Mini Kubernetes Control Plane v3.0.0 — Leader Election, Informers, and Scheduling

Mini Kubernetes Control Plane v3.0.0

This release completes the core control-plane architecture by introducing leader-elected controllers, resilient watch-based informers, and a replica scheduler. The system now follows Kubernetes control-plane patterns end-to-end.

Summary

Version 3.0.0 focuses on correctness, convergence, and operational safety. The control plane now supports leader election using etcd leases, MVCC-safe status updates, informer-driven reconciliation, and deterministic replica placement.

Changes

• Added leader election using etcd leases to ensure single-active controller behavior.
• Implemented watch-based informers with safe handling of update and delete events.
• Enforced MVCC-correct read-modify-write semantics for all status updates.
• Added a scheduler that assigns replicas to nodes based on desired state.
• Refined reconciliation logic to re-fetch latest objects before mutation.
• Improved separation between API server, controllers, scheduler, and storage layers.

Architecture

• API server persists desired state to etcd.
• Informers consume etcd watch events and propagate changes.
• Leader-elected controllers reconcile desired and observed state.
• Scheduler performs placement decisions and updates status atomically.
• All writes are guarded by resourceVersion to prevent lost updates.

Notes

This release emphasizes architectural completeness and control-plane correctness rather than feature breadth. It establishes a stable foundation aligned with Kubernetes controller and scheduling semantics.

The project remains experimental and is intended for learning and exploration of control-plane internals.