A turnkey docker compose setup that puts an SNI-aware TCP router
(HAProxy) in front of mtg and a real web server (Caddy with
automatic HTTPS).
Modern DPI systems actively probe suspected proxies. If the server closes the connection or returns something unexpected, the IP gets flagged. With this setup:
Because your domain’s DNS points to this server, the SNI/IP match is natural and passive DPI has nothing to flag.
# 1. Point your domain's DNS A/AAAA record to this server's IP.
# 2. Generate an mtg secret:
docker run --rm nineseconds/mtg:2 generate-secret --hex YOUR_DOMAIN
# 3. Edit the config files:
# - mtg-config.toml → paste the secret
# - haproxy.cfg → replace "example.com" in the SNI ACL
# - .env or export → DOMAIN=your.domain
# 4. (Optional) put your site content into www/
# 5. Start:
docker compose up -d
# 6. Verify:
# - Open https://YOUR_DOMAIN in a browser → you should see the web page
# - Configure Telegram with the proxy link from:
docker compose exec mtg mtg access /config/config.toml
HAProxy forwards TCP connections to mtg and Caddy with a PROXY protocol v2 header so both backends see the real client IP instead of HAProxy’s container address. The three pieces must stay in sync:
haproxy.cfg — send-proxy-v2 on the mtg and web backend server linesmtg-config.toml — proxy-protocol-listener = trueCaddyfile — listener_wrappers { proxy_protocol { ... } tls } on :8443If you disable one, disable all three, otherwise the backend will fail to parse the connection.
[domain-fronting] is set explicitly)When mtg sees TLS traffic that isn’t valid Telegram (a probe or a
browser hitting your domain on :443), it forwards that connection to
a real web server — “domain fronting”. By default mtg uses the
secret’s hostname as the fronting target and resolves it via DNS.
In this setup that hostname resolves back to this server, so mtg’s
fronting dial would hit HAProxy on :443, HAProxy would see the SNI
matching the secret and route the connection back to mtg → loop.
The trigger is DNS, not name equality: the loop reproduces whenever the secret’s hostname resolves to this host, regardless of how it relates to the domain Caddy serves (same name, subdomain, parent, or unrelated). In an SNI-router deployment the secret’s hostname has to point here for clients to reach mtg in the first place, so the loop is the default state unless mtg is steered away from HAProxy.
To break the loop, mtg-config.toml pins the fronting target to
Caddy’s container address directly:
[domain-fronting]
ip = "172.28.0.10"
port = 8443
proxy-protocol = true
The IP matches services.web.networks.sni.ipv4_address in
docker-compose.yml (mtg’s domain-fronting.ip only accepts a literal
IP, not a hostname, hence the static sni network). proxy-protocol =
true matches Caddy’s :8443 listener wrapper so the real client IP
still propagates to Caddy’s logs.
If you change Caddy’s pinned IP, update both files together. If
172.28.0.0/24 collides with another network on this host, change the
subnet in docker-compose.yml and the IP in mtg-config.toml to match.
Caveat for IPv6 clients: when an IPv6 probe is fronted, mtg’s outbound PROXY v2 header has an IPv6 source but an IPv4 destination (Caddy’s pinned address). Caddy may refuse the mixed-family header and log the docker-network address instead of the real client IP for that connection. Telegram traffic is unaffected.
HAProxy passes /.well-known/acme-challenge/ requests on :80 to
Caddy so that HTTP-01 validation works out of the box. Make sure your
domain’s DNS A/AAAA record points to this server before starting.
┌──────────────────┐
:443 ──────>│ HAProxy │
│ (TCP, SNI peek) │
└──┬───────────┬───┘
SNI match │ │ default
v v
┌─────────┐ ┌─────────┐
│ mtg │ │ Caddy │
│ :3128 │ │ :8443 │
│ FakeTLS │ │ real TLS│
└─────────┘ └─────────┘
| File | Purpose |
|---|---|
docker-compose.yml |
Service definitions |
haproxy.cfg |
SNI routing rules — edit the domain |
mtg-config.toml |
mtg proxy config — paste your secret |
Caddyfile |
Web server config (auto-HTTPS) |
www/ |
Static site content served by Caddy |