Closed bbigras closed 1 year ago
So the second issue is a byproduct of the first issue. With a clean shutdown - it will attempt to remove the interface for you. There is the --wireguard.force-interface-name
flag/config that will delete it on startup if it sees it. You can also customize the name of the interface. The Join error was it not caring what the kernel error was and trying to fall back to TUN, which I guess was not available wherever this was running.
I don't know exactly what that first issue was - but I'd be curious for logs. It feels pretty unavoidable, but maybe can at least make a clean exit happen. You took its data directory out from under it. On nodes where you don't care about persisting data locally you can use the --raft.in-memory
flag.
Just tidying things up. Gonna close this issue for now. If you happen to think of any ways disappearing storage can be dealt with, feel free to open a new issue.
Worth noting, that I hope to revamp a lot of how the storage works in favor of a distributed implementation.
If I delete /var/lib/webmesh by mistake while webmesh-node is running, I can't kill the node unless I use
-9
.And if I start webmesh-node again, I get
Error: failed to open mesh connection: join: fatal join error starting network manager: new wireguard: new interface: new tun: create tun: invalid argument
.I can start the node again if I run
sudo ip link delete webmesh0
.