AnnotatedConnectException: Connection timed out: no further information

[mantisgaming]

I'm getting the error in the title occasionally when trying to connect to the minecraft server I created using this chart. The server runs fine mostly. The cpu usage is lower than I would expect though, but it runs fine except that people have trouble connecting occasionally.

I am running the deployment on a single node microk8s server that I have port forwarded to a nodeport service. I have 80Mbps internet (though that doesn't seem to matter since I get the error on the same network), and the server has 16 threads with a max frequency of 3.7GHz (though it seems to run around 1.6GHz typically with only 2 mincraft servers on it)

[itzg]

Most likely this will be caused by something external to the container.

FWIW I have almost an identical setup with microk8s, etc and was also seeing sporadic connection issues. In my case though, the server host itself was a WiFi hop away from ISP router due to physical constraints. I used tcpdump and observed a total lack of packets making it from the router to the host when the issue occurred.

I have since moved the host onto an Ethernet connection to router and issue went away.

Sorry, I don't have solid advice but can commiserate.

[mantisgaming]

That's not the issue, since the server is connected to ethernet. My desktop, which has the same issue, is connected to the same switch as the server.

I agree, I don't think the issue is with the minecraft server specifically, but I've only seen connection issues with the minecraft server. It seems like the messages are not being forwarded through the nodeport or something occasionally, but people who have connection to the server don't have any issues.

[mantisgaming]

This is what I see in wireshark when I'm trying to connect sometimes, but people are still connected to the server with no issues

(192.168.1.20 is my desktop and 192.168.1.120 is the server)

[itzg]

The "TCP Port numbers reused" seems suspicious. I haven't seen that before.

[mantisgaming]

I believe that is just the TCP message retransmitting because there was no ACK response

[NeuronButter]

Sorry to revive a dead thread, but I'm running into the same issue now 2 years later, did you guys have any more tips on how to troubleshoot this?

[mantisgaming]

I have since switched to a single-node kubeadm deployment of k8s. I don't remember if I was able to resolve the issue on microk8s, but I have not seen it since revamping my setup.

I don't think I ended up finding a solution for microk8s though, sorry.

[NeuronButter]

No worries! I'm on a single-node k3s at the moment so I'm suspicious it's some weird Flannel or ServiceLB/Traefik thing (since I apparently wanted to make things hard for myself and make a TCP IngressRoute lol). Considering on changing to a different CNI

[NeuronButter]

For the benefit of those who may stumble across this in the future, my error had nothing to do with the CNI or Ingresses 😅. It actually had to do with the host storage stalling and the client interpreting this as a timeout, switching the PV location to one locally on NVMe fixed the issue