https://rancher.com/ logo
Join Slack
Powered by
This message was deleted.
# k3s
a
This message was deleted.
c
Do you have networkmanager or connman installed?
w
no, I don’t think so. They’re both disabled by default in NixOS and I haven’t enabled them.
c
Did you for some reason use 169.254.x as your cluster cidr?
I don’t know what would be adding all those other than some network management tool that’s not container network aware
w
No, the cluster CIDR is in the default range (you're asking about the one under 
--cluster-cidr
, yes?)
c
Are you using Docker alongside k3s? those kinda look like docker network interfaces
w
nope, it's containerd
almost everything in k3s is in the default configuration
c
you’re sure docker isn’t also running?
I don’t know what else those would be for
w
Docker isn't even installed
the number of 
vethXYZ
routes corresponds to the number of containers running on each host. Pretty sure this is flannel.
in fact I switched flannel from wireguard to vxlan mode and it added a new one named 
flannel.1
c
flannel won’t add those APIPA addresses though
if that’s not it, I’d ask the nixos folks. this is not anything k3s related.
w
hmm. OK, well, thanks for the pointers!
just to close the loop here, I investigated that DHCP stuff you mentioned and it turned out to be it. Turns out NixOS enables DHCP on all interfaces by default. I'm not sure about the consequences there still, but turning it off on 
veth*
seems to have fixed things.
c
That seems like something it should do by default.
w
it sounds like it is up for reconsideration in an upcoming release
c
Wrong place, can’t delete, so just edited. Sorry.
Open in Slack
PreviousNext
Powered by