Common issues and Debugging

This page provide the most common pitfalls users might have with VXLAN tunnel. This page is largely under construction as we are gathering more edge cases. Your feedback is much appreciated.

1. VXLAN Setup Check List

It is crucial to ensure that you can directly connect to the services or games you are hosting using their IPv4 address and port. For instance, in this setup documentation, the Ragnarok server was accessible via 108.61.149.182:6900. If your VXLAN tunnel is not functioning correctly, several potential causes may exist. Below are some best practices to follow to confirm that your service is properly set up and running.

1.1. Verifying tunnel creation

After running the setup script, you can run the following commands to check your tunnel configuration:

ip -s link show vxlan_<vxlan ID>

Example output:

root@admin:~# ip -s link show vxlan_47
418: vxlan_47: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1450 qdisc noqueue state UNKNOWN mode DEFAULT group default qlen 1000
    link/ether 12:cc:cb:ab:1f:e8 brd ff:ff:ff:ff:ff:ff
    RX: bytes  packets  errors  dropped missed  mcast   
    543784518  3037392  0       0       0       0       
    TX: bytes  packets  errors  dropped carrier collsns 
    154324621  1408584  0       0       0       0       

You should see 0 under the errors and dropped columns. Any other numbers would indicate packets being dropped (possibly from a firewall).

1.2. Interface traffic

To verify that your VXLAN tunnel has been properly established, you can ping the Private IP Address of your tunnel. You should see responses like the example below:

root@admin:~# ping 172.18.128.13
PING 172.18.128.13 (172.18.128.13) 56(84) bytes of data.
64 bytes from 172.18.128.13: icmp_seq=1 ttl=64 time=49.4 ms
64 bytes from 172.18.128.13: icmp_seq=2 ttl=64 time=49.4 ms
64 bytes from 172.18.128.13: icmp_seq=3 ttl=64 time=49.7 ms
64 bytes from 172.18.128.13: icmp_seq=4 ttl=64 time=49.4 ms
^C
--- 172.18.128.13 ping statistics ---
4 packets transmitted, 4 received, 0% packet loss, time 3005ms
rtt min/avg/max/mdev = 49.362/49.458/49.722/0.152 ms

The local IP address here corresponds to the VXLAN interface running on our anycast server, so the latency you observe will vary depending on your backend’s location. For further debugging, run:

arping -I vlxan_<vxlan ID> <private IP address>

as it can confirm whether the virtual link is properly resolving IP-to-MAC, something ping won't be able to show:

root@admin:~# arping -I vxlan_47 172.18.128.13
ARPING 172.18.128.13
42 bytes from 02:ab:cd:34:56:78 (172.18.128.13): index=0 time=44.839 msec
42 bytes from 02:ab:cd:34:56:78 (172.18.128.13): index=1 time=44.860 msec
42 bytes from 02:ab:cd:34:56:78 (172.18.128.13): index=2 time=44.852 msec
42 bytes from 02:ab:cd:34:56:78 (172.18.128.13): index=3 time=44.826 msec
42 bytes from 02:ab:cd:34:56:78 (172.18.128.13): index=4 time=44.827 msec
42 bytes from 02:ab:cd:34:56:78 (172.18.128.13): index=5 time=45.180 msec
42 bytes from 02:ab:cd:34:56:78 (172.18.128.13): index=6 time=45.047 msec
^C
--- 172.18.128.13 statistics ---
7 packets transmitted, 7 packets received,   0% unanswered (0 extra)
rtt min/avg/max/std-dev = 44.826/44.919/45.180/0.129 ms

In this example, the private interface IP resolves to a MAC address within the tunnel, which is 02:ab:cd:34:56:78

As for your tunnel's Public IP Address , the reply should looks something like this (with <1ms latency):

root@admin:~# ping 104.234.6.128
PING 104.234.6.128 (104.234.6.128) 56(84) bytes of data.
64 bytes from 104.234.6.128: icmp_seq=1 ttl=64 time=0.033 ms
64 bytes from 104.234.6.128: icmp_seq=2 ttl=64 time=0.046 ms
64 bytes from 104.234.6.128: icmp_seq=3 ttl=64 time=0.060 ms
64 bytes from 104.234.6.128: icmp_seq=4 ttl=64 time=0.061 ms

You can also check the output of:

ip route
root@admin:~# ip route
...
172.18.128.0/24 dev vxlan_47 proto kernel scope link src 172.18.128.13 
192.168.0.0/24 dev eth0 proto kernel scope link src 192.168.0.89 
...

