Microservices in production
Deploy, scale, and update each service independently.
What is Kubernetes? Container orchestration explained
Kubernetes (K8s) is an open-source platform that orchestrates containers across a cluster of machines: it schedules where they run, restarts failures, exposes services, handles scaling and rolling updates. Born at Google and governed by the CNCF, it is the de facto standard for running containerized apps in production at medium and large scale.
In one sentence
Kubernetes decides where, when, and how many containers run—and brings them back if they fail.
Key points
- A cluster = nodes (machines) + a control plane that schedules Pods.
- Declarative: you describe desired state (YAML/API); Kubernetes converges to it.
- Services, Ingress, ConfigMaps, and Secrets decouple networking and config from code.
- Powerful but complex: many SMEs start with a managed service (GKE, EKS, AKS).
Term at a glance
- Kubernetes
- K8s · Container orchestrator · Kubernetes cluster
- English term
- Kubernetes
- Domain
- Cloud & DevOps
- Category
- Orchestration
- Level
- Intermediate to advanced
What does “Kubernetes” mean exactly?
Docker (or another OCI runtime) runs a container on one machine. Kubernetes coordinates hundreds of containers across many machines: placement, health, scaling, rolling updates, and service discovery.
The model uses Pods (smallest deployable unit), Deployments/StatefulSets, and Services. You declare desired state; controllers fix drift (self-healing).
For a Montreal SME, K8s makes sense once you have multiple services, autoscaling needs, or an internal platform. Before that, Compose, a PaaS, or a managed container service often wins.
How do you adopt Kubernetes?
- 01
Containerize and stabilize
Healthy images, health checks, and externalized config before orchestrating.
- 02
Choose managed vs self-hosted
GKE/EKS/AKS reduce control-plane toil; self-managed needs dedicated ops.
- 03
Describe workloads
Manifests or Helm/Kustomize: Deployments, Services, Ingress, HPA, NetworkPolicies.
- 04
Observability and GitOps
Metrics, logs, traces; deploys via CI and often Argo CD / Flux.
Concrete Kubernetes example
A B2B SaaS in Quebec City runs API, worker, and frontend on GKE. A traffic spike doubles API pods via HPA. A rolling update ships a new version with no downtime. When a node dies, pods are rescheduled—clients see only a brief latency blip.
What is Kubernetes for?
High availability
Replicas, probes, and reschedule on node failure.
Internal platform
Standardize deploys for multiple product teams.
Batch / jobs
Jobs and CronJobs for one-off or scheduled work.
Pros and cons of Kubernetes
- Industry standard, huge CNCF ecosystem
- Native autoscaling, self-healing, rolling updates
- Relative multi-cloud portability
- Fine-grained network, quotas, and policies
- Steep learning curve and ops cost
- Overkill for a single small app
- Many concepts and YAML surfaces
- Cluster security (RBAC, policies) must be mastered
Kubernetes vs Docker Compose?
| Kubernetes | Docker Compose | |
|---|---|---|
| Scope | Multi-node cluster, production scale | Mostly local / single host |
| Self-healing | Reschedule, replicas, probes | Basic service restarts |
| Scaling | HPA, cluster autoscaler | Limited manual scale |
| Complexity | High | Low — great to start |
Why Kubernetes matters for a Quebec SME
When the product moves from “one API on a VPS” to several critical services, managed K8s avoids manual firefighting. In Quebec, talent and partners already speak this language. The costly mistake is adopting too early—or too late without a safety net.
FAQ
Do you need Kubernetes to use Docker?
No. Docker builds and runs containers. K8s orchestrates at cluster scale.
Is Kubernetes free?
The open-source software is; running it (managed cluster, engineering, monitoring) has real cost.
What if we are too small for K8s?
PaaS (Cloud Run, Render, Fly), ECS/Fargate, or Compose on one or two servers.
Is Helm required?
No, but Helm or Kustomize help manage complex manifests and multiple environments.
Related terms
Sources and references
Torn between Compose, a PaaS, and managed Kubernetes? We can scope architecture to your load and team.
Talk about orchestration