i have unstable connection / ping losses with goodcloud s2s.
The mainnode is an MT2500 which is connected via CAT7 directly to my home router (FritzBox 6591 Cable).
The Node is an XE300 which is cullular connected via internal modem to Telekom network (over LTE with nice values - RSSI - 38 excellent, RSRQ - 10 good, RSRP - excellent, TYPE LTE, SINR 14 excellent).
Mainnode has the lan-ip: 192.168.6.1 and is connected to his gateway 192.168.178.1 (fritzbox). Behind the Mainnode i have a Server with lan-ip: 192.168.2.
Node has lan-ip: 192.168.30.1 and there is a windows-laptop connected as client - 192.168.30.50.
When i now try to ping from the server (192.168.6.2) to the client (192.168.30.50) i see sometimes ping lossed. When i ping from server (192.168.6.2) to the node directly, same ping losses.
In these moments pings from server to 8.8.8.8 and from client-side to 8.8.8.8 will be stable, so both sites has no connection issues - but in goodcloud vpn s2s there are ping losses.
Anyone has an idea? Maybe an mtu issue?
Screenshot from Server site:
If you stop and start s2s on goodcloud, will things be better?
To diagnose the issue, you can ping from the main node to the subnode, both via public IP, to see if there’s any packet loss.
looks like the problem is back.
We have terrible connection lost to our VPN-Nodes, LTE Connection are perfect.
What can we do to debug these problem?
We check internet connection of main nodes site at the same time we see these issue by ping 8.8.8.8, no icmp request was lost…
I would have a wild guess: You are using Vodafone?
They are terrible since Covid-19 because the network is completely overbooked. This cause package loss. WireGuard is UDP, which doubles the danger of package loss because client & sender don’t know if the package goes through.
Try to ping the main IP (not the WG one) from each device - so you will test the modem2modem connection. Will it fail as well?
No, we are using Telekom at Node-Sites (cellular-modem). We run also a connection check from the node side (ping from an client behind the node onto 8.8.8.8) to check connectivity - this works without interruptions.
On the server side (main node) yes we are using vodafone. But here we also check connectivity to 8.8.8.8 while we see the issue.
Is there any better solution to use an TCP-based VPN for that?
Trying to ping 8.8.8.8 is not really a good way to find out if there are issues.
ICMP (ping) isn’t TCP nor UDP. So it might get handled differently by the network.
You could try to use iperf (iPerf3 for beginners - #2 by admon) to test the speed via UDP between both devices.
8.8.8.8 is Google, so it’s mostly higher prioritized within the Vodafone network. They are doing so many shady things to keep the network up and running without investing into the backbone
Switching to TCP could (!) solve the issue. But TCP is only available via OpenVPN.
Make sure when you use -u that you also set -b to a reasonable bandwidth until you start seeing regular and excessive packet loss. This will better allow you to see what your actual performance and quality are better than tcp iperf tests.