If the tunnel was properly created, you will see your private IP address in the output.

1.3 Firewall Configuration

There are two potential scenarios:

  1. Cloud Provider Firewall: Major cloud providers like Linode, AWS, and Azure have default security group or firewall settings that may block traffic before it even reaches your VPS. In this case, you can navigate to your provider's control panel and allow traffic from TCPShield to your backend ports.

  2. Server Firewall: If you are using a server firewall such as netfilter, iptables, or ufw, you will need to open both the backend port and the VXLAN port to accept external connections, or, in rare cases, explicitly whitelist the VXLAN interfaces.

Using iptables to Accept Connections to Your Backend Port

iptables -A INPUT -p udp --dport <PORT> -j ACCEPT
iptables -A INPUT -p tcp --dport <PORT> -j ACCEPT
iptables -A OUTPUT -p tcp --sport <PORT> -j ACCEPT
iptables -A OUTPUT -p udp --sport <PORT> -j ACCEPT

Using UFW :

ufw allow <port>/udp
ufw allow <port>/tcp

Whitelisting VXLAN Interface (Optional)

Although this step is typically not necessary, you can whitelist the VXLAN interface explicitly using iptables:

iptables -I FORWARD -i vxlan_<vxlan ID> -j ACCEPT
iptables -I FORWARD -o vxlan_<vxlan ID> -j ACCEPT

Using UFW:

sudo ufw route allow in on vxlan_<vxlan ID> to any
sudo ufw route allow out on vxlan_<vxlan ID> to any

2. Error: Nexthop has invalid gateway (Pterodactyl / Wings).

If you're running Pterodactyl (or other Docker-based services), you might encounter the error:

Nexthop has invalid gateway

This usually happens when the VXLAN tunnel attempts to bind to an IP address that is already reserved by Docker. For example, Pterodactyl’s default bridge network (pterodactyl0) may overlap with the subnet used by VXLAN (172.18.0.0/16), causing the tunnel setup to fail.

To debug:

  1. Check the Local IP Address for VXLAN: It should resemble something like 172.18.132.2.

  2. Run ip route and you may see an entry like:

172.18.0.0/16 dev pterodactyl0 proto kernel scope link src 172.18.0.1 

In this case, the 172.18.132.2 IP address might conflict with the Docker network configurations. For example, if the pterodactyl0 interface is using 172.18.0.1, this could block the gateway you are trying to use (172.18.132.2), as it lies within the Docker subnet.

✅Solution: Switch Pterodactyl to use host networking

To prevent Docker from reserving IPs that interfere with your VXLAN setup:

  1. Edit Pterodactyl config Open /etc/pterodactyl/config.yml and modify the Docker networking section:

    docker:
      network:
        name: host
        network_mode: host
  2. Restart the network stack and Wings daemon

    systemctl stop wings
    docker stop $(docker ps -q)
    docker network rm pterodactyl_nw
    systemctl start wings

This change puts all containers on the host network, removing the isolated Docker bridge and preventing subnet conflicts. Your VXLAN tunnel should now initialize without errors.

3. Tunnel deletion

In case you need to delete the tunnel and re-run the setup script, here is how:

ip link del dev vxlan_<vxlan ID>
ip rule del fwmark 9 table 200

You can also remove the iptables rules in the script, by replacing -A and -I with -D , example:

iptables -t mangle -D OUTPUT -s 104.234.6.137/32 -j MARK --set-xmark 0x9
iptables -t mangle -D POSTROUTING -s 104.234.6.137 -j MARK --set-mark 0

4. VXLAN tunnel stopped working on server reboot

Sometimes, after rebooting your VM or dedicated server, the VXLAN script may not re-run, causing the service to stop working. To fix this, you can set up a persistent service to automatically run the VXLAN script after each reboot.

