https://rancher.com/ logo
Join Slack
Powered by
This message was deleted.
# k3s
a
This message was deleted.
c
It should just be
Copy code
apiVersion: <http://provisioning.cattle.io/v1|provisioning.cattle.io/v1>
kind: Cluster
spec:
  rkeConfig:
    machineGlobalConfig:
      disable-cloud-controller: "true"
a
Thanks for the reply. I read that I shouldn't disable this globally but only on the CP nodes? Is that not true as well? Additionally is there a way to set this in the UI without having to edit the YAML? It's ok if the answer is now, I just like to stick to the form as much as possible if I am able.
c
technically it is a server arg and is only valid on servers but it is fine to set it for all nodes, the agents will just ignore it
👍 1
no, if there’s no UI option for it then you get to edit yaml
there are a LOT of advanced config options for k3s and rke2, most of them are not in the UI
a
Rancher is having trouble provisioning the new cluster. It's stuck waiting for the first master node to check-in. I filled out the form the exact same way that I have the last dozen or so times with the only change being 
disable-cloud-controller: true
. Am I missing anything else? Any node labels or taints I need to apply to the CP or worker nodes or both?
c
yes, if you disable the built in cloud controller the nodes will remain tainted as uninitialized until you deploy a replacement. beacuse they are tainted rancher cannot run
this is covered in the docs
You need to ensure that your replacement is deployed during initial cluster startup - the best way to do that is usually via HelmChart manifests that are set with 
bootstrap: true
a
Would I place that HelmChart manifest in the "Cluster Configuration" -> "Additional Manifest" section? I added it there but nodes still don't finish provisioning.
c
yes. what’s the YAML look like?
a
I've made several revisions to it as AI put on the wrong path. I think I'm getting close. Thanks again for getting back to me. I'll post more if I continue to hit roadblocks that I don't understand how to get around.
Hey Brandon. Just wanted to let you know that I was able to get my issue fixed. It took a lot of customization via the "Additional Manifest" section and two small tweaks via the YAML editor. I'll spare you the details unless you really want to know. Essentially I was able to use the Rancher UI/form to provision a whole k3s cluster on VMware infra that disables the built-in k3s ccm and replaces it with a fully functional 
vsphere-cpi
and 
vsphere-csi
. It is sort of particular to my env but that was always my goal. Only use Rancher with no post-provisioning user interaction required. It's pretty cool what I was able to accomplish by only using Rancher.
c
rad! FWIW that is all built in to RKE2, you just have to pick the rke2 cloud-provider in the UI and it deploys vsphere cpi and csi charts
a
Me personally, and my company, started off using k3s years ago so it's always been the most familiar to me and therefore I've stuck with it. I know RKE2 is suppose to be pretty similar but I've never researched the main differences between the two to give a second look. And I'm so busy and "in-bed" with k3s that I don't have the luxury of time to really look into it (nor the desire to refactor after being able to accomplish so much with k3s).
Thanks again for always being here to help when I run into a hard problem 🍻
c
oh don’t get me wrong, I love k3s and I think it is a great choice. RKE2 just has some stuff baked into it that K3s doesn’t
13 Views
Open in Slack
PreviousNext
Powered by