: [ "ReadWriteOnce" ]
      resources:
        requests:
          storage: 1Gi

Apply the StatefulSet configuration using:

kubectl apply -f mysql-statefulset.yaml

Example 2: Deploying a Kafka Cluster

Step 1: Create a Headless Service

apiVersion: v1
kind: Service
metadata:
  name: kafka
spec:
  ports:
    - port: 9092
  clusterIP: None
  selector:
    app: kafka

Create the service using:

kubectl apply -f kafka-service.yaml

Step 2: Create a StatefulSet

apiVersion: apps/v1
kind: StatefulSet
metadata:
  name: kafka
spec:
  serviceName: "kafka"
  replicas: 3
  selector:
    matchLabels:
      app: kafka
  template:
    metadata:
      labels:
        app: kafka
    spec:
      containers:
      - name: kafka
        image: wurstmeister/kafka:2.13-2.6.0
        ports:
        - containerPort: 9092
        env:
        - name: KAFKA_BROKER_ID
          valueFrom:
            fieldRef:
              fieldPath: metadata.name
        - name: KAFKA_ZOOKEEPER_CONNECT
          value: zookeeper:2181
        volumeMounts:
        - name: kafka-persistent-storage
          mountPath: /kafka
  volumeClaimTemplates:
  - metadata:
      name: kafka-persistent-storage
    spec:
      accessModes: [ "ReadWriteOnce" ]
      resources:
        requests:
          storage: 1Gi

Apply the StatefulSet configuration using:

kubectl apply -f kafka-statefulset.yaml

Scaling StatefulSets

Scaling StatefulSets is essential for managing workload changes, improving application performance, and maintaining availability. Kubernetes allows you to scale StatefulSets easily while preserving the guarantees of stable network identities and persistent storage.

Scaling Up

To scale up a StatefulSet, you simply increase the number of replicas. For example, to scale the MySQL StatefulSet from 3 to 5 replicas, you can use the following command:

kubectl scale statefulset mysql --replicas=5

This command increases the number of MySQL pods, creating additional pods with unique identities and stable storage.

Scaling Down

To scale down a StatefulSet, decrease the number of replicas. For example, to scale the Kafka StatefulSet from 3 to 2 replicas:

kubectl scale statefulset kafka --replicas=2

When scaling down, Kubernetes deletes pods in reverse order, ensuring data integrity and application stability.

Autoscaling StatefulSets

While Kubernetes supports Horizontal Pod Autoscaling (HPA) for Deployments and ReplicaSets, it does not natively support HPA for StatefulSets. However, you can achieve similar functionality using custom metrics and the Kubernetes metrics server.

Step 1: Install Metrics Server

Ensure the metrics server is installed in your cluster:

kubectl apply -f https://github.com/kubernetes-sigs/metrics-server/releases/latest/download/components.yaml

Step 2: Configure Custom Metrics

Create a custom metrics-based autoscaler using the Kubernetes API and tools like Prometheus. This setup involves configuring Prometheus to collect metrics and creating an autoscaler that scales the StatefulSet based on these metrics.

Example: Scaling MySQL Based on CPU Usage

apiVersion: autoscaling/v2beta2
kind: HorizontalPodAutoscaler
metadata:
  name: mysql-autoscaler
spec:
  scaleTargetRef:
    apiVersion: apps/v1
    kind: StatefulSet
    name: mysql
  minReplicas: 3
  maxReplicas: 10
  metrics:
  - type: Resource
    resource:
      name: cpu
      target:
        type: Utilization
        averageUtilization: 50

Apply the HPA configuration:

kubectl apply -f mysql-autoscaler.yaml

Conclusion

StatefulSets in Kubernetes provide a robust solution for managing stateful applications that require stable network identities and persistent storage. They are ideal for deploying databases, messaging systems, and other stateful applications. By leveraging the ordered deployment and scaling features, you can ensure the reliability and stability of your stateful services within a Kubernetes cluster. The practical examples provided demonstrate how to deploy and scale MySQL and Kafka clusters using StatefulSets, highlighting their ease of use and effectiveness in managing stateful workloads.

10
Subscribe to my newsletter

Read articles from Saurabh Adhau directly inside your inbox. Subscribe to the newsletter, and don't miss out.

Written by

Saurabh Adhau
Saurabh Adhau

As a DevOps Engineer, I thrive in the cloud and command a vast arsenal of tools and technologies: โ˜๏ธ AWS and Azure Cloud: Where the sky is the limit, I ensure applications soar. ๐Ÿ”จ DevOps Toolbelt: Git, GitHub, GitLab โ€“ I master them all for smooth development workflows. ๐Ÿงฑ Infrastructure as Code: Terraform and Ansible sculpt infrastructure like a masterpiece. ๐Ÿณ Containerization: With Docker, I package applications for effortless deployment. ๐Ÿš€ Orchestration: Kubernetes conducts my application symphonies. ๐ŸŒ Web Servers: Nginx and Apache, my trusted gatekeepers of the web.