Steps to Setup Persistent VXLAN Tunnel:

  1. Edit the rc.local file:

sudo nano /etc/rc.local
  1. Add the following lines to execute the VXLAN script at boot:

#!/bin/bash
# rc.local - VXLAN auto setup at boot
sleep 5
<INSERT YOUR VXLAN SCRIPT HERE>
echo "VXLAN script executed"
exit 0
  1. Save and exit the editor (Ctrl+X, then Y to confirm).

  2. Make the script executable:

sudo chmod +x /etc/rc.local

Now, the VXLAN script will run automatically after each reboot, ensuring that the tunnel is re-established.

5. Disable Reverse Path Filtering

By default, most Linux distributions have rp_filter (Reverse Path Filtering) enabled. This is a kernel security feature that drops packets if their source address does not match the expected interface route — a defense mechanism against IP spoofing.

However, in setups using VXLAN or asymmetric routing, this behavior can cause legitimate packets to be dropped if the kernel thinks they arrived from the "wrong" interface. During a tcpdump capture, you may see packets coming in, but no responses being sent by the kernel. For example:

root@admin:~# sudo tcpdump -nni any udp port 19132
tcpdump: data link type LINUX_SLL2
tcpdump: verbose output suppressed, use -v[v]... for full protocol decode
listening on any, link-type LINUX_SLL2 (Linux cooked v2), snapshot length 262144 bytes
15:15:30.707370 vxlan_47 In  IP 14.231.20.10.62338 > 104.234.6.137.19132: UDP, length 33
15:15:43.571573 vxlan_47 In  IP 14.231.20.10.63192 > 104.234.6.137.19132: UDP, length 1172
15:15:44.084121 vxlan_47 In  IP 14.231.20.10.63192 > 104.234.6.137.19132: UDP, length 1172
15:15:44.598862 vxlan_47 In  IP 14.231.20.10.63192 > 104.234.6.137.19132: UDP, length 1172

As shown above, there is only ingress traffic from 14.231.20.10, but no outgoing traffic from our VXLAN interface 104.234.6.137. This means the system is silently dropping the replies.

To fix this, disable reverse path filtering for all interfaces and for your VXLAN interface specifically:

sysctl net.ipv4.conf.all.rp_filter=0
sysctl net.ipv4.conf.vxlan_47.rp_filter=0

This disables strict source route validation, allowing the system to accept packets even if their return path differs — which is common in tunneling setups like VXLAN.

Once reverse path filtering is disabled, you should start seeing bi-directional traffic over your VXLAN tunnel. A successful exchange will show both incoming and outgoing packets during a tcpdump capture. Expected result:

root@admin:~# sudo tcpdump -nni any udp port 19132
tcpdump: data link type LINUX_SLL2
tcpdump: verbose output suppressed, use -v[v]... for full protocol decode
listening on any, link-type LINUX_SLL2 (Linux cooked v2), snapshot length 262144 bytes
20:02:32.691465 vxlan_47 In  IP 14.231.20.10.55407 > 104.234.6.137.19132: UDP, length 1164
20:02:32.691550 vxlan_47 In  IP 14.231.20.10.55407 > 104.234.6.137.19132: UDP, length 1164
20:02:32.691665 vxlan_47 In  IP 14.231.20.10.55407 > 104.234.6.137.19132: UDP, length 1112
20:02:32.696211 vxlan_47 Out IP 104.234.6.137.19132 > 14.231.20.10.55407: UDP, length 43
20:02:32.726227 vxlan_47 Out IP 104.234.6.137.19132 > 14.231.20.10.55407: UDP, length 550

Last updated

Was this helpful?