Awesome, I am glad to hear the provided info was helpful! Thank you for the quick investigation and explanation.

Again, thanks for the detailed reproducer; that made investigation straightforward. This appears to be simply an 'async' issue for FastAPI, the underlying web framework for vyos-http-api. FastAPI is very good at managing red/blue issues automatically, but in this case we need to explicitly annotate the endpoint method with async: an explicit lock does not appear necessary, though I will need to confirm. I'll provide a PR shortly. Thanks again !

PR: https://github.com/vyos/vyos-1x/pull/1820

Thanks Keving: https://vyos.dev/T5009

Error still present in 1.4: https://vyos.dev/T5004

FYI When I was troubleshooting I used:

In Both version, restarting relay service solved the issue:

PR with a fix: https://github.com/vyos/vyatta-cfg/pull/61

I can confirm this behavior, which occurs when changing IP address on listening interface (where dhcp-discover is captured).
Issue present in 1.3.2 .
It's also present in latest vyos-1.4-rolling-202302140317, regardless if old interface syntax is used, or if new upstream-interfces plus listen-interface commands are used.

@ammmze Thanks for the detailed report; allow me to investigate.

Here's the commands I ran for node-exporter:

https://github.com/FRRouting/frr/pull/12364
riw777 merged commit 91b6db4 into FRRouting:master Feb 14, 2023

You can either run both address-families through one tunnel

Is there any kind of ETA on this? It hasn't moved in a few months, and it is preventing me from being able to upgrade. I understand this probably isn't a huge priority, but an ETA would be nice.

In T5004#142577, @n.fort wrote:

Can you provide this configuration on both setups:

show config comm | grep relay

# And route to relay server
show ip route <relay_server>

Can you provide this configuration on both setups:

https://github.com/vyos/vyos-build/pull/306
https://github.com/vyos/vyos-1x/pull/1817
https://github.com/vyos/vyatta-cfg/pull/60
https://github.com/vyos/vyos-http-api-tools/pull/3

Then lets close it

I have prepared a configuration example using one of the latest 1.4 images, where more features were introduced.
Scenario and requirements:

• One vyos router
• 3 Uplinks to internet (eth0, eth1 and eth2). Static IP used on three links
• 2 VLANs
  • vif 2: + New Connections from vif-2 routed through WAN-2 + Server on vif 2 should accept ssh connections from internet, through dnat on 3 WAN interfaces (outside port 122)
  • vif 4: + NewConnections from vif-24routed through WAN-2 + Server on vif 4 should accept ssh connections from internet, through dnat on 3 WAN interfaces (outside port 222)

Don't see any issue with 1.3

vyos@r1:~$ monitor bandwidth-test initiate 192.168.122.14
------------------------------------------------------------
Client connecting to 192.168.122.14, TCP port 5001
TCP window size: 85.0 KByte (default)
------------------------------------------------------------
[  3] local 192.168.122.11 port 58042 connected with 192.168.122.14 port 5001
[ ID] Interval       Transfer     Bandwidth
[  3]  0.0-10.0 sec  6.37 GBytes  5.47 Gbits/sec
vyos@r1:~$ 
vyos@r1:~$ show version

Everything works on version 1.4

In T4774#142529, @c-po wrote:

Please note the WireGuard crypto Key routing concept: https://www.wireguard.com/#cryptokey-routing

Keys should not be re-used

Please note the WireGuard crypto Key routing concept: https://www.wireguard.com/#cryptokey-routing

@sdev @c-po @Alfa80 Can we look at rolling this back until it works fully, or fixing it so that it allows the above scenario please? Newer versions of 1.4 are unusable for me for a few months due to this. Let me know if I should file a separate bug.

Everything just worked fine.

Sorry about the trouble - fixed in next rolling release.

PR https://github.com/vyos/vyos-1x/pull/1814

PR https://github.com/vyos/vyos-1x/pull/1813

PR for 1.3 https://github.com/vyos/vyos-1x/pull/1813

PR for 1.3 https://github.com/vyos/vyos-1x/pull/1812