Operators

StatefulSets give you stable identities, ordered deployment, and persistent storage. But running a production PostgreSQL cluster on Kubernetes requires much more: configuring streaming replication, managing automatic failover, handling backup schedules, performing rolling upgrades that respect replication lag, and managing connection pooling.

A human database administrator knows all of this. They know that before upgrading the primary, you should ensure the replica is caught up. They know that failover requires promoting the replica, updating DNS, and reconnecting poolers. They know that backups should go to S3 with point-in-time recovery.

An Operator encodes this human knowledge into a Kubernetes controller. Instead of a StatefulSet that manages pods generically, the PostgreSQL Operator understands PostgreSQL-specific lifecycle operations. It knows how to initialize clusters, configure replication, perform failovers, manage backups, and handle upgrades, all automatically.

An Operator extends Kubernetes with:

1. Custom Resource Definition (CRD): Teaches Kubernetes a new resource type
2. Custom Controller: Watches the CRD and performs domain-specific reconciliation

Example. PostgreSQL Operator:

apiVersion: acid.zalan.do/v1
kind: postgresql
metadata:
 name: my-cluster
spec:
 numberOfInstances: 3
 postgresql:
 version: "16"
 volume:
 size: 50Gi
 enableConnectionPooler: true
 backup:
 s3Bucket: my-backups
 schedule: "0 */6 * * *"

From this single manifest, the Operator:

• Creates 3 StatefulSet pods (1 primary, 2 replicas)
• Configures streaming replication
• Sets up connection pooling (PgBouncer)
• Schedules backups every 6 hours to S3
• Monitors replication lag
• Handles automatic failover if the primary dies
• Manages rolling upgrades that respect replication

Popular Operators: