felhom.eu/documentation/audits/SPIKE-connectivity-wireguard-2026-07-03.md

# SPIKE — offsite-backup connectivity: WireGuard vs the home-NAT reality (2026-07-03)

> **STATUS: COMPLETE — verdict below.** This spike validates the offsite-backup *transport* before
> any production connectivity spec is written. No production code shipped; no Headscale on the real
> hub; no agent binary change; no persistent config on felhom-pve. All probes ran on a throwaway
> Hetzner box (`felhom-hetzner`, destroyed-or-parked after — see §6) and as additive, fully-removed
> interfaces/services on the demo-felhom host. The real DooPlex PBS and its customer snapshots were
> **never** a probe target.

**Class:** SPIKE (empirical validation; no product code). **Repos:** felhom.eu (this doc only);
felhom-agent read-only for grounding (`internal/pbs/{client,pin}.go`, `configs/build-golden.sh`,
`internal/hub/cloudflared.go`, `internal/escrow/identity.go`).
**Probe ends:** `felhom-hetzner` = Hetzner CX23, Debian 13.4, public IPv4 `167.233.158.164`,
global IPv6 `2a01:4f8:...::/64` (throwaway — NOT the live jarrs.eu box) ⟷ **demo-felhom** =
the real PVE 9.2.2 host on the operator's home line (One Hungary fixed cable, Budapest), driven
over the existing LAN SSH path; the tunnel itself always dialed **out**.

**Verdict (one line):** **GO at the lowest rung — H1: plain WireGuard, box-dials-out to a small
public endpoint, holds through the home NAT with zero coordinator/relay, carries the PBS API with
the TLS pin intact at ~wire-speed of the home uplink; Headscale is a separable fleet-management
option, not a traversal necessity — with the honest caveat that the operator's line turned out to
be plain-NAT (public IP), not CGNAT, so the CGNAT-traversal claim rests on mechanism + a deferred
second-ISP confirmation (Peti's VM 110), not on this line's evidence.**

---

## 1. Why this spike exists (the gap, grounded)

Offsite backup — the paid DR substrate (`architecture/01-topology-and-trust.md` §8) — is
structurally unreachable for a real customer today:

- Backup is outbound push to PBS (§7/§8), and the only PBS lives at a **LAN address**,
  `192.168.0.180:8007` (`felhom-agent/internal/pbs/client.go`).
- The public edge is Cloudflare-Tunnel → `nginx-internal` — HTTP(S) only, **no public inbound UDP
  anywhere**; DooPlex has no public IP.
- So offsite backup works only because the demo box shares a LAN with PBS. A real customer's box
  cannot reach `192.168.0.180` at all. The DR crypto/identity side is done and drilled
  (`SPIKE-dr-recipe-2026-06-16.md`); **transport is the only missing piece**.

**Architectural corollary (recorded as a finding):** production offsite backup requires a
**publicly-reachable endpoint that today's infrastructure does not provide**. The throwaway Hetzner
box in this spike is a preview of that production endpoint, not just a test rig (§4.4).

The hypothesis ladder tested, cheapest-first (the roadmap's "Headscale + DERP" sketch assumes the
hard problem — *peer-to-peer* traversal between two NAT'd hosts — but backup is **hub-and-spoke to
a public endpoint**, which any NAT'd box can reach outbound):

| Rung | Hypothesis | Result |
|---|---|---|
| H0 | both ends global IPv6 → direct, no NAT | **N/A — the home side has no IPv6 at all** (P0/P1) |
| H1 | plain WG over IPv4, box→public endpoint, keepalive | **CONFIRMED** on this line (P2/P3) |
| H2 | outbound UDP blocked → TCP/443 fallback needed | **not needed** — UDP 51820 *and* 443 both pass (P4) |
| H3 | fleet needs → Headscale | **separable** — fleet layer, not traversal (P5) |

