VPN Field Guides [A]

VPN field guide 5/5: diagnose a broken tunnel and retire a lost device

When a tunnel breaks, resist the urge to change the port, DNS and firewall all at once. Start with the last known good test from part 4 and change one variable. These symptoms point to different parts of the path.

No recent handshake: check whether the client is reaching the right public address and UDP port, whether the cloud firewall and host firewall both permit it, whether the server process is listening, and whether the public keys match the right peer. A home server behind CGNAT may have no usable inbound path at all; a port forward on your router cannot fix carrier NAT. A recent handshake with no traffic: check each side's tunnel address, AllowedIPs, routes and host firewall. Remember that the server must permit exactly the client's assigned tunnel address, such as 10.44.0.2/32 in the lab, not a broad range copied from a stranger's guide.

Tunnel address works, public IPv4 does not: check server forwarding, source NAT or return routing, and firewall forward rules. Public IPv4 works, hostnames do not: check DNS configuration on the client, server or active profile. Small pages work but large downloads stall: investigate path MTU and fragmentation instead of repeatedly rotating keys. If the connection dies only after an idle period behind NAT, a client-side PersistentKeepalive interval may help; WireGuard documents 25 seconds as a useful value in that case. If it fails only on one network, compare UDP reachability and network policy there. WireGuard does not gain native TCP transport simply by changing its port number.

For a lost phone, remove its peer public key from the server and make the change active; keep other devices' keys. This prevents new handshakes from that key, but it cannot erase data already seen or a client config copied elsewhere. Check server state and the device inventory to ensure you removed the right peer. If you suspect the server's private key was exposed, rotate the server key and distribute the new public key to every remaining client. If a client private key was exposed, replace that client's key pair and revoke the old public key. A preshared key, if you used one, is another secret to rotate. Protect backups containing configs and QR images too.

For OpenVPN deployments, key removal is a different workflow: revoke the affected certificate, publish the current certificate revocation list to the server and verify that a revoked client can no longer authenticate. Do not use “I deleted the file from my laptop” as a revocation test. For IKEv2, follow the authentication method actually deployed; a shared credential, device certificate and EAP account have different recovery steps.

After any repair or rotation, rerun the route, DNS, IPv6 and endpoint-failure tests. Record the result and the date. A fresh handshake is the start of verification, not the finish line. If your logs or screenshots are going into a public help thread, redact private keys, preshared keys, public endpoints tied to your identity and account details first.

Sources, checked 3 October 2026:
Ubuntu WireGuard troubleshooting — https://ubuntu.com/server/docs/how-to/wireguard-vpn/troubleshooting/
Ubuntu security tips — https://ubuntu.com/server/docs/how-to/wireguard-vpn/security-tips/
WireGuard Quick Start — https://www.wireguard.com/quickstart/
WireGuard known limitations — https://www.wireguard.com/known-limitations/
OpenVPN certificate revocation — https://openvpn.net/community-docs/revoking-certificates.html
strongSwan NAT traversal — https://docs.strongswan.org/docs/latest/features/natTraversal.html

↑ 10

Обсуждение

Комментарии

0

Комментариев пока нет.

Написать Join — чтобы участвовать в обсуждении.