What is a controller in k8s?

### What is a Controller?

In Kubernetes, a controller is a control loop that continuously watches the state of your cluster and makes or requests changes to move the current state closer to the desired state.

From a platform engineering perspective, controllers are the implementation of declarative configuration. Instead of executing a linear list of commands (imperative), you provide a "Record of Intent" (the object .spec), and the controller works tirelessly to ensure that reality matches that intent,.

How it works (The Reconciliation Loop):

1. Watch: The controller tracks at least one Kubernetes resource type (e.g., Pods).
   • Example: The Job controller watches for new Job objects.
2. Compare: It compares the actual state (what is running) against the desired state (what is in the YAML).
3. Reconcile: It takes action to fix the difference.
   • Example: If a Job spec asks for a task to run, the Job controller tells the API server to create new Pods to do that work.

This design allows the system to handle failure automatically. If a component fails, the controller notices the deviation from the desired state and attempts to fix it (e.g., restarting a crashed container).

---

How many controllers are there?

There is no single fixed number of controllers in a Kubernetes cluster because they are modular. Controllers are compiled into the kube-controller-manager binary, but specific controllers can be enabled, disabled, or added via Custom Resource Definitions (CRDs) and Operators,.

However, the core kube-controller-manager typically ships with over two dozen built-in controllers responsible for the native behaviors of the cluster-.

Below is a table of the core built-in controllers found in a standard Kubernetes control plane:

Category	Controller Name	Responsibility
Workload	Deployment	Manages rolling updates and scaling for stateless applications.
	ReplicaSet	Ensures the correct number of Pod replicas are running.
	StatefulSet	Manages stateful applications with unique network identities.
	DaemonSet	Ensures a copy of a Pod runs on all (or some) nodes.
	Job / CronJob	Manages one-off tasks and recurring scheduled tasks,.
	Replication	(Legacy) Manages the lifecycle of ReplicationControllers.
Node & Core	Node	Monitors node health, handles evictions, and manages cloud node lifecycle,.
	Namespace	Handles Namespace creation and deletion cleanup.
	ServiceAccount	Creates default ServiceAccounts for new Namespaces.
	Garbage Collector	Cleans up resources (like Pods) that no longer have an owner,.
Network	Service	Manages load balancers and service accessibility,.
	Endpoint / EndpointSlice	Populates Endpoints (links Services to Pod IPs),.
	Route	Sets up network routes in the underlying cloud infrastructure,.
Storage	PersistentVolumeBinder	Binds PersistentVolumeClaims to PersistentVolumes.
	AttachDetach	Manages the attachment and detachment of volumes to nodes.
Security	Certificate	Manages signing of certificates (CSRs).
	RootCACertPublisher	Publishes the root CA certificate.

Note: In cloud environments, some of these loops (like Node, Route, and Service controllers) may be moved out of the kube-controller-manager and into a specialized cloud-controller-manager to integrate with specific cloud provider APIs,.