---

## 2. Probe results (verbatim, redacted)

### P0 — NAT classification of the demo-felhom line — **plain single-NAT with a public IPv4, no IPv6**

- Egress IP seen by `ifconfig.me` from the host: `37.191.56.193`.
- Source IP observed **at the Hetzner box** for an inbound TCP connection from demo-felhom
  (HTTP listener log): `37.191.56.193` — same address, i.e. one NAT, stable mapping.
- `ipinfo.io`: `37.191.56.193` = `catv-37-191-56-193.catv.fixed.one.hu`, AS21334 One Hungary Ltd.
  — a *fixed* (subscriber-routable) cable address, not shared-pool space.
- Traceroute from the host (first hops):

  ```
   1  192.168.0.1      0.239 ms      ← home router
   2  89.133.164.70   12.485 ms      ← already PUBLIC ISP space (catv...one.hu, same AS)
   3  89.135.220.78   22.316 ms
  ```

  **No `100.64.0.0/10` (or private) hop after the home router** — no CGNAT tier on the path.
- IPv6: `ip -6 addr` shows link-local only; `curl -6` fails (`NO-IPV6-EGRESS`); a **15-minute
  passive `tcpdump` capture for router advertisements on `vmbr0` caught zero RAs**. (Caveat noted:
  `accept_ra=1` + `forwarding=1` means the kernel would ignore RAs anyway, which is why the
  wire-level capture, not the address absence, is the evidence.) No IPv6 exists on this LAN.

**Consequences:** the operator's line is a *softer* premise than the task assumed — H1 success here
does **not** prove CGNAT traversal (see §5). H0 (IPv6-direct) is dead on arrival for this line and
cannot be the plan's foundation: production must assume v4-only homes.

### P1 — H0 IPv6-direct — **N/A** (recorded gap)

Not testable: the home side has no global IPv6 (P0). The Hetzner side has native v6, so the
server-side of a dual-stack endpoint is free — worth keeping the endpoint dual-stack so v6-capable
customer lines use it, but it solves nothing this spike needed solving.

### P2 — H1: plain WG over IPv4, outbound-initiated — **CONFIRMED, first try, no coordinator/relay**

Setup: WG keypair per side (private keys generated on-box, 0600, never transited or logged);
Hetzner `wgspike` = `10.99.0.1/24`, `ListenPort 51820`; demo-felhom `wgspike` = `10.99.0.2/24`,
`Endpoint 167.233.158.164:51820`, `PersistentKeepalive = 25`, config under `/tmp` (nothing in
`/etc/wireguard`). PVE 9.2 needed nothing beyond `modprobe wireguard` (module ships in the kernel)
+ the `wireguard-tools` package.

```
demo-felhom# wg-quick up /tmp/wgspike.conf && ping -c 3 10.99.0.1
64 bytes from 10.99.0.1: icmp_seq=1 ttl=64 time=51.2 ms
3 packets transmitted, 3 received, 0% packet loss
rtt min/avg/max/mdev = 33.411/40.659/51.204/7.628 ms
```

Server-side view: peer endpoint `37.191.56.193:32873` (the NAT-translated source port), i.e. a
classic outbound UDP mapping — exactly the mechanism that also works through CGNAT.

**Keepalive hold (idle test):** the link carried **zero data traffic for 11.4 minutes**
(keepalives only — total transfer counters after the window: 1 888 B rx / 3 744 B tx). Result:
handshake stayed fresh throughout (age 120 s at check — keepalive-driven rekeys continuing on
schedule), the NAT mapping never moved (`37.191.56.193:32873` unchanged), and the first ping after
the idle window went straight through:

```
3 packets transmitted, 3 received, 0% packet loss
rtt min/avg/max/mdev = 30.119/30.812/31.279/0.499 ms
```

The tunnel **holds through idle with no help** — the single mechanism CGNAT UDP timeouts would
threaten, working as designed.

