You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
This may be a mistake on my side but I've set up a cluster with only one master node. I've installed the NVIDIA drivers as well as the container runtime. I've verified that containers can access GPU both through docker as well as containerd.
Environmental Info:
K3s Version:
➜ / k3s -v
k3s version v1.29.6+k3s2 (b4b156d9)
go version go1.21.11
Node(s) CPU architecture, OS, and Version:
uname -a
Linux psduyqnrwtts 5.15.0-113-generic #123-Ubuntu SMP Mon Jun 10 08:16:17 UTC 2024 x86_64 x86_64 x86_64 GNU/Linux
Cluster Configuration:
Just one master node:
NAME STATUS ROLES AGE VERSION INTERNAL-IP EXTERNAL-IP OS-IMAGE KERNEL-VERSION CONTAINER-RUNTIME
psduyqnrwtts Ready control-plane,master 23m v1.29.6+k3s2 10.64.4.86 184.105.5.117 Ubuntu 22.04.2 LTS 5.15.0-113-generic containerd://1.7.17-k3s1
Describe the bug/Steps To Reproduce
Installed container runtime & drivers
verified nvidia-smi throuhg host machine, and containers through both docker & containerd
started the master node
curl -sfL https://get.k3s.io | INSTALL_K3S_EXEC="--node-external-ip=$(master_ip) --flannel-backend=wireguard-native --flannel-external-ip" sh -
kubectl describe node <my node> doesn't show the gpu:
I applied the pod specified on the website itself, and upon inspecting it:
Events:
Type Reason Age From Message
---- ------ ---- ---- -------
Warning FailedScheduling 23m default-scheduler 0/1 nodes are available: 1 Insufficient nvidia.com/gpu. preemption: 0/1 nodes are available: 1 No preemption victims found for incoming pod.
Warning FailedScheduling 2m59s (x4 over 17m) default-scheduler 0/1 nodes are available: 1 Insufficient nvidia.com/gpu. preemption: 0/1 nodes are available: 1 No preemption victims found for incoming pod.
As the docs say, all we natively support is the runtimes. If you want things that the operator adds, including GPU resources in the node status, you need to install the plugin.
This may be a mistake on my side but I've set up a cluster with only one master node. I've installed the NVIDIA drivers as well as the container runtime. I've verified that containers can access GPU both through docker as well as
containerd
.Environmental Info:
K3s Version:
Node(s) CPU architecture, OS, and Version:
Cluster Configuration:
Just one master node:
Describe the bug/Steps To Reproduce
nvidia-smi
throuhg host machine, and containers through both docker & containerdkubectl describe node <my node>
doesn't show the gpu:Running
nvidia-smi
fromcontainerd
Running
nvidia-smi
from dockerExpected behavior:
GPU should be automatically detected.
Actual behavior:
It doesn't get detected.
Additional context / logs:
Per the website's instructions, I ran this;
I can see that
nvidia
&nvidia-experimental
runtime classes are visible:I applied the pod specified on the website itself, and upon inspecting it:
Full output of
kubectl node describe
The text was updated successfully, but these errors were encountered: