[K3s on Jetson] Seeking Guidance on Deploying 2 Replicas of a YOLO Pod to Different Worker Nodes (Orin & AGX Xavier)

Robotics & Edge Computing Jetson & Embedded Systems Jetson AGX Xavier
1

I am working with a K3s cluster composed of a Xavier NX as the control-plane node, and an Orin NX and an AGX Xavier as worker nodes.

My objective is to deploy a CPU-based YOLO image recognition service (with the option to incorporate GPU support later) or other AI algorithms. To test the cluster’s reliability and parallel processing capabilities, this service needs to be deployed as two replicas (Pods), with one replica running on each of the two worker nodes.

Could you please let me know if there are any relevant examples or case studies for this? Alternatively, how should I proceed?

xavier-nx@xaviernx-desktop:~$ sudo kubectl get nodes -o wide
[sudo] password for xavier-nx:
NAME STATUS ROLES AGE VERSION INTERNAL-IP EXTERNAL-IP OS-IMAGE KERNEL-VERSION CONTAINER-RUNTIME
agx-xavier-work Ready 4d23h v1.29.15+k3s1 10.42.0.22 Ubuntu 20.04.6 LTS 5.10.216-tegra containerd://1.7.26-k3s1
orin-nx-worker Ready 4d23h v1.29.15+k3s1 10.42.0.33 Ubuntu 22.04.5 LTS 5.15.148-tegra containerd://1.7.26-k3s1
xavier-nx-master Ready control-plane,master 5d v1.29.15+k3s1 10.42.0.11 Ubuntu 20.04.6 LTS 5.10.216-tegra containerd://1.7.26-k3s1

3

Hi,

Thanks for reaching out.

We don’t have an official K3s tutorial for Xavier.
But if you search for it, there are some shares from the community that meet your requirements.

Thanks.