問題文
A platform team assumes that the control plane will take over the work of their own controller if it fails, the way it does for the built-in ones. Which distinction does the Kubernetes documentation draw, and what does the team have to plan for?
選択肢
- The distinction does not matter, since controllers are designed to fail safely.
- The resilience described applies to the built-in controllers inside the control plane, while a controller written to extend the cluster runs outside it, so its own availability has to be planned for.
- The control plane takes over any controller that manages a registered kind, because the registration is what tells the control plane which loops it is responsible for keeping alive on behalf of the platform team.
- Controllers written to extend the cluster must run inside the control plane.