Kubernetes Operators and Custom Resources

Operator Pattern Fundamentals

The operator pattern extends Kubernetes by encoding operational knowledge into software. Core principles:

Controller pattern: Watch resources, detect drift between desired and actual state, take corrective action.
Reconciliation loop: A single Reconcile function called for every relevant event. Must converge toward desired state.
Declarative API: Users declare intent in .spec; the controller drives reality to match.
Level-triggered, not edge-triggered: React to current state, not to the event that caused it. Always re-read the resource before acting. Never cache assumptions across reconcile calls.

The reconcile function receives a Request (namespace/name), fetches the current object, computes desired state, and applies changes.

The operator pattern extends Kubernetes by encoding operational knowledge into software. Core principles:

Controller pattern: Watch resources, detect drift between desired and actual state, take corrective action.
Reconciliation loop: A single Reconcile function called for every relevant event. Must converge toward desired state.
Declarative API: Users declare intent in .spec; the controller drives reality to match.
Level-triggered, not edge-triggered: React to current state, not to the event that caused it. Always re-read the resource before acting. Never cache assumptions across reconcile calls.

The reconcile function receives a Request (namespace/name), fetches the current object, computes desired state, and applies changes.