### P2b — PBS `:8007` through the tunnel, TLS pin intact — **CONFIRMED (positive + negative)**

A real PBS 4.2.2 was installed on the throwaway box (datastore `scratch` at `/srv/pbs-scratch`,
API token `root@pam!spike`, secret 36 chars — never logged; box firewalled to SSH + WG/iperf ports
only, so `:8007` is **not publicly reachable — tunnel-only by construction**). From demo-felhom,
with `PBS_FINGERPRINT` = the scratch PBS leaf-cert SHA-256 (the exact pin model of
`felhom-agent/internal/pbs/pin.go`):

```
=== public :8007 reachability (must FAIL, firewall proof):
000 / UNREACHABLE (as intended)
=== tunnel :8007 API version (pin intact):
client version: 4.2.0
server version: 4.2.2
=== NEGATIVE pin test (wrong fingerprint, must FAIL):
WARNING: certificate fingerprint does not match expected fingerprint!
```

The exact-cert pin validates **through** the tunnel (WG is below TLS; the pinned leaf is
unchanged), and a wrong pin is detected — the red-proof for the pin surviving the transport
change. (Deploy gotcha found + fixed on the scratch box: a datastore under `/root` fails with
`Permission denied` — the PBS `backup` user can't traverse `/root`; use a world-traversable
parent like `/srv`.)

### P3 — throughput — home uplink is the bottleneck; tunnel overhead ≈ 5% on the backup direction

Direct (no tunnel) `iperf3` demo-felhom ⟷ Hetzner (two runs each):

```
UPLOAD  (home→Hetzner):  42.1 / 42.0 Mbit/s sender    (receiver 40.7 both runs)
DOWNLOAD (Hetzner→home):  203 / 230 Mbit/s            (retransmits: 3 / 285)
```

Through the WG tunnel:

```
UPLOAD  (tunnel):   39.1 Mbit/s sender / 38.4 receiver   ← ~5% under direct
DOWNLOAD (tunnel):  565 Mbit/s, 0 retransmits            ← FASTER than direct TCP
```

The download anomaly is real and repeatable-in-session: plain TCP from Hetzner→home tops out at
~200–230 Mbit/s with hundreds of retransmits, while the same TCP stream **inside** the WG/UDP
encapsulation ran at 565 Mbit/s with zero retransmits — whatever middlebox/shaping hurts inbound
TCP on this cable line does not see the UDP tunnel. Download is the restore direction; a pleasant
surprise, not a design input. (Single-session observation; not load-tested further.)

**PBS-native measurements over the tunnel** (the actual workload):

```
proxmox-backup-client benchmark:  TLS (maximal backup upload speed): 4.94 MB/s
                                  AES256-GCM 3 904 MB/s, SHA256 2 152 MB/s (CPU is never the limit)

REAL backup, 2 GiB of /dev/urandom (worst case — incompressible, non-dedupable):
  spike.pxar: had to backup 2 GiB of 2 GiB in 480.75 s (average 4.26 MiB/s)
  Duration: 481.26s — sustained, zero stalls, saturating the uplink for 8 minutes straight
```

4.94 MB/s TLS-in-tunnel ≈ 41 Mbit/s = the full measured uplink: **the tunnel + TLS stack costs
effectively nothing against the line's own ceiling.**

**First-backup window math (at the measured ~40 Mbit/s ≈ 5 MB/s effective uplink):** 10 GB ≈ 35 min,
50 GB ≈ 2.8 h, 100 GB ≈ 5.6 h — a realistic first backup fits an overnight window; incrementals
(PBS dedup) are far smaller. The *download* direction (restore) is ~5× faster on this line.

### P4 — H2: outbound UDP reality — **51820 AND 443/UDP both pass**

UDP 51820 egress is proven by P2 itself. The same tunnel re-pointed at **`:443/UDP`**
(`wg set … listen-port 443` server-side, endpoint updated client-side) handshook immediately and
carried pings at the same RTT:

```
3 packets transmitted, 3 received, 0% packet loss
rtt min/avg/max/mdev = 30.300/32.612/35.044/1.938 ms
endpoint: 167.233.158.164:443
```

No TCP/443 fallback (DERP-over-HTTPS etc.) is needed on this ISP; and 443/UDP working means even a
hypothetical 51820-blocking ISP has a first, zero-cost escalation (same protocol, friendlier port)
before any TCP fallback machinery is justified.

