Kubernetes version support windows: what they mean for a small team
Kubernetes ships a new minor version roughly three times a year, and each minor version receives patches for about fourteen months. Those two numbers decide how much of a small team's year goes into cluster upgrades.
The math
If each version is supported for about fourteen months and new ones arrive every four months or so, you need to upgrade at least once a year just to stay supported. Skipping versions is not allowed for the control plane: you upgrade one minor version at a time.
What one upgrade involves
- Reading the changelog for removed APIs and deprecations.
- Checking every Helm chart and manifest for those APIs.
- Upgrading the control plane, then nodes, one pool at a time.
- Upgrading add-ons: ingress controller, CNI, cert-manager, monitoring. Each has its own compatibility matrix.
- Testing workloads after each step.
A management layer automates the mechanics. It does not remove the reading, checking and testing.
Managed Kubernetes changes the shape, not the duty
Cloud providers handle the control plane, but they also enforce the support window and will eventually upgrade clusters for you. You still own the workloads and add-ons.
The questions to ask
- Who on the team does this, and how many days per year?
- What breaks if an upgrade slips by six months?
- Would a simpler Docker host with a deployment platform need less of that time?
When it is worth it
If you already run several clusters, need fine-grained scheduling or rely on operators, the upgrade work is the price of features you use. If not, it is overhead.
For a direct comparison of a cluster management approach and a Docker-based platform, see Peon vs Rancher. I work on Peon, an open-source self-hosted deployment platform.