StatefulSets

Deployments are usually used to deploy stateless components of your application. For stateful components, Kubernetes provides another API Object named StatefulSet. The principle of this operation is very similar to Deployment – it manages ReplicaSets and Pods in a declarative way and provides smooth rollouts and rollbacks. However, there are some key differences:

• StatefulSets ensure a deterministic (sticky) ID of Pods, which consists of <statefulSetName>-<ordinal>. For Deployments, you would have a random ID consisting of <deploymentName>-<randomHash>.
• For StatefulSets, the Pods are started and terminated in a specific, predictable order while scaling the ReplicaSet.
• In terms of storage, Kubernetes creates PersistentVolumeClaims based on volumeClaimTemplates of the StatefulSet Object for each Pod in the StatefulSet and always attaches this to the Pod with the same ID. For Deployments, if you choose to use volumeClaimTemplates, Kubernetes will create a single PersistentVolumeClaim and attach the same to all the Pods in the Deployment.
• You need to create a headless Service Object that is responsible for managing the deterministic network identity (DNS names) for Pods. The Headless Service allows us to return all Pod IPs behind the Service as DNS A records instead of a single DNS A record with a Service Cluster IP.

StatefulSets use a similar Spec to Deployments – you can find out more regarding StatefulSets by looking at the official documentation: https://kubernetes.io/docs/concepts/workloads/controllers/statefulset/.