Usually when running a cloud environment, the nodes have an internal IP that is actually bound to an interface, and an external non-RFC1918 address that the internal IP is natted to for inbound access from the internet.

All of those address types can be set by flags. The hostname and internal IP assume reasonable defaults, but the external IP must be set manually or via a cloud provider

It's taking quite a while...

You're not really supposed to change node IPs once the cluster is up.

Wait! So I must set the

node-ip

before starting it?

I just uninstall and re-install the server. Then I use the

config.yaml

file from the previous installation before starting it. The server fails to start.

I re-install then run the server with the following:

sudo rke2 server --node-ip 192.168.56.101 --node-external-ip 192.168.56.101

Afterwards I can see the nodes up and running, some with the IP specified above. However when I run

systemctl status rke2-server

, it shows:

● rke2-server.service - Rancher Kubernetes Engine v2 (server)
     Loaded: loaded (/usr/local/lib/systemd/system/rke2-server.service; enabled; vendor preset: enabled)
     Active: inactive (dead)
       Docs: <https://github.com/rancher/rke2#readme>

Jan 18 14:35:52 vm-01 rke2[156065]: time="2023-01-18T14:35:52+07:00" level=info msg="Active TLS secret kube-system/rke2-serving (ver=321) (count 10): map[listene>
Jan 18 14:35:52 vm-01 rke2[156065]: I0118 14:35:52.581635  156065 event.go:294] "Event occurred" object="kube-system/rke2-metrics-server" fieldPath="" kind="Helm>
Jan 18 14:35:52 vm-01 rke2[156065]: I0118 14:35:52.602986  156065 event.go:294] "Event occurred" object="kube-system/rke2-metrics-server" fieldPath="" kind="Helm>
Jan 18 14:47:20 vm-01 systemd[1]: Stopping Rancher Kubernetes Engine v2 (server)...
Jan 18 14:47:20 vm-01 rke2[156065]: W0118 14:47:20.130271  156065 reflector.go:442] <http://k8s.io/client-go@v1.24.9-k3s1/tools/cache/reflector.go:167|k8s.io/client-go@v1.24.9-k3s1/tools/cache/reflector.go:167>: watch of *v1.Endp>
Jan 18 14:47:20 vm-01 rke2[156065]: time="2023-01-18T14:47:20+07:00" level=info msg="Shutting down <http://k3s.cattle.io/v1|k3s.cattle.io/v1>, Kind=Addon workers"
Jan 18 14:47:20 vm-01 rke2[156065]: time="2023-01-18T14:47:20+07:00" level=fatal msg="context canceled"
Jan 18 14:47:20 vm-01 systemd[1]: rke2-server.service: Main process exited, code=exited, status=1/FAILURE
Jan 18 14:47:20 vm-01 systemd[1]: rke2-server.service: Failed with result 'exit-code'.
Jan 18 14:47:20 vm-01 systemd[1]: Stopped Rancher Kubernetes Engine v2 (server).

What must I do to get it to work properly after setting

node-ip

?

So I run into this Github issue and follow the instructions there:

$ systemctl stop rke2-server
$ sudo rke2 server --cluster-reset
$ systemctl restart rke2-server

Afterwards, I can see RKE2 server status as ACTIVE. However running

kubectl get pods --output=wide -A

shows that

etcd-vm-01

/

kube-apiserver-vm-01

/

kube-proxy-vm-01

are now using the old

10.0.2.15

IP again. This seems to be the kubelet default behavior stated in this issue. Looks like everything will be fixed if there's a way to specify the interface that kubelet must use. How can it be done?

I have made some progress by using the following steps:

$ systemctl enable rke2-server
$ systemctl start rke2-server
... WAIT FOR ALL PODS TO BE READY
$ sudo touch /etc/rancher/rke2/config.yaml
$ sudo nano /etc/rancher/rke2/config.yaml
... EDIT CONFIG FILE
$ systemctl stop rke2-server
$ sudo rke2 server --cluster-reset --node-ip 192.168.56.101 --node-external-ip 192.168.56.101  --advertise-address 192.168.56.101
$ sudo reboot

When the pods are all good, there are 2 pods that still bind to the

10.0.2.15

interface/IP:

helm-install-rke2-canal
helm-install-rke2-coredns

They are shown as Completed. Is that something I must worry? How do I set them to the desired

192.168.56.101

?