docs: add reverse-proxy topology + external-TLS conventions

Capture the cert + edge-proxy conventions worked through deploying the
helexa-bench UI:

- external-tls.md — publicly-trusted certs via Let's Encrypt (certbot,
  Cloudflare DNS-01, ECDSA, /root/.certbot-internal); the external
  counterpart to internal-tls.md. Decision rule: public name → LE,
  *.internal → internal CA.
- reverse-proxies.md — names the per-site edge proxies (oolon for
  kosherinata, hanzalova.internal for the office) and what sits behind
  each, the public-vs-mesh access paths + the "public names don't
  hairpin from inside the mesh" gotcha, per-vhost cert choice, nginx
  conventions, and the bench (bench.helexa.ai + bench.internal) worked
  example.
- readme + generic.md §11 cross-reference both.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
This commit is contained in:
2026-06-14 15:50:57 +03:00
parent 200c41b4f1
commit 746c55fe94
4 changed files with 203 additions and 1 deletions

113
external-tls.md Normal file
View File

@@ -0,0 +1,113 @@
# External TLS: public certs for WAN-facing vhosts
Extends `generic.md` §11 (TLS / PKI). That section and `internal-tls.md` cover the
**internal** PKI (Smallstep `step-ca`, `*.internal` names, mesh-only). This doc covers
the other half: **publicly-trusted certs for names served to the public internet** at a
site's WAN edge — e.g. `bench.helexa.ai`, `qapish.ai`, `*.zap.pics`.
Decision rule (the whole strategy in one line):
> **Public, internet-resolvable name → Let's Encrypt. Mesh-only `*.internal` name →
> internal CA (`internal-tls.md`).** A service reached both ways gets one vhost of each
> (see `reverse-proxies.md`).
Public certs must chain to a publicly-trusted root (browsers off the mesh don't trust
the `lair` internal root), so these come from Let's Encrypt — never `step-ca`.
---
## 1. Issuance: certbot + Cloudflare DNS-01, ECDSA
Our public DNS zones are on Cloudflare, so we use the **DNS-01** challenge via the
`certbot-dns-cloudflare` plugin. DNS-01 is deliberate:
- **No inbound :80 needed.** The challenge is a TXT record, not an HTTP hit — so a cert
can be issued (or renewed) even while nginx is stopped or the host isn't yet reachable
from the WAN. (This is why a dormant edge proxy doesn't block issuance.)
- **Wildcard-capable**, if a zone ever wants `*.example.com`.
Keys are **ECDSA** (`--key-type ecdsa`), matching the rest of the fleet.
```sh
sudo certbot certonly \
-m ops@zap.pics --agree-tos --no-eff-email --noninteractive \
--cert-name <domain> \
--key-type ecdsa \
--dns-cloudflare \
--dns-cloudflare-credentials /root/.certbot-internal \
--dns-cloudflare-propagation-seconds 60 \
--keep-until-expiring \
-d <domain>
```
- **`/root/.certbot-internal`** holds the Cloudflare API token. One token covers all the
zones we manage (`helexa.ai`, `zap.pics`, …), so new sub-domains under an existing zone
need no new credential — just run the command.
- **`--keep-until-expiring`** makes scripted/repeated runs idempotent (no-op if the cert
is still valid), so this is safe to call unconditionally from `infra-setup.sh`.
- `--cert-name <domain>` pins the lineage name so the cert lands at a predictable path
regardless of `-d` ordering.
## 2. Paths
certbot's standard layout (do **not** relocate — the renew timer expects it):
| Path | Contents |
| --- | --- |
| `/etc/letsencrypt/live/<domain>/fullchain.pem` | cert + intermediate chain |
| `/etc/letsencrypt/live/<domain>/privkey.pem` | private key |
These live under root-only `/etc/letsencrypt/live` (`0700`). Scripts that check for an
existing cert must `sudo test -d /etc/letsencrypt/live/<domain>` — an unprivileged
`test` silently returns false and will wrongly conclude the cert is missing.
## 3. Renewal
Automatic via certbot's own `certbot-renew.timer` (systemd) — **no per-cert unit**,
unlike the internal `step@<name>` template. certbot renews any lineage within 30 days of
expiry and runs the configured deploy hook. Ensure nginx reloads after renewal with a
deploy hook (once per host):
```sh
# /etc/letsencrypt/renewal-hooks/deploy/reload-nginx.sh (chmod +x)
#!/bin/sh
systemctl reload nginx 2>/dev/null || true
```
## 4. nginx wiring
```nginx
server {
listen 443 ssl;
http2 on;
server_name <domain>;
ssl_certificate /etc/letsencrypt/live/<domain>/fullchain.pem;
ssl_certificate_key /etc/letsencrypt/live/<domain>/privkey.pem;
ssl_protocols TLSv1.2 TLSv1.3;
}
```
Keep an `:80` server for the same name only if you want an HTTP→HTTPS redirect; the
cert itself needs no `:80` (DNS-01). Never reference a cert path before the cert exists —
`nginx -t` fails on a missing `ssl_certificate` file and blocks **all** of nginx from
(re)starting. Issue first, then install the TLS vhost (gate the vhost install on
`sudo test -d /etc/letsencrypt/live/<domain>`).
## 5. Gotchas
- **SAN, not CN.** Modern clients ignore CN; the served name must be in the SAN. certbot
sets SAN from `-d`, so this is automatic — but if `curl` reports *"no alternative
certificate subject name matches target hostname"*, the listener answering isn't the
one holding this cert (see next point).
- **Wrong cert on the public endpoint = a routing problem, not a cert problem.** If a
public name returns something like `CN=opnsense.<site>.internal`, the WAN `:443`
forward (or HAProxy SNI route) on OPNsense isn't landing on the site's nginx. Fix the
edge route (`reverse-proxies.md` §2), not the cert.
## 6. Checklist for a new public vhost
1. Add the public DNS record on Cloudflare (unproxied by default — `generic.md` §11).
2. Issue the cert (§1), from `infra-setup.sh`, idempotently.
3. Point the nginx vhost at the `live/<domain>` paths (§4); `nginx -t` && reload.
4. Confirm the site's OPNsense forwards WAN `:443` to this nginx (`reverse-proxies.md`).

