Hello, I need some help setting up a K3S node using the "Create Cluster" functionality. I want to know if I can set the following config so it's ready to use the GPU: • Add

--default-runtime=nvidia

to the K3S cli. In a regular install this is done by adding it to the

ExecStart

parameter of

/etc/systemd/system/k3s.service

• Set the following label to the node:

<http://nvidia.com/gpu.present=true|nvidia.com/gpu.present=true>

.

👋 Hi everyone!

Hi, Is it possible to run Rancher (Community Edition) in Kubernetes Cluster? Can someone help me sharing the link as I could only find docker based deployment. Thanks in advance. DSK.

Hey folks, Just a quick heads-up – I'm not sure if you're already aware or working on it, but https://get.k3s.io seems to be down. Cheers!

Cloudflare is experiencing an outage

image.png

Looks like someone from the Cloudflare dev team was vibe coding a bit too hard 😄. Time to take a coffee break, then.

I am currently evaluating Rancher CE to provision a RKE2 Cluster on two VMs. I am using 1 VM as Control plane and etcd. And I am using another VM as Worker Node. Rancher CE is hosted in single node K3s Cluster on a VM All 3 VMs mentioned above have full connectivity with eachother. Now the issue is when I run poisoning script my etcd and control plane node is stuck at

waiting for probes: kube-scheduler

. I have tried reinstalling as well that didn't help either. I have attached screenshot of provisioning logs

Hi folks, After downgrading from 1.19 to 1.9, following the faq, I cannot start the VM now. This is the error information. Anyone has idea for it?

After starting the laptop, the starting VM lasted 5min, then showed this error

Hi folks, I have a rancher deployed RKE2 (1.32.5) cluster... I need to create a service account on this downstream cluster that can list all pods in all namespaces Below is what I have, but it does not seem to work when I check with:

kubectl auth can-i get pods --all-namespaces --as my-service-account

What am I missing?

apiVersion: v1
kind: ServiceAccount
metadata:
  name: my-service-account
  namespace: default  # You can specify any namespace you prefer
---
apiVersion: <http://rbac.authorization.k8s.io/v1|rbac.authorization.k8s.io/v1>
kind: ClusterRole
metadata:
  name: get-pods-clusterrole
rules:
- apiGroups: [""]
  resources: ["pods"]
  verbs: ["get", "list"]
---
apiVersion: <http://rbac.authorization.k8s.io/v1|rbac.authorization.k8s.io/v1>
kind: ClusterRoleBinding
metadata:
  name: get-pods-clusterrolebinding
subjects:
- kind: ServiceAccount
  name: my-service-account
  namespace: default  # Use the same namespace where the service account was created
roleRef:
  kind: ClusterRole
  name: get-pods-clusterrole
  apiGroup: <http://rbac.authorization.k8s.io|rbac.authorization.k8s.io>

Hi all, I migrated to a harvester cloud provider setup with my downstream rancher guest cluster. I'm struggling to get my workloads to be accessible via a harvester loadbalancer. Issue is my Devs are waiting as this is the QA environment and we can't access our workload now.. I need some clarification on if I can use a vlan cluster network for the load balancer? IE the VM network vs the management network? I can't seem to get it to be routable via the VM network, I have my service in the downstream guest and I have the load balancer automatically provisioned in harvester with IP on the VM vlan network but it can't seem to get up to the guest cluster. Not sure how to troubleshoot from here or what's required for the networking. My rancher master VMs are on the same network as my upstream guest cluster nodes which are connected via the vm-net harvester bridge and bonded interface to my switch as a vlan trunk. I'm not sure if the harvester nodes need to have a vlan interface on that network too? Is that best practice? I thought harvester would setup what it needed in that case as well. The alternative seems to be to use the management network and route the cluster gateway traffic over that but It doesn't seem like best practice either.. Any more info about the harvester cloud provider guest cluster load balancer setup would be much appreciated.

hello, is it possible to create a cluster-wide role via role templates? E.g., I would like to create a role to allow list all CRDs, which must be the clusterwide role. If I just add it as another roletemplate, it does not work. Creating some namespace scoped role works as expected. I did not find a word about this in rancher documentation.

Hello everyone;

I having some trouble with longhorn, after the installation I keep getting "Request failed with status code 502"

I have went into the logs of the pods but they are not showing anything..

I believe I'm moving forward, the 503 error is no longer showing now I'm getting "Instance manager error" on 3 of my nodes..

But now I run into another problem

problem*

I have added extra storage that's in a folder /mnt/bababa/

Using the UI I added them and the show "Schedulable" but when I go back to the dashboard tab they shows "Reserved" why?

Hi everyone! I built this "Ask AI" chatbot where I gave a custom LLM access to all SUSE dev docs to help answer technical questions for people using SUSE. I tried it on a couple of questions here in the forum, and it answered it within seconds. Feel free to try it out here: https://demo.kapa.ai/widget/suse Would love to hear your thoughts on it!

how do you guys manage UPS stuff (like nut) in harvester? Do you just like run something that ssh's in and tells nodes to isolate/shutdown/whatever?

@astonishing-address-98404 depends on the UPS, I use Vertiv systems here they provide an rpm, systemd service etc

Yeah this is my house. i have some random cyberpower things that work with Nut. i didn't know if it was sane to try to install stuff on harvester nodes. i was considering usb pass through to a vm or something similar but i've struggled with passing through usb devices (like sdr, radios, things that i turn off/on).

i'm thinking like a rpi5 or something plugged into the ups and then just having it ssh into nodes and telling them to shutdown. i built this to play with k8s/rook-ceph. which means the rest of my life is now consumed with adding more nodes and disks because ceph.

@astonishing-address-98404 if using kvm, USB passthrough shouldn't be a problem...

i run some windows/linux vm(s) to do radio junk (like sdr, and icom radios, etc). I don't like to leave my radios running when i'm not messing with them so the usb devices changes. i have to go like add/remove the devices which means it has to restart the vm(s).

@astonishing-address-98404 I have ads-b setups running 24/7

neat! I run unittrunker/trunking recorder and wsjtx/gridtracker.