Closed herbetom closed 3 years ago
I'd propose a config-parameter called vpn_proto
or something.
It'd be a string, with currently the valid value 'fastd'
and None
if unset defaults to 'fastd' for legacy purposes,
and then wrap the code @herbetom marked above with something like if 'fastd' == vpn_proto:
that way fastd would not be called if configured out.
Furthermore herbetom and I could evaluate, what we use instead of fastd and commit PRs for those protocols. Similar to how @genofire did in #27, respecting babel intends to replace batman while we replace the higher vpn protocols..
Any takers for the idea? @TobleMiner
The Node i got that error message from is one of our DNS Resolvers:
https://map.ffrn.de/#!/de/map/6aff94005402 (Domain 0) https://map.ffrn.de/#!/de/map/6aff94015402 (Domain 1) https://map.ffrn.de/#!/de/map/6aff94025402 (Domain 2)
We use Wireguard+VXLAN in between our VMs to get them to mesh with each other. This method is used to connect our Gateways, Map Server, the DNS Resolvers and the Unifi Controller.
Our Nodes currently still use fastd to connect to the Gateways. But we're looking into alternatives.
So the Gateways currently have two VPN Methods. But I have currently no need to represent both wg+vxlan and fastd. I'm fine with just no errors in the Logs of the Servers :sweat_smile:
It's a similar setup on our side as well. We've got supernodes that run fastd for clients as well as to communicate with each other and our related-Servers, and we've got supernodes, that speak WG as well as fastd for clients, and fastd for backend stuff.
I'm not sure, if it's a feasable approach to make the proposed parameter a list. For the software path that'd work well, but I've got to look into the statistics tab first; whether the counted packets are just batman related, which would be quite easy for us; or if they get counted on fastd-level.
@TobleMiner I don't think there's a problem with having multiple vpn protocols at the same time, do you? I'd therefore revise this to become a list of protocols. A short feedback would be greatly appreciated in order to go on in the right direction.
At least for what we intend for now; drop in replacements for fastd without layer breaking changes like for babel.
Currently mesh-announce doesn't really like it if fastd isn't installed on a system. This should probably be quieter: