beszel via tailscale

[ziggysdomain]

I could have sworn I had this working for a short while via tailscale but now, despite numerous container and image deletions and reinstallations, I cannot get beszel to display my two VPSs in different locations to the hub. I'm using the tailscale ip for the VPS in the hostname. I've just discovered Alex Kretzschmar's tailscale video for beszel:
https://www.youtube.com/watch?v=O_9wT-5LoHM
So in the meantime hoping there's a quick solution here. Thanks.

[henrygd]

Please let me know what the connection error is on the Logs page in PocketBase (<your-url>/_/#/logs).

[henrygd]

Same command when you got the SSH-2.0-Go message, just run from inside the VM.

telnet <tailnet address> <port>

[ziggysdomain]

Yeah telnet working between all those tailscale addresses. I have installed a new ubuntu server and interestingly the local hub is showing its data and one of the remote VPSs, but the second remote is not. This latter one is headless (as they all are). The logs for that one report:
ssh: handshake failed: ssh: unable to authenticate, attempted methods [none publickey], no supported methods remain
I can ssh into that with the tailscale ip no problem. Probably something obvious I'm not thinking of here.

[henrygd]

unable to authenticate means it reached the agent but the key was denied.

Please double check that the KEY env var on the agent has the correct value. If you installed the hub fresh then maybe the agent still has the value of the key from the previous hub.

[ziggysdomain]

I checked and all key values are the same on all servers. I removed the container and image on the remote, redid the add system on the dashboard, copied the yml file to the remote, redid docker compose up -d, and still the log says:
ssh: handshake failed: ssh: unable to authenticate, attempted methods [none publickey], no supported methods remain
One day I hope to put zfs on that remote server so that I can revert a snapshot to a previous uncorrupt state as I suspect this would fix issues like this. Could be wrong, however. Again, appreciate your time.

[ziggysdomain]

How do I do that? Bear in mind that another remote server using tailscale is working, this one isn't. If tailscale were intercepting traffic would it not be doing it on both? It has just occurred to me - in case this is relevant - the non-reporting server has an exit node for tailscale. Could this be a problem?

[BreckanM]

check that the tailscale replacement SSH isnt intercepting your traffic!

[MisterSeagull]

I'd like to chime in and say that I am having the same issue accessing my nodes connected over Tailscale. I've had Beszel up for a week with no issues. This morning at 3:20am all of my Tailscale nodes suddenly went offline simultaneously in Beszel. I am able to access them through all other channels but can no longer connect those agents to my Beszel hub. I have not made any changes to my ACLs, and I have double checked everything.

My logs display the same dial tcp 100.xx.xxx.xxx:45876: i/o timeout error that @ziggysdomain was getting and was able to get the same SSH-2.0-Go message when I tried to telnet <ip> <port>

I'm guessing that whatever changed has changed on the Tailscale side of things since I've made no configuration changes or updates.

I have multiple other nodes and the only nodes offline are the Tailscale connected ones.

UPDATE:
I was able to resolve this by switching my hub network_mode to host.

[henrygd]

Thanks for the update!

I don't understand enough about how Tailscale works to know why that would fix it. I'll try to do some research this weekend if I have time.

Are you using hostnames for these systems or IP addresses?

[MisterSeagull]

I’m using the tailscale IP addresses

[ziggysdomain]

I'd like to chime in and say that I am having the same issue accessing my nodes connected over Tailscale. I've had Beszel up for a week with no issues. This morning at 3:20am all of my Tailscale nodes suddenly went offline simultaneously in Beszel. I am able to access them through all other channels but can no longer connect those agents to my Beszel hub. I have not made any changes to my ACLs, and I have double checked everything.

My logs display the same dial tcp 100.xx.xxx.xxx:45876: i/o timeout error that @ziggysdomain was getting and was able to get the same SSH-2.0-Go message when I tried to telnet <ip> <port>

I'm guessing that whatever changed has changed on the Tailscale side of things since I've made no configuration changes or updates.

I have multiple other nodes and the only nodes offline are the Tailscale connected ones.

UPDATE: I was able to resolve this by switching my hub network_mode to host.

Yeah, and the strange bit for me is that I have two external (to the LAN on which the beszel hub sits) VPSs with the agent running, both using tailscale as the ip and one works and the other no.