Multi-Node Architecture Working
The controller + node-agent architecture is functional. Nodes self-register with the controller on boot and report their capabilities (CPU, memory, storage). The controller tracks cluster state in SQLite and routes VM operations to the appropriate node agent via gRPC.
Current topology: one controller (runs on your workstation or a management node) and N compute nodes. Each compute node runs NixOS with libvirtd and the kcore node-agent service.
What's next
- mTLS between controller and agents (currently using insecure gRPC for development)
- Automatic certificate provisioning on node registration
- Node health checks and heartbeat monitoring