View File

@@ -536,7 +536,7 @@ templated `step@<name>` unit. That pattern is documented separately in
**`internal-tls.md`**.
### Ingress
- Per-site nginx reverse proxy terminates all WAN inbound 443.
- Per-site nginx reverse proxy terminates all WAN inbound 443 (`oolon` for kosherinata, `hanzalova.internal` for the office). The named topology, the public-vs-mesh access paths (and the hairpin gotcha), and the per-vhost cert choice are in **`reverse-proxies.md`**; external (Let's Encrypt) cert provisioning in **`external-tls.md`**.
- Public DNS via Cloudflare, **unproxied by default** (CF's mTLS origin-pull has been unreliable). Revisit if/when that changes.
- nginx serves static frontends directly from `/var/www/<app>` and reverse-proxies API traffic to the internal host:port from `manifest.yml`.

View File

@@ -13,6 +13,8 @@ The goal is boring consistency: the same crate layout, the same deploy flow, the
- **`generic.md`** — the baseline. Applies to every project unless that project explicitly overrides a section. Covers workspace layout, separation of concerns, configuration, secrets, deployment, service accounts, firewalld, SELinux, and code quality.
- **`deployment-gitea-actions.md`** — CI-driven deployment via a Gitea Actions workflow, as an alternative to the `deploy.sh` + `manifest.yml` flow in `generic.md` §7. The workflow is the source of infra truth; the runner deploys as a scoped `gitea_ci` user.
- **`internal-tls.md`** — provisioning and renewing per-service internal TLS certs (`<service>.internal`) for mesh-only nginx vhosts, extending the PKI conventions in `generic.md` §11.
- **`external-tls.md`** — publicly-trusted certs for WAN-facing vhosts via Let's Encrypt (certbot, Cloudflare DNS-01, ECDSA). The external counterpart to `internal-tls.md`.
- **`reverse-proxies.md`** — the per-site nginx edge proxies (`oolon` for kosherinata, `hanzalova.internal` for the office), what sits behind each, the public-vs-mesh access paths, and the per-vhost cert choice. Names the topology behind `generic.md` §11 Ingress.
More files will appear here over time as guidance that's more specific than `generic.md` gets extracted — per-stack, per-deployment-target, or per-problem-domain documents. When a project needs guidance that isn't generic, it belongs in a new file here, not buried in one project's repo.

