https://rancher.com/ logo
#rke2
Title
# rke2
a

adamant-kite-43734

09/01/2022, 1:53 PM
This message was deleted.
v

victorious-analyst-3332

09/01/2022, 4:09 PM
they can technically be different, but it will lead to the nodes behaving differently 🤷
b

broad-farmer-70498

09/01/2022, 4:11 PM
kubelet-arg was probably a bad example, as that should be tweaked/set on the agents as well
v

victorious-analyst-3332

09/01/2022, 4:13 PM
examples of what I’ve tweaked in the past for things like metrics scraping from outside the cluster
Copy code
etcd-expose-metrics: true
kube-controller-manager-arg:
  - "bind-address=0.0.0.0"
kube-proxy-arg:
  - "metrics-bind-address=0.0.0.0:10249"
b

broad-farmer-70498

09/01/2022, 4:13 PM
yeah, I'm mostly referring to the things that are 'cluster' options like
cni: none
what happens if one of the server nodes doesn't have that and another does or they have different values etc? or should I only set that on the bootstrap node and then have quite minimal configs for all the other servers similar to the agent configs
v

victorious-analyst-3332

09/01/2022, 4:15 PM
those values aren’t propagated that I know of, so you’ll need them on all nodes that you plan to interact with in that way
things like
cni: none
are part of the rancher cluster config that get automatically pushed down to the clusters I believe
b

broad-farmer-70498

09/01/2022, 4:16 PM
so that's a 'last man wins' scenario as well?
v

victorious-analyst-3332

09/01/2022, 4:17 PM
no idea in that scenario
I think best practice is to make sure they match to prevent issues if something goes offline
👍 1
22 Views