Update docs to match live cluster state

Audited the running cluster and fixed all .md files:
- Node info: Fedora 43, Lima (not OrbStack), worker IP 10.0.1.58
- Networking: fixed public/internal hostname tables, all *.dog internals
- Storage: removed Longhorn refs (not deployed), documented hostPath/local-path
- Services: moved Seerr to media chart, utils is Zerobyte only
- Bootstrap: reordered steps, MetalLB/traefik-internal as manual pre-deploy
- Headlamp.md/MetalLB.md: added context and explanations

Made-with: Cursor
This commit is contained in:
Alvin Wang 2026-04-22 14:59:34 -07:00
parent 2dfa47403d
commit 4301877f33
4 changed files with 117 additions and 77 deletions

View File

@ -1,10 +1,9 @@
NAME STATUS ROLES AGE VERSION INTERNAL-IP EXTERNAL-IP OS-IMAGE KERNEL-VERSION CONTAINER-RUNTIME
dogbox Ready control-plane 3h31m v1.34.6+k3s1 10.0.1.2 <none> Fedora Linux 40 (Server Edition) 6.9.6-200.fc40.x86_64 containerd://2.2.2-bd1.34
mac-worker Ready <none> 3h13m v1.34.6+k3s1 192.168.139.12 <none> Ubuntu 25.10 6.17.8-orbstack-00308-g8f9c941121b1 containerd://2.2.2-bd1.34
localhost.localdomain Ready control-plane v1.34.6+k3s1 10.0.1.2 <none> Fedora Linux 43 (Server Edition) 6.17.1-300.fc43.x86_64 containerd://2.2.2-bd1.34
lima-mac-worker Ready <none> v1.34.6+k3s1 10.0.1.58 <none> Ubuntu 25.10 6.17.0-22-generic containerd://2.2.2-bd1.34
The mac-worker is running inside orbstack linux VM if that matters.
The mac-worker is running inside a Lima VM on macOS.
I have a DNS rewrite pointing *.internal to 10.0.1.250 which is traefik-internal.
@ -24,8 +23,8 @@ separated by `loadBalancerClass` so they don't conflict.
request a pool via the `metallb.io/address-pool` annotation.
| Service | loadBalancerClass | LB | External IPs |
|------------------|-------------------|----------|-------------------------|
| traefik | (none) | klipper | node IPs (10.0.1.2 etc) |
|------------------|-------------------|----------|---------------------------|
| traefik | (none) | klipper | 10.0.1.2, 10.0.1.58 |
| traefik-internal | metallb | MetalLB | 10.0.1.250 |
`loadBalancerClass` is immutable on k8s Services. Changing it requires deleting

View File

@ -1,6 +1,22 @@
# Headlamp — Manual Token Access
The `charts/headlamp` Helm chart deploys Headlamp with its own in-cluster
ServiceAccount (`headlamp`) and a `cluster-admin` ClusterRoleBinding. That
SA is used by the running pod and does not require manual setup.
To generate a **bearer token** for logging in to the Headlamp UI (e.g. from
a browser), create a separate short-lived token:
```bash
kubectl -n apps create token headlamp --duration=48h
```
If you need a dedicated SA for external/long-lived access instead:
```bash
kubectl -n apps create serviceaccount headlamp-admin
kubectl create clusterrolebinding headlamp-admin \
--serviceaccount=apps:headlamp-admin \
--clusterrole=cluster-admin
kubectl -n apps create token headlamp-admin
kubectl -n apps create token headlamp-admin
```

View File

@ -1,7 +1,17 @@
# MetalLB — Manual Setup
MetalLB is **not** included in `bootstrap.sh deploy`. It must be installed
manually before deploying `traefik-internal` (which depends on the MetalLB
`loadBalancerClass`).
```bash
helm repo add metallb https://metallb.github.io/metallb
helm repo update
helm search repo metallb
helm dependency build charts/metallb
helm upgrade --install metallb charts/metallb -n kube-system --wait
```
The chart wraps the upstream MetalLB subchart and adds a custom
`IPAddressPool` + `L2Advertisement` (defined in `charts/metallb/templates/pool.yaml`).
The pool assigns a single IP (`10.0.1.250`) with `autoAssign: false`, so only
services that explicitly request the `internal` pool via annotation get that IP.