87
reverse-proxies.md Normal file
View File

@@ -0,0 +1,87 @@
# Reverse proxies and edge ingress
Extends `generic.md` §11 (Network / Ingress). That section says "per-site nginx reverse
proxy terminates all WAN inbound 443"; this doc names the proxies, maps what sits behind
each, and pins down the two access paths and the per-vhost cert choice — plus the one
gotcha that bites every time (a public name doesn't work from *inside* the mesh).
---
## 1. The proxies (one per site)
Each WireGuard site has a single nginx edge proxy. All WAN-inbound 443 for that site is
port-forwarded by the site's OPNsense router to its proxy, which terminates TLS and
fans out to internal upstreams.
| Site | Edge proxy (nginx host) | Notable hosts behind it |
| --- | --- | --- |
| **kosherinata** (DC) | `oolon.kosherinata.internal` | `magrathea` (Postgres primary), `nikola`, `gramathea`, … |
| **hanzalova** (office) | `hanzalova.internal` | GPU/inference: `beast`, `benjy`, `quadbrat`; `bob` (helexa-bench API + Agent Zero); `frankie` (Postgres streaming standby); `trillian`; the workstation |
Site octet encodes the mesh subnet (`10.<site>.0.0/16`); see `generic.md` §11. New
office services front on `hanzalova.internal`; new DC services on `oolon`.
## 2. Two access paths — and the mesh hairpin gotcha
A service can be reached two ways, and they are **not** interchangeable:
- **Public (from the WAN):** public DNS (Cloudflare, unproxied by default) → site WAN IP
→ OPNsense forwards `:443` → site nginx → upstream. Cert: **Let's Encrypt**
(`external-tls.md`).
- **Internal (from the mesh):** split-horizon `.internal` DNS → the host/proxy directly
over WireGuard → nginx. Cert: **internal CA** (`internal-tls.md`).
> **Gotcha — public names don't hairpin.** From *inside* the mesh, a public name still
> resolves (via public DNS) to the site's **WAN** IP, so the packet hits the OPNsense
> **LAN** interface — which only forwards `:443` inbound from the **WAN**, not from the
> LAN. The connection dead-ends (or worse, gets OPNsense's own default cert). So a
> service that mesh clients also need must be published under a **`*.internal` name with
> its own internal-CA vhost**, in addition to its public vhost.
This is why dual-audience services get **two vhosts** on the same proxy — one public
(LE), one internal (`lair` CA) — usually sharing one webroot and one upstream.
## 3. Per-vhost cert choice
| vhost audience | name | cert | doc |
| --- | --- | --- | --- |
| Public / WAN | `<svc>.<public-zone>` (e.g. `bench.helexa.ai`) | Let's Encrypt (certbot, Cloudflare DNS-01, ECDSA) | `external-tls.md` |
| Mesh-only | `<svc>.internal` | internal CA (`step ca`, `lair` provisioner, `step@` renewal) | `internal-tls.md` |
Provisioner credentials for the internal CA (`~/.step/secrets/provisioner`, shipped to
the host transiently and removed) are covered in `internal-tls.md` §4.
## 4. nginx conventions on the proxies
- **`sites-available/` + `sites-enabled/` symlink**, included via
`/etc/nginx/conf.d/sites-enabled.conf` (`include /etc/nginx/sites-enabled/*.conf;`).
One file per `server_name`; enable with a relative symlink
(`ln -sf ../sites-available/<name>.conf /etc/nginx/sites-enabled/`).
- **Static SPA** served from `/var/www/<name>` with SPA fallback
(`try_files $uri $uri/ /index.html;`); **API** reverse-proxied to the internal
`host:port`. Internal vhosts add `ssl_trusted_certificate <internal root>` and pin
`ssl_protocols TLSv1.3`.
- **SELinux (enforcing):** webroots must be labelled `httpd_sys_content_t` or nginx
returns **403**. After creating/populating `/var/www/<name>`, run
`restorecon -R /var/www/<name>`; rsynced files inherit the dir's type.
- **Never reference a cert path before the cert exists** — `nginx -t` fails on a missing
`ssl_certificate` and blocks the whole server from (re)starting. Issue the cert, then
install the TLS vhost (gate on the cert's presence; serve an http-only bootstrap until
then if needed).
- Config + cert/renewal wiring is installed idempotently from each project's
`infra-setup.sh` (`deployment-gitea-actions.md` §2); the recurring artifact rsync
(e.g. built SPA `dist/`) rides in the deploy workflow.
## 5. Worked example: helexa-bench UI
The bench visualisation is reached both ways, fronted by `hanzalova.internal`:
| vhost | cert | DNS |
| --- | --- | --- |
| `bench.helexa.ai` (public) | Let's Encrypt | Cloudflare A → office WAN IP; OPNsense forwards WAN `:443``hanzalova` |
| `bench.internal` (mesh) | internal `lair` CA, renewed by `step@bench.timer` | split-horizon `bench.internal → hanzalova` mesh IP |
Both vhosts share one webroot (`/var/www/bench.helexa.ai`, the built SPA) and proxy
`/api` to the helexa-bench read API on `bob.hanzalova.internal:13132`. The internal vhost
exists precisely because of §2: from a workstation on the mesh, `bench.helexa.ai`
hairpins to the OPNsense LAN interface and fails, so mesh users hit `bench.internal`.