### P5 — H3: what Headscale actually buys — **a fleet layer, not traversal**

Headscale 0.29.2 stood up on the throwaway box (systemd service + `/etc/headscale/config.yaml` +
sqlite at `/var/lib/headscale/db.sqlite`): created users `customer-a`/`customer-b`, minted a
preauth key, inspected the config surface. Observations:

- **Operational surface:** a public control-plane service (TLS **required** — the embedded DERP
  server refuses a non-HTTPS `server_url`), a database to back up, user + preauthkey + node
  lifecycle, an ACL policy file, plus a **client-side daemon (tailscaled) on every box**. The
  default DERP config points at **Tailscale's public derpmap**
  (`controlplane.tailscale.com/derpmap/default`) — a third-party dependency unless you run your
  own DERP.
- **For hub-and-spoke backup it is redundant:** traversal is already solved by rung H1 (the spoke
  dials a public endpoint). Headscale's real value is fleet mechanics — key rotation, node expiry,
  per-customer users/ACLs, automatic endpoint re-resolution, and DERP *fallback* for UDP-hostile
  lines (H2 — which this ISP does not need).
- **Recorded gap:** no client was joined to the throwaway headscale (joining = installing
  tailscaled on the production PVE host; declined). Surface inspection + docs only.

### P6 — per-customer isolation

- **Plain WG (recommended rung):** isolation is inherent in hub-and-spoke: the server holds one
  peer entry per box with `AllowedIPs = <that box's /32>` — cross-peer traffic requires the
  *server* to forward between WG peers, which the endpoint's firewall simply does not allow
  (forwarding stays off; only `INPUT` to the PBS port is open on the WG interface). A box can
  reach the PBS API and nothing else; boxes cannot see each other. PBS-side tenancy = the existing
  per-customer namespace + token model (unchanged).
- **Headscale (if ever adopted):** the same property must be *written* as ACL policy
  (per-customer users + a policy file allowing only spoke→PBS) — more expressive, but a policy to
  maintain and test rather than a topology that cannot express the attack.

### P7 — guest-side WG refutation — **hypothesis REFUTED: kernel WG works in the unprivileged guest; host placement stands on architecture, not infeasibility**

Against the real guest 9201 (`features: nesting=1,keyctl=1`, `unprivileged: 1`):

```
# pct exec 9201 -- ls -la /dev/net/
ls: cannot access '/dev/net/': No such file or directory     ← no TUN, as expected
# pct exec 9201 -- ip link add wgtest type wireguard
rc=0                                                          ← SUCCEEDS
wgtest           DOWN           <POINTOPOINT,NOARP>
# pct exec 9201 -- ip link del wgtest                         ← removed immediately
```

Kernel WireGuard is **not** a TUN consumer — an unprivileged LXC with the stock golden features can
create a WG interface as long as the host kernel has the module. So the "no TUN → guest can't"
argument is dead. The **host-placement recommendation stands on architecture instead** (§4.2):
backup traffic *originates on the host* (the agent drives vzdump/PBS push with host-side
credentials), the tunnel must survive guest death/rebuild by construction (the `cloudflared`
precedent, 01 §7), and the guest must keep holding **zero** infra credentials (01 §3). A guest-side
WG key would violate the trust model, not the kernel. (`/dev/net/tun` absence *does* still block a
guest-side **tailscaled**, relevant only if the Headscale rung were ever chosen.)

