Proposal: Simplify All-in-one Implementation with kine(sqlite) + k8s-apiserver
Background: Based on the design principles of the All-in-one architecture (abstracting Kubernetes concepts while maintaining Kubernetes API compatibility), we propose re-evaluating the implementation approach described in the current design documentation.
Key Observations:
​User Perspective: All-in-one users expect a Kubernetes-transparent experience without interacting with low-level k8s components.
​Design Principle: The system should maximally reuse Kubernetes native concepts and APIs to ensure compatibility.
​Implementation Suggestion: Using kine (with SQLite) + kube-apiserver might provide a cleaner implementation compared to the current documented approach, as:
Kine already implements the Kubernetes storage interface (etcd v3 API)
SQLite embedded database aligns with All-in-one's single-node deployment scenario
Reduced maintenance cost through reuse of upstream Kubernetes components
Perhaps it's a good idea to migrate the apiserver component in standalone to kine, but we need to check if the kine deployment is lightweight enough.
I'd like to hear your opinion @CH3CHO
Perhaps it's a good idea to migrate the apiserver component in standalone to kine, but we need to check if the kine deployment is lightweight enough.
I'd like to hear your opinion @CH3CHO
I think it's a good idea. We can dig into it.
It looks like kine and k3s only support Linux. So we will still need to rely on a docker container environment to do the visualization for us.
maybe we no need k3s but noly need kine here
maybe we no need k3s but noly need kine here
However, if we use k3s to simulate a K8s environment, there will be much easier.
easier
yes, but if use k3s here,we also need docker in docker ,that's not easy