Kubernetes Operators — Automating Application Operations

Deploying PostgreSQL to Kubernetes is relatively easy. But what about failover, backup, scaling, upgrades? All of that was traditionally done by a DBA manually. Kubernetes Operators promise to automate these operations.

What Is an Operator¶

An Operator is a Kubernetes controller that understands the domain of a specific application. Instead of a generic “ensure 3 replicas,” an Operator knows how to properly perform a PostgreSQL failover or add a replica to an Elasticsearch cluster.

How an Operator Works¶

Technically, an Operator is a combination of a Custom Resource Definition (CRD) and a custom controller.

apiVersion: postgres.core.cz/v1alpha1
kind: PostgreSQLCluster
metadata:
  name: orders-db
spec:
  version: "9.6"
  replicas: 3
  storage:
    size: 50Gi
  backup:
    schedule: "0 2 * * *"
    retention: 7d

Prometheus Operator — Our First Production Operator¶

The CoreOS Prometheus Operator is excellent. Instead of manually configuring Prometheus, you define ServiceMonitor objects that the Operator automatically translates into scrape configurations. New service? Create a ServiceMonitor.

The Future¶

Operators are the future of running complex applications in Kubernetes. Instead of runbooks and on-call shifts, Operators will automatically handle most operational scenarios.

Operators = Next-Generation DevOps Automation¶

Kubernetes Operators transfer operational knowledge from human heads into code. Start with existing Operators and gradually learn the pattern for your own applications.