---

## 3. The twelve §7 answers

1. **CGNAT or plain-NAT?** Plain single-NAT with a fixed public IPv4 (`37.191.56.193`, One Hungary
   catv-fixed); **no IPv6 at all** (no address, no egress, no RAs on the wire). The CGNAT premise
   did not hold *on this line*.
2. **H0 (v6-direct)?** N/A — no home IPv6. Keep the production endpoint dual-stack, but v4 is the
   planning baseline.
3. **H1 (plain WG through the NAT)?** Yes — outbound-initiated, `PersistentKeepalive=25`, zero
   coordinator/relay, held through an 11.4-minute fully-idle window with a stable NAT mapping,
   and carried the PBS API with the exact-cert pin validating through the tunnel (positive +
   negative pin proof, P2b).
4. **Throughput?** Tunnel costs ~5% on the ~40 Mbit/s uplink bottleneck; PBS-native TLS speed
   through the tunnel = the full uplink (4.94 MB/s); a real worst-case 2 GiB backup sustained
   4.26 MiB/s for 8 min with zero stalls. First backup of 50–100 GB ≈ 3–6 h (overnight window);
   incrementals trivial.
5. **H2 (UDP blocked)?** No — UDP 51820 **and** 443 both pass on this ISP. TCP/443 fallback is a
   *contingency for other ISPs*, not a launch requirement; note it as the escalation if a customer
   line ever proves UDP-hostile.
6. **H3 (Headscale needed)?** Not for backup. It is a separable fleet-management layer (key
   lifecycle, ACLs, endpoint re-resolution, DERP fallback) purchasable later without re-architecting
   — the spoke config is the same WG primitive either way. Its cost: a public TLS control plane,
   a DB, tailscaled on every box, and (by default) a Tailscale-operated DERP dependency.
7. **Isolation?** Plain WG hub-and-spoke: per-peer `/32` `AllowedIPs` + no forwarding on the
   endpoint = no lateral path by topology (P6).
8. **Placement?** Host — same placement as `cloudflared` (agent-managed systemd service). The P7
   probe *refuted* the infeasibility argument (kernel WG works in the guest), so the decision rests
   on: backups originate host-side, tunnel survives guest death, guest keeps zero infra creds.
9. **Key custody + DR:** mint the box's WG keypair at install into the agent state dir (0600,
   `felhom-agent` user), register the pubkey with the hub, and add the private key to the
   escrowed `IdentityBundle` (`felhom-agent/internal/escrow/identity.go`) so a re-provisioned box
   re-establishes its tunnel from the same recovery flow that already restores PBS identity. The
   endpoint's own pubkey+address are non-secret config in the hub record.
10. **Where does the production endpoint live?** A small public relay VM (this spike's CX23-class
    box ≈ €4–8/mo) — see §4.4 for the two variants (PBS-on-VM vs VM-as-rendezvous-relay with
    DooPlex PBS dialing out too). DooPlex itself cannot be the endpoint (no public IP/UDP).
11. **Trust model preserved?** Yes — the box only ever dials **out** (UDP to the endpoint); no new
    inbound listener, no new inbound attack surface on the box; the endpoint exposes exactly one
    UDP port publicly. Operator-inbound access (B) and game-server ingress (C — public players are
    not tunnel members) are **explicitly separate problems, out of scope here**, and nothing in
    this design presumes them.
12. **Go/no-go:** **GO** on rung **H1** — plain WireGuard, host-side, box-dials-out, to a small
    public endpoint carrying PBS. Headscale/DERP only if (a) a real customer ISP proves
    UDP-hostile (H2) or (b) fleet size makes static peer config painful (H3). No pivot to an
    outbound-mTLS PBS proxy needed — the simplest thing worked.

---

## 4. Architecture recommendation (input to the production spec — NOT the spec)

### 4.1 Transport rung
Plain kernel WireGuard, one tunnel per box, box-initiated, `PersistentKeepalive=25`, endpoint =
a public relay VM. No coordinator, no relay, no mesh.

### 4.2 Placement
On the Proxmox **host**, as an **agent-managed systemd service** — the exact `cloudflared`
precedent (01 §7, `internal/hub/cloudflared.go` prober pattern). Interface + config under the
agent's ownership; guest untouched.

### 4.3 Identity & custody
Per-box keypair minted at install; private key 0600 in the agent state dir; pubkey → hub record;
private key → escrowed `IdentityBundle`. Server pubkey/endpoint = plain config from the hub.
Revocation = delete the peer entry at the endpoint (hub-driven, auditable).

