Stop 09 · Containers & OrchestrationTopics · 3/9

Container Orchestration Basics

In short

Running one container with docker run is easy. Running 60 of them across 10 machines, restarting the ones that crash and rolling out updates without downtime is a full-time job. Orchestrators do that job.

Managing many containers across many machines — scheduling, scaling, networking, and self-healing — a problem that becomes unmanageable to handle manually beyond a handful of containers.

The blob character loads colour-sorted tiffin boxes into a crate on a bicycle.

Think of it like this

The dabbawala network

Thousands of tiffins, hundreds of runners, and almost nothing goes missing. Each box has a code, each runner a route. If a runner falls sick, another picks up his boxes. Nobody tracks single tiffins by hand. The system does.

Watch it happen

Desired state, kept true

You (kubectl)

Control plane

Node A

Node B

apply: replicas 3

1/5You don't say where to run things. You declare what you want: three copies of the web app.

The big ideas

What you need to know

Declarative

You write the desired state in YAML. A control loop keeps comparing it with reality and fixes the gap.

Scheduling

The orchestrator decides which machine runs each container based on resources and rules you set.

Self-healing

Crashed containers restart, failed health checks get replaced, lost nodes get their work moved.

Rolling updates

New versions replace old pods a few at a time, and you can roll back if the new one misbehaves.

See it in code

Walk through the code

yaml

1apiVersion: apps/v12kind: Deployment3metadata:4  name: web5spec:6  replicas: 37  selector:8    matchLabels: { app: web }9  template:10    metadata:11      labels: { app: web }12    spec:13      containers:14        - name: web15          image: nginx:1.2716          ports: [{ containerPort: 80 }]

1/4A Deployment named web: Kubernetes' standard way to run stateless apps.

Save as web.yaml and run kubectl apply -f web.yaml. Then kubectl delete pod one of them and watch a new one appear.

Save yourself the pain

Common mistakes

Oops

Using Kubernetes for a single small app on one server.

Do this instead

Orchestration pays off with many services or machines. One app may be happier on Cloud Run or a single VM.

Oops

SSHing into a node to restart a container by hand.

Do this instead

Change the YAML and apply it. Manual fixes get undone by the control loop anyway.

Your mission

Watch self-healing

~30 min

Quick check

You delete one of three pods managed by a Deployment. What happens?

Go deeper

1List 3 problems orchestration solves that become painful to handle manually at scale.
2Learn the basic concepts shared across orchestrators: scheduling, desired state, and self-healing.
3Understand why orchestration exists as a distinct layer above just 'running containers'.