121
README.md
View File

@ -1,6 +1,6 @@
# Homelab — k3s Cluster
2-node k3s cluster (1 manager, 1 worker) running a self-hosted homelab stack on `ratboo.me`.
2-node k3s cluster (1 control-plane, 1 worker) running a self-hosted homelab stack on `ratboo.me`.
## Architecture
@ -8,8 +8,8 @@
| Node | Role | OS | IP | Runtime |
|------|------|----|----|---------|
| **dogbox** | control-plane | Fedora 40 Server | `10.0.1.2` | k3s server + containerd |
| **mac-worker** | worker | Ubuntu 25.10 (OrbStack VM) | `192.168.139.12` | k3s agent + containerd |
| **localhost.localdomain** (dogbox) | control-plane | Fedora Linux 43 (Server Edition) | `10.0.1.2` | k3s v1.34.6 + containerd |
| **lima-mac-worker** | worker | Ubuntu 25.10 (Lima VM on macOS) | `10.0.1.58` | k3s v1.34.6 + containerd |
### Overview
@ -20,62 +20,70 @@
*.ratboo.me
┌──────────────────────────┼──────────────────────────┐
dogbox (manager)
│ Fedora 40 · 10.0.1.2
localhost.localdomain (dogbox)
│ Fedora 43 · 10.0.1.2
│ │
│ ┌─────────────────┐ ┌──────────────────────┐ │
│ │ k3s server │ │ Traefik (k3s) │ │
│ │ control-plane │ │ :443 websecure │ │
│ └─────────────────┘ │ Let's Encrypt + CF │ │
│ └──────────┬───────────┘ │
│ ┌─────────────────┐ │ │
│ ┌─────────────────┐ │
│ │ traefik-internal │ Routes to pods across │
│ │ :80 LB 10.0.1.250│ both nodes via CNI │
│ │ (MetalLB L2) │ │ │
│ └─────────────────┘ │ │
│ Longhorn │ │
└──────────────┬─────────────────────┼─────────────────┘
│ │ :80/:443 MetalLB │ both nodes via CNI │
│ │ LB 10.0.1.250 │ │ │
│ └──────────────────┘ │ │
└───────────────┬───────────────────┼──────────────────┘
│ │
NFS /dogstore k3s cluster
│ │
┌──────────────┴─────────────────────┼─────────────────┐
│ mac-worker (worker)
│ Ubuntu 25.10 · OrbStack VM
192.168.139.12
┌──────────────┴───────────────────┼──────────────────┐
lima-mac-worker (worker) │
│ Ubuntu 25.10 · Lima VM on macOS
10.0.1.58
│ │
Longhorn · workload pods
└──────────────────────────────────────────────────────┘
workload pods
└──────────────────────────────────────────────────────
```
### Networking
**Public ingress** — k3s bundles Traefik, configured via `HelmChartConfig` in `traefik-config`. TLS terminates at Traefik using Let's Encrypt with Cloudflare DNS-01 challenge. HTTP automatically redirects to HTTPS.
**Public ingress** — k3s bundles Traefik, configured via `HelmChartConfig` in `traefik-config`. TLS terminates at Traefik using Let's Encrypt with Cloudflare DNS-01 challenge. HTTP automatically redirects to HTTPS. klipper (servicelb) exposes the public Traefik on every node IP.
| Public hostname | Service |
|-----------------|---------|
| `plex.ratboo.me` | Plex |
| `sonarr.ratboo.me` | Sonarr |
| `radarr.ratboo.me` | Radarr |
| `watch.ratboo.me` | Seerr |
| `paperless.ratboo.me` | Paperless-ngx |
| `mealie.ratboo.me` | Mealie |
| `watch.ratboo.me` | Seerr |
**Internal ingress** — A separate Traefik instance (`traefik-internal`) listens on `10.0.1.250:80`, served by MetalLB L2. A DNS rewrite points `*.internal` to that IP. Internal services use Traefik `IngressRoute` CRDs with `ingressClass: traefik-internal`.
**Internal ingress** — A separate Traefik instance (`traefik-internal`) listens on `10.0.1.250` (ports 80 and 443), served by MetalLB L2. A DNS rewrite points `*.internal` to that IP. Internal services use Traefik `IngressRoute` CRDs with `ingressClass: traefik-internal`. Every service with a `*-ingressroute.yaml` template gets an `*.dog` hostname on this Traefik.
| Internal hostname | Service |
