Container orchestration

Container orchestration is the automated work of running containers across a cluster of machines: deciding which machine runs each container, restarting the ones that fail, scaling them with load and routing traffic to them. You describe what should run, and the orchestrator keeps the cluster matching that description. Kubernetes is the most widely used orchestrator.

How it works

A control plane stores desired state, such as "three copies of the checkout service, each with 500m CPU". A scheduler picks a node for each container based on free resources and constraints. Controllers then compare desired and actual state in a loop and act on the difference.

For example, a Deployment declares replicas: 3. If node-2 fails and takes one replica with it, the controller sees 2 of 3 running and starts a replacement on a healthy node, with no one paged to do it by hand.

Why it matters in production

Orchestration removes the manual work of placing and restarting containers, but it adds its own systems to run: the control plane, cluster upgrades, networking, autoscaling and observability. Most production incidents on orchestrated platforms come from those layers, not from the containers themselves.