Phase 7 -- Long-Term Vision¶
Status: Not started
Items worth tracking but not planned in detail. These become relevant as earlier phases are completed and the platform matures.
| Item | What | When It Makes Sense |
|---|---|---|
| Third compute host | 3-node Proxmox cluster with Ceph or shared storage | When workload density outgrows 2 hosts, or if distributed storage is desired |
| Declarative infrastructure | Crossplane or Cluster API to manage Proxmox VMs and AWS resources from Kubernetes CRDs | When Terraform/Ansible maintenance overhead becomes a pain point, or as a platform engineering learning exercise |
| Multi-cluster GitOps | Single ArgoCD managing staging + production + future clusters | When the second cluster is stable and patterns are proven |
| Dedicated GPU | Low-profile GPU in the MS-01's PCIe x16 slot (e.g., Intel Arc A380) | When iGPU transcoding hits limits, or for ML/AI workloads |
| Full 10G fabric | Replace USW-16-PoE with a switch that has SFP+ uplinks; 10G to NAS | When NFS throughput is a bottleneck or a NAS with 10GbE is added |
| IPv6 | Dual-stack networking across VLANs and Kubernetes | When ISP provides native IPv6 and external access is in use |
| GitOps for network config | Version-control UniFi firewall rules and VLAN config | When network changes are frequent enough to warrant it |
| IoT VLAN | Dedicated VLAN for smart home devices, isolated from household and homelab traffic | When IoT devices are added to the network |
| HPA for workloads | Horizontal Pod Autoscaler for Authentik, Grafana, and high-traffic apps | When traffic patterns justify scaling beyond fixed replica counts |