|-------------------|---------|
| `homepage.rat` | Homepage |
| `glance.rat` | Glance |
| `plex.dog` | Plex |
| `sonarr.dog` | Sonarr |
| `radarr.dog` | Radarr |
| `bazarr.dog` | Bazarr |
| `prowlarr.dog` | Prowlarr |
| `qbittorrent.dog` | qBittorrent |
| `seerr.dog` | Seerr |
| `paperless.dog` | Paperless-ngx |
| `mealie.dog` | Mealie |
| `homepage.dog` | Homepage |
| `glance.dog` | Glance |
| `headlamp.dog` | Headlamp |
| `zerobyte.dog` | Zerobyte |
**Cluster-only (no ingress):** Prowlarr, Bazarr, qBittorrent, Zerobyte.
**No ingress:** unpackerr (background download-extraction daemon, no web UI).
### Storage
| Mechanism | Use |
|-----------|-----|
| **Longhorn** (`storageClass: longhorn`, replica count 2) | Small config/state PVCs — Traefik ACME (128Mi), app configs (120Gi), Paperless Postgres/Redis, Mealie data, Seerr, Zerobyte |
| **NFS via hostPath `/dogstore`** | Large/shared data — Plex media + transcode, Sonarr/Radarr/qBittorrent/unpackerr data trees, Paperless documents, Homepage/Glance configs |
| **NFS via hostPath `/dogstore`** | Large/shared data — Plex media + transcode, Sonarr/Radarr/qBittorrent/unpackerr data trees, Paperless documents, Homepage/Glance configs, ACME cert storage |
| **hostPath `/home/alvin/service-data`** | App config directories on dogbox (Seerr, etc.) |
| **local-path (default StorageClass)** | k3s built-in provisioner for any PVCs (rancher.io/local-path) |
### Secrets
@ -85,24 +93,23 @@ SOPS + age encryption. All secrets live in `secrets/secrets.enc.yaml`, encrypted
| Namespace | Contents |
|-----------|----------|
| `kube-system` | k3s Traefik, `traefik-config` (HelmChartConfig + redirect middleware) |
| `longhorn-system` | Longhorn storage |
| `media` | Plex, Sonarr, Radarr, Bazarr, Prowlarr, qBittorrent, unpackerr |
| `apps` | Mealie, Homepage, Glance, Headlamp, Seerr, Zerobyte, Paperless-ngx + Postgres + Redis |
| `kube-system` | k3s Traefik + `traefik-config` (HelmChartConfig + redirect middleware), `traefik-internal`, MetalLB controller + speakers, CoreDNS, metrics-server |
| `media` | Plex, Sonarr, Radarr, Bazarr, Prowlarr, qBittorrent, unpackerr, Seerr |
| `apps` | Paperless-ngx + Postgres + Redis, Mealie, Homepage, Glance, Headlamp, Zerobyte |
## Services
| Chart | Namespace | Services | Notes |
|-------|-----------|----------|-------|
| traefik-config | kube-system | Traefik HelmChartConfig overlay | Cloudflare DNS-01, ACME on Longhorn |
| traefik-internal | | Internal Traefik instance | LB via MetalLB at `10.0.1.250` |
| metallb | — | MetalLB L2 pool | Single-IP pool for internal LB |
| media | media | Plex, Sonarr, Radarr, Bazarr, Prowlarr, qBittorrent, unpackerr | Media stack with `/dogstore` data paths |
| traefik-config | kube-system | Traefik HelmChartConfig overlay | Cloudflare DNS-01, ACME on hostPath `/dogstore` |
| traefik-internal | kube-system | Internal Traefik instance | LB via MetalLB at `10.0.1.250`, ports 80/443/9095 |
| metallb | kube-system | MetalLB L2 pool | Single-IP pool (`10.0.1.250`) for internal LB |
| media | media | Plex, Sonarr, Radarr, Bazarr, Prowlarr, qBittorrent, unpackerr, Seerr | Media stack with `/dogstore` data paths |
| paperless | apps | Paperless-ngx, Redis, PostgreSQL | Postgres 15, Redis 7 |
