Running WireGuard?

WireGuard tunnel silent or dropping packets? We bring it back.

WireGuard is wonderful until the handshake stops completing or the tunnel comes up but nothing flows. Small, boring misconfigurations cause both: an AllowedIPs range that overlaps an existing route, an MTU that was fine on Ethernet and drops every reply on a cellular link, or a persistent keepalive that was never set on the roaming side. We work through the kernel module output and the routing table until the issue is visible and then close it.

  • Since 2015Fixing technology
  • FreeEstimates
  • US teamReal engineers

wg-quick.log

still broken

What it is

WireGuard at a glance

WireGuard is a modern VPN protocol written by Jason Donenfeld, upstreamed into the Linux kernel in version 5.6 in March 2020. It uses a small set of curated cryptographic primitives, Noise IK for the handshake, ChaCha20Poly1305 for the data channel, and Curve25519 keys, with no cipher negotiation and no X.509 anywhere.

The configuration model is deliberately plain: each peer has a private key, a public key, a set of AllowedIPs and an optional endpoint. There are userspace implementations for every mainstream operating system, an Android and iOS client, and a growing set of commercial mesh products like Tailscale and Netbird that build their control plane on top of the protocol.

Type
Modern layer-3 VPN protocol
First released
2016 prototype, 2020 mainline Linux
License
GPLv2 kernel module, dual-licensed userspace
Maintained by
Jason Donenfeld and the WireGuard contributors
Typically used for
Site to site tunnels, remote access, mesh overlays, roaming client VPN

The honest version

WireGuard pros and cons

No technology is all upside. This is where WireGuard earns its place and where it tends to cause trouble.

Where WireGuard is strong

  • Short, auditable codeThe kernel module is a few thousand lines compared to the hundreds of thousands in IPsec and OpenVPN stacks, which is one of the reasons it was accepted upstream so quickly.
  • Handshake and throughput are both fastSetup is one round trip and the data path uses ChaCha20Poly1305, which outperforms AES-GCM on devices without hardware acceleration, including most phones and routers.
  • Roaming is built into the protocolThe remote endpoint for a peer can change at any time; the tunnel picks it up on the next handshake rather than needing to be torn down and recreated.
  • Configuration is small and readableAn entire site-to-site setup fits on one screen per side. Reviewing a change is straightforward, which is unusual in VPN land.
  • Modern primitives onlyNo cipher negotiation, no downgrade, no legacy MD5 or 3DES. The opinionated choice makes WireGuard dull to deploy in the best way.
  • Well supported client ecosystemOfficial clients for Windows, macOS, iOS, Android and Linux, and the protocol is the foundation under Tailscale, Netbird and several commercial zero-trust offerings.

Where WireGuard bites

  • No built-in key distribution or revocationEvery peer is a public key, and rotating or revoking a key means editing config on every other peer. Teams end up building their own control plane or buying one.
  • MTU gotchas on the first deploymentThe default 1420 MTU is fine on clean Ethernet and causes silent TCP timeouts over PPPoE, cellular or anywhere else path MTU is lower, because handshakes work but large packets vanish.
  • Failed handshakes are quietA wrong preshared key, a clock skew beyond the replay window, or a public key typo produces almost no useful log output; you see retries and nothing else.
  • Dynamic routing is out of scopeThere is no OSPF, no BGP and no built-in dynamic peer discovery. For mesh topologies with many peers, you end up pairing WireGuard with something else.
  • NAT traversal needs keepalivesWithout a PersistentKeepalive value, the NAT mapping on an outbound peer expires after around a minute and inbound traffic silently drops.
  • Not every router's kernel is new enoughOlder OpenWrt, Mikrotik and pfSense builds either need a backport module or a userspace implementation, which erodes the throughput advantage.

Sound familiar?

Common WireGuard problems we fix

The WireGuard problems that bring people to us, and what you have once each one is fixed.

Handshake did not complete after 5 seconds

The tunnel will not come up

A wrong peer public key, a listening port that is firewalled on the server side, or a server clock that drifted far enough to fail the replay window check.

Handshake clean and quick

MTU 1420, path MTU 1280, large packets dropped

The tunnel is up but nothing bigger than ping works

The classic MTU fence: handshake and small packets succeed, every full-size TCP packet is silently dropped on the path, and connections sit until they time out.

MTU correct for the real path

persistent keepalive not set

Connections die after idle

A roaming client behind NAT with no keepalive loses its NAT binding after about ninety seconds, so inbound traffic from the server stops reaching it until the client speaks first.

