Use uc ctx conn if your uncloud cluster is unresponsive
So, uncloud is obviously still a work in progress, so bugs will happen. Because I've drunk the cool-aid and dived into using it on my homelab, I'm doing my best to be a good netizen and helping troubleshoot any issues I find.
Today, I had a great session with the author, where my cluster was unresponsive, except for running commands like uc machine ls. The services were still running, but I'd lost the ability to perform new deploys or check the state of services. Seems like a bug, and might be related to WireGuard + Tailscale + Uncloud and IP6 addresses.
Along the way, though, I learned a new command, which got me completely unblocked:
uc ctx conn
Run this and you get a list of all the machines in your cluster and you can choose a different default machine to proxy traffic through. In my case, it seems like there's something wrong with only the Ubuntu machines I have… the others on DietPi work just fine.
One caveat I had, though: because my machines are behind NAT on a home network, I have ports 80 and 443 redirected to the “default machine”, so I had to change that redirection to point to the local IP address of the new default machine, then do a new deploy of whatever services I needed to update. I believe the issue here is probably to do with registering the Let's Encrypt certificate… they require the machine to be discoverable for that to complete.