| mealie | apps | Mealie (v3.14.0) | Gemini API integration for recipes |
| mealie | apps | Mealie (v3.16.0) | Gemini API integration for recipes |
| dashboards | apps | Homepage, Glance | Internal-only via `traefik-internal` |
| headlamp | apps | Headlamp | K8s dashboard, internal-only via `traefik-internal` |
| utils | apps | Seerr, Zerobyte | Seerr public, Zerobyte cluster-only |
| utils | apps | Zerobyte | Backup service, internal-only via `traefik-internal` |
## Prerequisites
@ -126,13 +133,7 @@ This prints the worker join command at the end.
K3S_URL="https://<manager-ip>:6443" K3S_TOKEN="<token>" ./scripts/bootstrap.sh agent
```
### 3. Install Longhorn
```bash
./scripts/bootstrap.sh longhorn
```
### 4. Set up SOPS encryption
### 3. Set up SOPS encryption
Generate an age keypair (run on each node):
@ -147,39 +148,53 @@ Copy the public key into `.sops.yaml`, replacing the placeholder. Then encrypt y
sops -e -i secrets/secrets.enc.yaml
```
### 5. Apply secrets
### 4. Apply secrets
```bash
./scripts/bootstrap.sh apply-secrets
```
### 6. Deploy all charts
### 5. Deploy MetalLB and internal Traefik (manual)
These are deployed separately before the main charts because other services depend on them:
```bash
helm dependency build charts/metallb
helm upgrade --install metallb charts/metallb -n kube-system --wait
helm upgrade --install traefik-internal charts/traefik-internal -n kube-system --wait
```
### 6. Deploy all application charts
```bash
./scripts/bootstrap.sh deploy
```
This installs (in order): `traefik-config`, `media`, `paperless`, `mealie`, `dashboards`, `utils`, `headlamp`.
Or deploy individually:
```bash
helm upgrade --install metallb charts/metallb -n kube-system --wait
helm upgrade --install traefik-internal charts/traefik-internal -n kube-system --wait
# Traefik config goes in kube-system (managed by k3s)
helm upgrade --install traefik-config charts/traefik-config -n kube-system
kubectl create namespace apps
helm upgrade --install headlamp charts/headlamp -n apps
helm upgrade --install dashboards charts/dashboards -n apps
helm upgrade --install paperless charts/paperless -n apps
helm upgrade --install mealie charts/mealie -n apps
helm upgrade --install utils charts/utils -n apps
kubectl create namespace media
helm upgrade --install media charts/media -n media
```
helm upgrade --install utils charts/utils -n apps
### Optional: Install Longhorn
The bootstrap script includes a Longhorn install command, but it is not currently deployed:
```bash
./scripts/bootstrap.sh longhorn
```
## Verifying
@ -190,6 +205,7 @@ kubectl get pods -A
# Check ingress routes
kubectl get ingress -A
kubectl get ingressroute -A
# Test a specific service
curl -I https://mealie.ratboo.me
@ -216,13 +232,12 @@ homelab/
│ ├── traefik-config/ # k3s Traefik overrides (HelmChartConfig)
│ ├── traefik-internal/ # Separate internal Traefik instance
│ ├── metallb/ # MetalLB L2 for internal LB IP
│ ├── media/ # Plex, *arr stack, qBittorrent, unpackerr
│ ├── media/ # Plex, Sonarr, Radarr, Bazarr, Prowlarr, qBittorrent, unpackerr, Seerr
│ ├── paperless/ # Paperless-ngx + Postgres + Redis
│ ├── mealie/ # Mealie recipe manager
│ ├── dashboards/ # Homepage + Glance (internal only)
│ ├── headlamp/ # Headlamp K8s dashboard (internal only)
│ └── utils/ # Seerr + Zerobyte
│ └── utils/ # Zerobyte backup
└── secrets/
└── secrets.enc.yaml
```