Lucas Servén Marín

Thanks @hhstu so there is indeed a kilo0 interface available. Some things that come to mind: What differences are there between this node and the one that is working? Different...

Also, @hhstu does this work if you pin Kilo to 0.3.1? I wonder if this might be due to the switch to using a different WireGuard client library.

Thanks @hhstu those logs are a bit more helpful. `Unable to access interface: Protocol not supported` seems to be a pretty common symptom of WireGuard problems.

Hi @hhstu yes, I'd be very interested in taking a look at a PR for that 👍. I'm curious how/why it's different. Our E2E tests run on KinD, which uses...

These are all great ideas for optimizations @SerialVelocity. 1 is certainly trivial to implement, however it will continue to produce very spiky behavior, albeit less often. 3 would be the...

> There is another option which is to just have blanket rules for the node subnet, and the cluster subnet. That way you only have two rules? This is tricky...

btw, if you'd like to try out the test build a made, you can try running this image: `squat/kilo:13a25891de81a4ae8a0079285799fce492c06ef4`

I'll try to post tonight 👍 Btw is there any chance you could open a PR to merge your branch into upstream and let the kg agent set up non-node...

I would like to avoid adding extra flags for every possible re-sync period. Maybe we can have one global flag that sets the re-sync periods for all of Kilo's control...

There are three separate controllers: * iptables * ip routes + rules * main Kilo controller, which in turn sets: * wireguard config * ip address The next controller that...