This message was deleted.
# rke2a
adamant-kite-43734
05/06/2026, 10:35 AMThis message was deleted.
b
best-appointment-66450
05/06/2026, 11:47 AMI rather side-stepped that with keepalived and a floating IP. Not great, since it's only one at a time, but it was something.
best-appointment-66450
05/06/2026, 11:49 AMThat was for the Manager cluster. I haven't needed to get through to the managed clusters directly other than through the command line - all web interface things go through Rancher Manager.
b
bland-article-62755
05/06/2026, 2:25 PMCurious as to managed how? IIRC you could a custom cluster or have it be AWS EC2, Azure, Digital Ocean, Havester, Linode, or vSphere. I thought the Harvester LB mechanism was different from like how it would be set up for a custom one. (we side stepped this with kubevip and fqdns)
c
creamy-pencil-82913
05/06/2026, 4:01 PMWhile you MAY be able to override the address with your own config drop in on the nodes, it is definitely not supported and bad things will probably happen if you ever try to do a restore from snapshot via the UI.
✅ 1
creamy-pencil-82913
05/06/2026, 4:03 PMI would recommend just leaving it alone. The server address is only needed when joining the cluster, after that nodes use a local LB with cached endpoints for all comms with server nodes. So it'd be a lot of work for no real benefit.
✅ 1
❤️ 1
e
eager-refrigerator-66976
05/06/2026, 5:16 PMGot it, thanks! I was under impression that when kubelet / kubeproxy talking to 127.0.0.1:6443 (I assume it is rke2-agent/server) it just forwards it to the "server:" address.
eager-refrigerator-66976
05/06/2026, 5:17 PMIf it is using some internal LB than yeah there is no reason to do this.
c
creamy-pencil-82913
05/06/2026, 5:50 PMNope. There is a lb proxy. You can even see it in the logs.
✅ 1
2 Views