Keepalive set where NAT demands it

AllowedIPs 10.0.0.0/8 overlaps

Routing stops working for unrelated hosts

A broad AllowedIPs entry on one peer that swallows traffic that should go to a different peer or to the local network entirely.

AllowedIPs tight and correct

wg: Unable to access interface

wg-quick fails to start on boot

A systemd unit ordering issue where wg-quick runs before the network stack is ready, or a non-privileged user running the command without CAP_NET_ADMIN.

Interface comes up on every reboot

endpoint for peer changed

Peers keep flapping endpoints

A roaming client on mobile data switching between carriers, or a NAT in front of the server mapping to different public addresses over time, which is normal but noisy if nobody expected it.

Flap handled, logs quiet

resolvconf: command not found

DNS settings are not applied

wg-quick relies on resolvconf or systemd-resolved to push the DNS line into the resolver, and distributions that ship neither leave the client with the old DNS after bringing the tunnel up.

DNS applied with the tunnel

File exists: RTNETLINK

Routes from a previous session block the new tunnel

A wg-quick down that did not run cleanly, leaving a route in the main table that refuses to be replaced on the next up.

Routes tidied before each start

What we do

WireGuard help, from one tunnel to a full mesh

Call for one broken peer, or hand us the whole VPN estate.

Rescue
Tunnels that will not handshake, flows that die under load, or client VPNs that broke on the last OS update. We stabilise the link first and write up why second.
Site to site design
AllowedIPs plans that do not overlap, MTU verified against the real path, failover between two upstreams, and routing that keeps working if a peer reboots.
Roaming client rollout
Mobile and laptop clients, provisioning that does not involve emailing config files, and keepalive values tuned to the actual NAT in front of each user.
Mesh and overlays
Choosing between hand-rolled WireGuard, Tailscale, Netbird or a self-hosted control plane, and running the migration when the answer changes.
Key rotation and audit
Rolling peer keys without downtime, revoking lost laptops, and leaving an audit trail that holds up on the next compliance review.
Managed Tech
Tunnel health alerts, key rotation on a schedule, config under version control, and a number to call when a remote office cannot reach the datacentre.

How it works

From first message to fixed

  1. Look Free estimate

    Tell us what's going wrong. We look at it, then tell you plainly what's wrong and what it will cost.

  2. Fix

    We fix the cause, not the symptom. If something is down, getting it back up comes first.

  3. Harden

    Backups, monitoring, updates and access, so the same problem can't quietly come back.

  4. Hand back or stay on

    You get a plain-language write-up. Or we stay on for hosting, upkeep and new work.

Often found together

Running WireGuard alongside something else?

We work on the whole stack, so the problem can be wherever it actually is.

Questions

WireGuard questions

Something else? Call+1 737 727 7327

Why will my WireGuard tunnel not complete a handshake?

In practice it is almost always one of four things: the peer public key does not match what the other side thinks it should, the UDP port is not reachable through the firewall on the listening side, the system clocks are more than a few minutes apart, or the preshared key was set on one side and not the other. The kernel logs will not tell you which, so we work through them in order.

Why does my WireGuard link work for ping but not for real traffic?

That is the classic MTU problem. The handshake and small packets get through, but every full size TCP packet is slightly too large for the path and is silently dropped. The fix is to measure the real path MTU and set the WireGuard MTU below it, often around 1280 on cellular or PPPoE.

Do we need a keepalive on every peer?

You need PersistentKeepalive on any peer that sits behind a NAT and wants inbound traffic to keep flowing when it is idle. Without it, the NAT mapping ages out after roughly ninety seconds and the server cannot reach the client until the client speaks. For server-to-server links on fixed IPs, keepalives are not needed.

Should we use raw WireGuard or Tailscale?

Raw WireGuard is excellent when you have a small number of peers and someone to own the key distribution. Tailscale is the sensible call once you have more than a handful of users, want single sign-on, or need ACLs and device posture without writing a control plane yourself. We help pick between the two rather than default to either.

How do we rotate a WireGuard key without downtime?

Add the new key as an additional peer entry on each side, let both sides accept it, cut clients over, then remove the old key. On a small deployment you can do it by hand; once there are more than twenty peers, it is worth generating config from a template and pushing it through Ansible or a similar tool.

How much does WireGuard help cost?

Pricing an engagement starts with a free call. Describe what the tunnel is doing and we come back with a written estimate before any hours are booked.

Free estimate

WireGuard giving you trouble?

Tell us what you're running and what's going wrong. A person reads it and comes back with an estimate.

  • No charge to find out what it will take
  • If it's urgent, say so

Replies come from a person. By sending, you authorize us to email you back. No marketing.