Operator

🔧 Problem Operators Solve

❌ The Problem

Kubernetes manages stateless apps (like NGINX, APIs) easily via Deployments and Services. But stateful apps (like databases) need:

• Backups

• Restores

• Scaling with persistent volume migration

• Health checks + failover

• Version-aware upgrades

Manually doing this is error-prone and non-repeatable.

Operators can be simple (e.g., restart app if it crashes) or advanced (e.g., auto-backup before upgrade).

🔄 Real-Time Example

🐘 Example: PostgreSQL Operator

Problem: PostgreSQL isn’t natively supported in Kubernetes for:

• Safe backup/restore

• Replication/failover

• Version upgrades

• PVC migration

Solution: Use an Operator like Zalando Postgres Operator

💡 What It Does:

• Deploys PostgreSQL clusters using a CRD like:

yamlCopyEditapiVersion: "acid.zalan.do/v1"
kind: postgresql
metadata:
  name: my-postgres
spec:
  teamId: "myteam"
  volume:
    size: 10Gi
  numberOfInstances: 3
  postgresql:
    version: "15"
  resources:
    requests:
      cpu: 500m
      memory: 500Mi

• Automatically:

  • Creates StatefulSets with correct configs

  • Sets up replication

  • Monitors health

  • Performs failover

  • Manages backups (e.g., to S3)

---

📚 Operator Examples in Production

Operator	Manages	Features
Prometheus Operator	Prometheus, Alertmanager	Auto config, upgrades, alerting
Cert-Manager	TLS certs via Let's Encrypt	Auto renewals, validation
ArgoCD Operator	GitOps CD tool	Installs and maintains ArgoCD
Elastic Operator	Elasticsearch & Kibana	Handles cluster upgrades, scaling, snapshots
Kafka Operator	Apache Kafka	Handles topics, partitions, failover, scaling

---

🧠 Summary

Concept	Description
CRD	Custom resource you define (e.g., PostgreSQL)
Controller	Logic to watch CRD and take action
Operator	Combination of CRD + controller logic, often packaged as a Helm chart or container
Why Use It?	To manage complex, stateful, or repetitive workloads like a human would