Skip to main content

ETCD

Description​

Etcd is a distributed, reliable key-value store for distributed systems. It is used as Kubernetes' backing store for all cluster data. It's designed to be highly available and consistent, ensuring that the cluster state is reliably stored and accessible.

Use-Case​

The primary use-case for etcd in the context of Kubernetes is as the cluster's backing store. Specifically:

  • Kubernetes Backing Store: etcd stores critical Kubernetes cluster state, including:
    • Configuration data (e.g., Deployments, Services)
    • Cluster state (e.g., the current status of pods)
    • Secrets
    • Service discovery information
    • Metadata about pods, services, deployments, etc.

Etcd essentially acts as the central database for a Kubernetes cluster. Without a healthy etcd cluster, the Kubernetes control plane cannot function correctly.

Example​

While etcd itself can be interacted with directly using its API, its primary interaction within Kubernetes is through the Kubernetes API server. Here's a conceptual example of how Kubernetes uses etcd:

  1. kubectl apply -f deployment.yaml: A user runs a kubectl command to create or update a deployment.

  2. API Server: The Kubernetes API Server receives this request.

  3. Validation & Mutation: The API Server validates the request and potentially mutates it based on admission controllers.

  4. etcd Storage: The API Server writes the desired state of the deployment (as a Kubernetes Deployment object) to etcd.

  5. Controllers Watch: Kubernetes controllers (e.g., the Deployment Controller) watch etcd for changes.

  6. Reconciliation: The Deployment Controller notices the new or updated deployment object in etcd. It then takes actions to reconcile the actual state of the cluster with the desired state stored in etcd (e.g., creating or updating ReplicaSets and Pods).

Why It's Important​

etcd is critical for Kubernetes for the following reasons:

  • Centralized State: It provides a single, consistent view of the entire cluster's state. This is essential for coordination and decision-making by Kubernetes components.

  • Reliability & Consistency: etcd is designed for high availability and fault tolerance. It uses the Raft consensus algorithm to ensure data consistency, even if some etcd nodes fail.

  • Durability: Data is persisted to disk, protecting the cluster state in case of control plane restarts.

  • Source of Truth: etcd serves as the authoritative source for the desired state of the cluster. Controllers work to ensure the actual cluster state matches what's stored in etcd. Without etcd, the cluster cannot properly maintain its state or recover from failures.