### 4.4 The production public endpoint (the §1 corollary)
Two viable shapes, decision deferred to the production spec:
- **(a) PBS on the public VM** — simplest ops; storage cost scales on the VM; client-side
  encryption means at-rest exposure is ciphertext-only anyway.
- **(b) VM as dumb rendezvous relay** — DooPlex PBS *also* dials out a WG spoke to the VM; the VM
  forwards spoke↔PBS traffic and stores nothing. Keeps bulk storage on DooPlex's existing disks;
  the VM needs only bandwidth; **both** sides remain outbound-only (this also solves DooPlex's own
  no-public-UDP problem). Ciphertext-only transits the VM.
Either way the endpoint is small, dual-stack, firewalled to WG-UDP-only, and *is* the thing the
subscription's offsite promise runs through — it becomes managed infrastructure, not a lab box.

### 4.5 Isolation
Per-peer `/32` AllowedIPs; endpoint forwarding off; WG-interface firewall admits only the PBS port.
Per-customer PBS namespaces/tokens unchanged on top.

---

## 5. Scope caveats + NOT-validated list

- **The CGNAT claim is NOT proven by this spike** — the operator's line turned out to be
  plain-NAT. What *is* proven: outbound-initiated WG through a NAT with keepalive, which is the
  same mechanism CGNAT permits (CGNAT breaks *inbound* and *p2p*, not outbound-to-public). The
  honest residual risk is an ISP that breaks long-lived UDP mappings aggressively — exactly what
  the deferred second-ISP vantage (**Peti's Debian VM 110**, currently unreachable — "No route to
  host") should confirm once fixed. **Follow-up recorded.**
- Single ISP (One Hungary catv), single evening, single geography. No long-horizon (days) hold
  data; no home-IP-rotation event observed (WG roaming handles it by design — untested here).
- No Headscale client join (tailscaled kept off the production host); Headscale assessed on
  server-side surface + docs only.
- No real vzdump→PBS run over the tunnel (synthetic `proxmox-backup-client` traffic only — real
  guest snapshots stay on the real PBS).
- IPv6 absence is a One-Hungary-line fact, not a market fact; the endpoint should stay dual-stack.
- The tunnel-download anomaly (UDP encapsulation outrunning direct TCP) is a single-session
  observation — noted, not relied on.
- Restore-direction (PBS → box) bulk transfer was measured only as iperf3, not as a
  `proxmox-backup-client restore` run.
- Relayed-mode (DERP) throughput was NOT measured — moot, since no relay is recommended.

## 6. Cleanup assertion

**demo-felhom host (all verified after removal):** `wgspike` interface deleted (`wg-quick down`,
"Device does not exist" confirmed); WG private key + PBS token secret **shredded**; `/tmp` spike
files (config, RA capture log, 2 GiB dataset) removed; the `iperf3` package I installed purged;
`wireguard` kernel module unloaded (`lsmod` count 0); `/etc/wireguard/` empty and untouched
(mtime Apr 2025); guest 9201's probe link `wgtest` deleted seconds after creation ("Device does
not exist" confirmed); **no persistent config, unit, or package change remains from this spike**
(`wireguard-tools` was already installed before the spike and was left as found).

**Throwaway Hetzner box (`felhom-hetzner`, 167.233.158.164):** WG interface down + config/private
key/PBS token **shredded**; headscale + PBS services stopped and disabled; headscale DB
(`/var/lib/headscale`) and the scratch datastore (`/srv/pbs-scratch`, random-data chunks only)
deleted; then the box was **powered off** (`shutdown -h`). Disposition: **kept powered-off** for
the deferred second-ISP follow-up (Peti's VM 110) — the operator can start it from the Hetzner
console for that probe or destroy it outright; nothing secret survives on it either way. The live
jarrs.eu Hetzner server was never touched.

**No repo/hub/agent/manifest production change** — this commit is docs-only.

---

## Backlog resolution

`runbooks/day0-install.md` "PBS host is LAN-only until the Headscale/WireGuard work lands" → now
points at this spike: the transport decision is **made** (plain WG to a public endpoint); what
remains is the production connectivity spec + implementation (endpoint VM, agent-managed WG
service, escrow join, hub peer registry).