Files
home-network-k3s-impl/platform/vpn-egress/README.md
T
jakman85andClaude Opus 4.8 34f41a0fba phase3(vpn): egress-class conformance fixtures + steering design finding
Author the internet/vpn egress-class fixture pods for the egress-class + vpn-killswitch
conformance tests. Record the datapath finding: ADR-0013's preferred steering (Cilium
egress gateway -> gluetun tun0) is NOT viable because gluetun must run hostNetwork to
put tun0 on the node, and gluetun then redirects the NODE's default route through the
VPN (breaks k3s/etcd). Escalated: remaining viable path is option 2 (gateway-pod
routing). No steering datapath committed until the mechanism is chosen + validated.

Co-Authored-By: Claude Opus 4.8 <[email protected]>
2026-08-24 00:26:26 -05:00

71 lines
5.4 KiB
Markdown

# platform/vpn-egress — shared VPN egress gateway (Phase 3, ADR-0013)
Single shared **Gluetun** gateway: **PIA over OpenVPN-UDP** (ADR-0013 amendment — Gluetun does not
support PIA over WireGuard), built-in kill-switch, native port forwarding. `vpn`-class services
(sonarr, radarr, jackett, qbittorrent) egress **only** through this tunnel; `internet`-class services
go out the normal WAN. Serves **R-NET-05/06/07**.
> **STATUS: enabled in `platform/kustomization.yaml` (Phase 3).** `pia-credentials` SOPS secret is
> provisioned. Remaining work: validate the tunnel comes up + correct egress IP, then the
> egress-steering mechanism below (the hard part) and the kill-switch conformance test.
## Pieces
- `gluetun-gateway.yaml` — the Gluetun Deployment (`replicas: 1`, `Recreate`) + the `gluetun-control`
Service (control API :8000 for forwarded-port lookup + tunnel-state health). Kill-switch = `FIREWALL=on`.
- `pia-credentials` (SOPS) — PIA username/password only; Gluetun ships PIA's OpenVPN configs.
## Operator hand-off (when deploying Phase 3)
1. Provision `pia-credentials` (see `bootstrap/sops/secret-templates/pia-credentials.yaml.example`,
namespace `vpn-egress`, keys `username`/`password`) → `sops -e -i` → uncomment it in `kustomization.yaml`.
2. cluster-vars (already defined in `clusters/base/cluster-vars.yaml`; a ConfigMap — values are plain
strings under `data:`, injected into the `${...}` tokens by Flux `postBuild.substituteFrom`). Confirm
or override:
- `PIA_REGION` — a **port-forwarding-capable** PIA region (default `Sweden`; e.g. `Netherlands`,
`CA Toronto`, `Switzerland`; NOT a US region — those lack PF). Gluetun `SERVER_REGIONS`.
- `CLUSTER_POD_CIDR` — the **Cilium cluster-pool** pod CIDR (default `10.0.0.0/8`; pods get `10.x`,
confirmed live — NOT the k3s `10.42.0.0/16`). Gluetun `FIREWALL_OUTBOUND_SUBNETS`.
Per-env overrides go in `clusters/<env>/kustomization.yaml` as a cluster-vars JSON patch
(`- op: replace, path: /data/PIA_REGION, value: "…"`). With the self-managing root, a `git push`
auto-applies (no manual `kubectl apply -k`).
3. Enable `- vpn-egress` in `platform/kustomization.yaml` (Phase 3).
## Egress steering — the DESIGN DECISION (needs live validation)
The hard part: routing `vpn`-class pods through this ONE shared tunnel. Cilium Egress Gateway
(`CiliumEgressGatewayPolicy`) SNATs matched pods out of a **gateway node's interface** — it does not
natively route through a *pod's* tunnel. Options considered:
1. **Cilium Egress Gateway → the Gluetun node's tunnel interface (recommended).** Pin Gluetun to a
node; a `CiliumEgressGatewayPolicy` selects pods labeled `homelab.keithjackson.dev/egress-class=vpn`
and egresses them via that node's tunnel interface (`tun0`, OpenVPN). If the tunnel drops, `tun0`
is gone → traffic has no egress path → **fail-closed** (the kill-switch behavior R-NET-06 needs). This is the
most Cilium-native fit but requires Gluetun's tunnel interface to be visible to the node datapath
(hostNetwork Gluetun, or a shared netns) — the detail to VALIDATE.
2. **Gateway-pod routing:** client pods default-route to the Gluetun pod IP (init-container +
NET_ADMIN, or a Multus secondary interface); Gluetun forwards+NATs via the tunnel. Works without
Cilium egress gateway but is more moving parts and pod-IP-fragile.
3. **Per-pod Gluetun sidecar** — explicitly REJECTED by ADR-0013 (single shared gateway preferred).
### FINDING (2026-08-23): option 1 is NOT viable with Gluetun — needs an architect decision
With the tunnel now live (custom-mode Gluetun, PIA Stockholm egress), option 1 fails on inspection:
`CiliumEgressGatewayPolicy interface: tun0` requires `tun0` on the **node**, which means running Gluetun
`hostNetwork: true`. But Gluetun *owns its netns routing* — on connect its `[routing]` logic adds
`0.0.0.0/0` via the tunnel. In hostNetwork that would **redirect the NODE's own default route through the
VPN**, breaking k3s/etcd/kubelet on that node. Gluetun has no supported "create tun0 but don't touch the
default route" mode, so option 1 can't expose `tun0` on the node safely. (The external-router-offload
pattern some use — MikroTik/RouterOS policy routing per pod-subnet — is out of scope: ADR-0013 mandates an
*in-cluster* gateway on a dumb SOHO edge.)
**→ Escalated.** The remaining viable path is **option 2 (gateway-pod routing)**: `egress-class=vpn` pods
default-route to the Gluetun pod via an init-container (+NET_ADMIN), and Gluetun forwards+NATs out `tun0`.
More moving parts, and it requires Gluetun to forward for *foreign* pods (not its native same-netns model),
so it needs validation. Kill-switch (R-NET-06) then relies on Gluetun's `FIREWALL=on` dropping non-tunnel
egress. Fixtures for the `egress-class`/`vpn-killswitch` conformance tests are authored; the
`CiliumEgressGatewayPolicy` / steering datapath is intentionally NOT committed until the mechanism is chosen
+ validated, to avoid shipping an untested (and potentially node-breaking) datapath.
## Port forwarding (qBittorrent) — follow-up
PIA assigns a forwarded port that can change on reconnect. A small **port-sync helper**
(`qbittorrent-natmap`) reads the current port from the `gluetun-control` API (:8000
`/v1/openvpn/portforwarded`) and writes it into qBittorrent, keeping it hands-off (R-UPD). Authored
alongside the qBittorrent app (Phase 6 apps / or here when qBittorrent lands).