What is Kubernetes and when do you need it?
Kubernetes automates running containers on multiple servers, but it also brings management effort that not every team needs.
Kubernetes is open-source software that runs containers across a group of servers, restarts them on failure and scales up the number of copies. You need it with many components, multiple teams or varying load. For a few applications, a simpler platform is often enough.
Who is it for: Developers, devops engineers and technical buyers who want to know whether Kubernetes fits their application and team.
What is Kubernetes?
A container is a packaged application with everything it needs to run, so that it works the same everywhere. Kubernetes is an orchestrator: software that decides which containers run on which server and keeps watch over that. You describe the desired state, for example three copies of a service, and Kubernetes works towards it.
The system consists of a control layer, which makes decisions, and worker nodes, where the containers run. The smallest unit is a pod: one or more containers that are scheduled together. A deployment describes how many pods of a version should run and how you update.
What does Kubernetes solve?
It restarts containers that crash, spreads traffic across multiple copies, rolls out new versions gradually and scales the number of pods with load. You manage network, configuration, secrets and storage with the same descriptive files, which fits infrastructure as code: you record the setup in version control instead of clicking by hand.
The platform is also widely supported, so you can move work between environments. Keep in mind that storage, networking and access management differ per environment.
When do you need it?
Kubernetes usually pays off when you manage many services, multiple teams use a shared platform, you deploy regularly, or traffic varies strongly. Then the automation outweighs the learning curve.
- Many separate services that need to scale and roll out independently
- Multiple teams that want a shared, standardised platform
- A need for automated recovery and gradual rollout
- A wish not to be tied to a single environment
What does it cost in management?
Kubernetes requires knowledge of networking, security, storage and monitoring. Platform updates, certificates and access rights remain management work, even if a provider supplies the control layer. Consider a simpler route when you run one or a few applications: a virtual machine with containers, or a platform that takes care of running them for you.
Key terms in plain language
| Term | What is it | Why it is relevant |
|---|---|---|
| Container | Packaged application with dependencies | Ensures the same behaviour in every environment |
| Pod | Smallest unit that gets scheduled | This is where your containers run |
| Node | Server that runs pods | Determines capacity and failure risk |
| Deployment | Description of the desired version and number | Makes rolling out and rolling back repeatable |
| Service and ingress | Fixed address and access from outside | Distribute traffic and determine what is reachable |
Who is this less suitable for?
- Small teams with one or two applications often have less management work left with a simpler platform.
- Applications that are strongly tied to a single server or expect a file system with specific requirements first need a redesign.
- Teams without time for monitoring, security and updates risk ending up with a platform that nobody manages well.
Frequently asked questions
Is Kubernetes the same as Docker?
No. Docker is a way to build and run containers. Kubernetes manages many containers across multiple servers.
Can I run databases on Kubernetes?
It is possible, but it requires attention to persistent storage, backups and recovery. Many teams choose a separate, managed database service.
What is a manifest?
A text file, usually in YAML, that describes the desired state of a component. You keep it in version control.
How long does it take to learn Kubernetes?
It depends on your experience with containers and networking. Allow time to practise in a test environment before you touch production.
