Running GPU payload on Jetson device with K3s
Briefly

Installing K3s directly on the Jetson device simplifies GPU access, as the Docker layer complicates GPU passthrough and adds unnecessary overhead in managing containers.
By defining specific configuration options in a config file, such as changing kubeconfig permissions and setting a TLS SAN, I can manage the remote K3s cluster effectively.
K3s provides flexibility in its configuration, allowing for a custom template file to modify essential settings, in contrast to K3d where the Nvidia device plugin is mandatory for GPU utilization.
Read at Medium
[
]
[
|
]