- Watches the current state of the cluster (from the API server),
- Compares it with the desired state (from YAML manifests),
- Acts to fix differences (create/update/delete resources).
Example:
You define “3 Pods” → if only 2 are running, the controller starts 1 more.
🧩 2. Main Categories of Controllers
There are 3 broad categories:
| Category | Description | Examples |
|---|---|---|
| Workload Controllers | Manage Pods and how applications run. | Deployment, ReplicaSet, StatefulSet, DaemonSet, Job, CronJob |
| Infrastructure Controllers | Manage nodes, networking, namespaces, etc. | Node Controller, Service Controller, Namespace Controller |
| Custom / Operator Controllers | Created by users to manage specific apps or CRDs. | Prometheus Operator, Argo CD Operator, AWS Controllers for Kubernetes (ACK) |
⚙️ 3. Core Workload Controllers (Most Common)
These are the controllers you’ll use daily.
| Controller | Purpose | Typical Use Case |
|---|---|---|
| Deployment | Manages ReplicaSets and performs rolling updates/rollbacks. | Stateless web apps, APIs |
| ReplicaSet | Ensures a specific number of identical Pods are running. | Low-level controller used by Deployment |
| StatefulSet | Ensures unique, ordered Pods with stable storage and network IDs. | Databases (MySQL, MongoDB), Kafka, Zookeeper |
| DaemonSet | Ensures one Pod runs on each node (or selected nodes). | Log collectors, monitoring agents (Prometheus Node Exporter, Fluentd) |
| Job | Runs Pods to completion. | Batch tasks, data processing |
| CronJob | Runs Jobs on a schedule. | Backups, cleanup jobs, periodic reports |
🏗️ 4. Cluster & Infrastructure Controllers
These are system-level controllers running in the kube-controller-manager (on the control plane).
| Controller | Description |
|---|---|
| Node Controller | Detects when nodes go down and manages node lifecycle. |
| Service Controller | Creates or removes cloud load balancers when Services of type LoadBalancer are created. |
| Namespace Controller | Cleans up resources when a namespace is deleted. |
| EndpointSlice Controller | Maintains network endpoints for Services efficiently. |
| PersistentVolume Controller | Manages PersistentVolume and PersistentVolumeClaim binding. |
| PersistentVolumeBinder | Handles dynamic provisioning of storage. |
| ServiceAccount Controller | Creates default service accounts and API tokens. |
| ReplicationController (Legacy) | Older controller replaced by ReplicaSet. |
| Job Controller | Manages Pod creation for Job resources. |
| CronJob Controller | Manages Job scheduling for CronJobs. |
All these run inside one process:
kube-controller-manager
🧰 5. Cloud-Specific Controllers (on Managed Clusters)
When you use EKS, GKE, or AKS, additional controllers integrate Kubernetes with the cloud provider:
| Controller | Role |
|---|---|
| Cloud Controller Manager | Connects Kubernetes with the underlying cloud APIs. |
| Route Controller | Manages networking routes between cluster nodes. |
| AWS Load Balancer Controller | Provisions AWS ALB/NLB for Ingress or Services. |
| External DNS Controller | Automatically manages DNS records in Route53 or Cloud DNS. |
🧬 6. Custom Controllers
Developers can create their own controllers to automate any workflow.
Example:
You define a Custom Resource Definition (CRD) called Database, and a custom controller ensures that:
- When a
Databaseobject is created → a Pod and PVC are provisioned. - When it’s deleted → resources are cleaned up.
This is how Operators are built.
🧠 7. Operator Controllers (Advanced)
Operators are custom controllers that encode domain-specific operational logic.
| Example Operator | What It Manages |
|---|---|
| Prometheus Operator | Deploys and configures Prometheus and Alertmanager |
| Argo CD Operator | Manages Argo CD GitOps setup |
| Kafka Operator | Manages Kafka clusters |
| PostgreSQL Operator | Automates PostgreSQL database deployment |
| AWS Controllers for Kubernetes (ACK) | Manages AWS resources (S3, RDS, etc.) directly from Kubernetes |
Operators use:
- Custom Resources (CRDs)
- Custom Controllers (logic)
- Often written in Go, Python, or with frameworks like Kubebuilder.
🧮 8. Summary Table — All Controller Types
| Type | Example Controllers | Purpose |
|---|---|---|
| Workload Controllers | Deployment, StatefulSet, DaemonSet, Job, CronJob | Manage Pods and app workloads |
| System Controllers | Node, Namespace, Service, Endpoints, PV/PVC | Maintain cluster infrastructure |
| Cloud Controllers | AWS Load Balancer, Route, Cloud Controller Manager | Integrate with cloud provider |
| Custom Controllers | CRD-based logic | Extend Kubernetes |
| Operators | Prometheus Operator, Argo CD Operator | Automate complex apps |
🧩 9. Where They Run
All default controllers (like Deployment, StatefulSet, etc.) are part of the kube-controller-manager process on the control plane.
Custom and Operator controllers run as Pods inside the cluster.
🧠 10. How to See Active Controllers
You can check which controllers are running:
kubectl get pods -n kube-system | grep controller
You might see:
kube-controller-manager-minikube
aws-load-balancer-controller-xxxx
ingress-nginx-controller-xxxx
Top comments (0)