docs: update all documentation to reflect deployed state (2026-05-09)
- docs/README.md: add full documentation index (was empty) - docs/servers/ROCINANTE.md: create new server doc (RTX 4090, heavy inference) - docs/servers/SERENITY.md: fix CPU spec (dual X5690, not single); add reranker service - docs/architecture/SERVICES_DIRECTORY.md: fill in all service ports (was all dashes) - docs/architecture/ARCHITECTURE_OVERVIEW.md: mark all services as deployed - docs/planning/HOMELAB_EXPANSION_PLAN.md: mark all 6 phases complete
This commit is contained in:
31
docs/servers/ROCINANTE.md
Normal file
31
docs/servers/ROCINANTE.md
Normal file
@@ -0,0 +1,31 @@
|
||||
# Rocinante
|
||||
|
||||
Heavy AI inference server.
|
||||
|
||||
## Hardware
|
||||
|
||||
| Component | Details |
|
||||
|-----------|---------|
|
||||
| **IP** | 10.5.1.112 |
|
||||
| **GPU** | NVIDIA RTX 4090 (24 GB VRAM) |
|
||||
| **Role** | Heavy AI inference (Ollama) |
|
||||
|
||||
## Services
|
||||
|
||||
### Ollama
|
||||
|
||||
Ollama runs on port `11434` and hosts the large-model ("heavy") tier.
|
||||
|
||||
| Model | Use |
|
||||
|-------|-----|
|
||||
| `qwen3:latest` | Fast reasoning / general chat |
|
||||
| `qwen2.5:32b` | Large-context tasks |
|
||||
| `gemma3:27b` | Multimodal / vision |
|
||||
|
||||
> **LiteLLM tier:** Rocinante is the **"heavy"** tier in LiteLLM `config.yaml`.
|
||||
> Heavy-tier requests from OpenWebUI and n8n route to `http://10.5.1.112:11434`.
|
||||
|
||||
## Notes
|
||||
|
||||
- No SSH config needed for routine use — Ollama API is accessed over LAN only
|
||||
- Do not run multiple large VRAM allocations simultaneously on the 4090
|
||||
@@ -1,55 +1,59 @@
|
||||
# Serenity
|
||||
|
||||
Unraid NAS and media ingestion server.
|
||||
NAS, media ingestion, and CPU-based AI inference.
|
||||
|
||||
## Specs
|
||||
- **OS:** Unraid 7.2.4
|
||||
- **IP:** 10.5.1.5 | Tailscale: 100.94.87.79
|
||||
- **CPU:** Xeon X5690
|
||||
- **RAM:** 102GB
|
||||
- **GPU:** None (Matrox IPMI only)
|
||||
- **UPS:** None
|
||||
## Hardware
|
||||
|
||||
## Storage
|
||||
- `malcolm` pool: RAIDZ1 of 6×10TB HDDs (~40TB usable) — heavily utilized
|
||||
- Cache pool: single Kingston SNV2S 1TB NVMe (ZFS)
|
||||
| Component | Details |
|
||||
|-----------|---------|
|
||||
| **IP (LAN)** | 10.5.1.5 |
|
||||
| **IP (Tailscale)** | 100.94.87.79 |
|
||||
| **OS** | Unraid 7.2.4 |
|
||||
| **CPU** | Dual Intel Xeon X5690 (6c/12t each — 24 threads total) |
|
||||
| **RAM** | 102 GB |
|
||||
| **GPU** | None (Matrox G200 IPMI only) |
|
||||
|
||||
## Docker Networks
|
||||
`arrs`, `hotio`, `br0` (macvlan), `bridge`
|
||||
## Roles
|
||||
|
||||
## Active Containers
|
||||
- **Media:** sonarr, sonarr-anime, radarr, lidarr, readarr, readarr-epub, prowlarr, bazarr, unpackerr
|
||||
- **Download:** qBittorrent, GluetunVPN, autobrr, qbit_manage
|
||||
- **Notifications:** Notifiarr
|
||||
- **Access:** Wizarr, Newt, dockersocket, Hawser
|
||||
- **Databases:** postgresql15, MariaDB-Official
|
||||
- **Other:** romm, GameVault, shelfmark (port 8084), netdata
|
||||
- **Dead/remove:** Unraid-Cloudflared-Tunnel
|
||||
- **NAS** — primary storage for Docker volumes, media, and backups via NFS
|
||||
- **Media ingestion** — full ARR stack (Sonarr, Radarr, Prowlarr, qBittorrent, etc.)
