Windows tunnel shows established but traffic sometimes does not move after recycling tunnel

[Sander Saares]

Hi!

I tested again with latest Wireguard (.35) and Windows 2019 (.914) and the problem remains easy to reproduce.

In short, after restarting a Windows that is doing NAT for a peer behind a Wireguard tunnel, the tunnel stops correctly moving packets. I can see the Wiregurad "sent" and "received" counts increment to some degree but the end result is that pings do not work after some restarts of the tunnel/server).

I remain ready to assist in further diagnosis if a more informed person can guide me to what data might be useful.


Cheers,

Sander Saares, Advisor, Axinom
phone: +49 911 80109-54 | saares at axinom.com

-----Original Message-----
From: Sander Saares 
Sent: esmaspäev, 5. august 2019 10:55
To: 'wireguard at lists.zx2c4.com' <wireguard at lists.zx2c4.com>
Subject: Windows tunnel shows established but traffic sometimes does not move after recycling tunnel

Hi!

I submit a report on a problem encountered attempting to use WireGuard in a Windows-to-Windows "VPN gateway/proxy" deployment. I have a test deployment available in case I can provide further data for ease of debugging.

Scenario:
* Server A set up as WireGuard server, accepting connections from server B.
* Traffic from WireGuard network is forwarded and NATed by server A using built-in Windows networking features.
* Server B connects through WireGuard tunnel to access the internet.

For purpose of experimentation, the internet is defined as 8.8.8.8/32.

Expected result: tunnel is successfully established, internet traffic of server B is forwarded through server A.
Actual result: tunnel is successfully established (at least as shown in WireGuard GUI) but sometimes the expected traffic flows do not occur.
Occasionally, actual result matches expected result.

Method of observation: mutual ping on private IP address; ping from server B (WG client) to 8.8.8.8.
In failure case:
* both pings time out (server A and server B cannot ping each other on private IP)
* ping to 8.8.8.8 times out, EXCEPT for the first ping after tunnel is re-established (server B always seems to get 1 response before connectivity vanishes; possibly this is a ping not routed through the VPN, so it just goes directly out from server B to the internet?)
In success case, all pings work fine and get expected responses.

I suspect some startup/lifecycle/timing issue disrupting proper operation of the tunnel and/or associated routing configuration. If I can provide more data that may prove useful, I am happy to do so when instructed on how to collect it.

Configuration and experiment log follows.

Both systems are Windows 2019 (17763.652) running in clean Azure VMs, fully patched. WireGuard 0.0.19.

WireGuard server (server A)
=======================
[Interface]
PrivateKey = 
ListenPort = 9000
Address = 172.16.16.1/24

[Peer]
PublicKey = 
AllowedIPs = 172.16.16.0/24

WireGuard client (server B)
=======================
[Interface]
PrivateKey = 
Address = 172.16.16.2/24

[Peer]
PublicKey = 
AllowedIPs = 172.16.16.0/24, 8.8.8.8/32
Endpoint = xxx:9000

Forward+NAT setup (server A)
=================
PS C:\Users\saares> $interfaces = Get-NetIPInterface
PS C:\Users\saares> $interfaces[4]

ifIndex InterfaceAlias                  AddressFamily NlMtu(Bytes) InterfaceMetric Dhcp     ConnectionState PolicyStore
------- --------------                  ------------- ------------ --------------- ----     --------------- -----------
3       wg-test                         IPv4                  1420               5 Disabled Connected       ActiveStore

PS C:\Users\saares> $interfaces[4] | Set-NetIPInterface -Forwarding Enabled
PS C:\Users\saares> New-NetNat -Name NAT -InternalIPInterfaceAddressPrefix "172.16.16.0/24"

Name                             : NAT
ExternalIPInterfaceAddressPrefix :
InternalIPInterfaceAddressPrefix : 172.16.16.0/24
IcmpQueryTimeout                 : 30
TcpEstablishedConnectionTimeout  : 1800
TcpTransientConnectionTimeout    : 120
TcpFilteringBehavior             : AddressDependentFiltering
UdpFilteringBehavior             : AddressDependentFiltering
UdpIdleSessionTimeout            : 120
UdpInboundRefresh                : False
Store                            : Local
Active                           : True

Experiment log
===============
Immediate after setup -> all OK
Recycle tunnel on server -> all OK
Restart server PC -> tunnel reestablished but traffic does not move
Recycle tunnel on server -> all OK
Recycle tunnel on server -> all OK
Recycle tunnel on server -> all OK
Recycle tunnel on server -> all OK
Recycle tunnel on server -> all OK
Recycle tunnel on server -> all OK
Restart server PC -> all OK
Restart server PC -> tunnel reestablished but traffic does not move
Recycle tunnel on server -> tunnel reestablished but traffic does not move
Recycle tunnel on server -> tunnel reestablished but traffic does not move
Recycle tunnel on server -> tunnel reestablished but traffic does not move
Recycle tunnel on server -> tunnel reestablished but traffic does not move
Recycle tunnel on server -> all OK
Recycle tunnel on server -> tunnel reestablished but traffic does not move
Recycle tunnel on server -> tunnel reestablished but traffic does not move
Recycle tunnel on server -> tunnel reestablished but traffic does not move



Cheers,

Sander Saares
Advisor 
 
Axinom | Soola 8 | 51004 Tartu | Estonia
phone: +49 911 80109-54
saares at axinom.com | www.axinom.com 
 
Managing Directors: Sergei Gussev, Oleg Knut
Tartu Circuit Court, Reg. 11046287