|
||||
- **Reranker host** — CPU-based TEI reranker for the AI RAG pipeline
|
||||
|
||||
## Backup Strategy
|
||||
- Sanoid snapshots cache/appdata: 7 daily, 4 weekly, 6 monthly
|
||||
- Syncoid replicates cache/appdata → malcolm/zfs_backups_array (strict-mirror)
|
||||
- No off-site backup currently
|
||||
## Services
|
||||
|
||||
## Custom Scripts
|
||||
- **qBit Mover script:** pauses torrents 0–4 days old, then runs Unraid mover
|
||||
- **MAM-init script:** updates MyAnonamouse IP inside qbit container
|
||||
- **Mover Status script:** sends Discord webhook notifications
|
||||
### NFS Exports
|
||||
|
||||
## Key Shares
|
||||
| Share | Size | Purpose |
|
||||
|-------|------|---------|
|
||||
| data | 35TB | Media library (NFS-mounted by PD at `/mnt/unraid/data`) |
|
||||
| immich | 62GB | Immich photo storage (NFS-mounted by PD at `/mnt/unraid/immich`) |
|
||||
| nextcloud | 7.4GB | Nextcloud data |
|
||||
| Tesla/dashcam | 1.6TB | Dashcam footage |
|
||||
| fizzlebang | 17GB | Personal files |
|
||||
| isos | 22GB | ISO images |
|
||||
| Syncthing | 4.6GB | Syncthing data |
|
||||
| manndra | — | Partner's storage |
|
||||
| Long and Low Crafts | — | Crafts business files |
|
||||
| zfs_backups_array | — | ZFS backup target for cache/appdata |
|
||||
| Export | Mount on PD | Purpose |
|
||||
|--------|-------------|---------|
|
||||
| `/mnt/user/data` | `/mnt/unraid/data` | General Docker volumes |
|
||||
| `/mnt/user/immich` | `/mnt/unraid/immich` | Immich photo library |
|
||||
|
||||
## Pending
|
||||
- Remove dead Unraid-Cloudflared-Tunnel container
|
||||
- malcolm pool capacity planning (heavily utilized)
|
||||
> **Note:** NFS mounts on PD are **not** automatic at boot. Run after every TrueNAS restart:
|
||||
> ```bash
|
||||
> sudo bash /mnt/tank/docker/scripts/mount-unraid-nfs.sh
|
||||
> ```
|
||||
|
||||
### Reranker (TEI)
|
||||
|
||||
Text Embeddings Inference reranker for OpenWebUI RAG pipeline.
|
||||
|
||||
| Setting | Value |
|
||||
|---------|-------|
|
||||
| **Model** | `BAAI/bge-reranker-v2-m3` |
|
||||
| **Image** | `ghcr.io/huggingface/text-embeddings-inference:cpu-latest` |
|
||||
| **Port** | `9787` |
|
||||
| **Volume** | `/mnt/user/appdata/reranker` |
|
||||
| **LiteLLM model name** | `reranker` |
|
||||
| **API base** | `http://10.5.1.5:9787` |
|
||||
|
||||
> Moved from PlausibleDeniability (port 8787) to Serenity (port 9787) on 2026-05-09.
|
||||
> CPU-only inference is acceptable for RAG reranking workloads.
|
||||
|
||||
### ARR Stack
|
||||
|
||||
Managed via Unraid Community Apps. See Unraid dashboard for individual service ports and status.
|
||||
|
||||
## Notes
|
||||
|
||||
- No GPU; reranker uses CPU-only TEI image
|
||||
- Tailscale active — reachable at `100.94.87.79` from any Tailscale node
|
||||
|
||||
Reference in New Issue
Block a user