This message was deleted.
# rke2
a
This message was deleted.
c
you should be able to use a containerd configuration template to add that to the containerd bundled with rke2
I don’t see anything about how to use this with Kubernetes though, so I’m not sure what you’d do with it if you enabled it.
b
gotta be something
that whole project is..well k8s, and the most recent release is all about cdi
c
looks like the nvidia device plugin only added cdi support in 0.14
b
I'm just getting ready to blow this out more fully and want to use what's current to not need to go through some ugly upgrade path across a bunch of clusters in the near future 😄
c
which came out like, just under 2 months ago
so it’s a relatively recent thing
b
agreed
using cdi seems like the path forward so if you'd like to explore it jointly I'm happy to do so
c
rke2 will autodetect the nvidia runtimes if they are installed when it starts up, I guess we could autodetect CDI too if there is a standard way to check for it
b
yeah, looks like it might be a new binary
c
and if there is anything else in the ecosystem that actually uses it
b
speaking of upgrades, we're almost through all the rke1 -> rke2 transition 😄
been an adventure for sure
c
hopefully rke2 is more enjoyable to admin than rke 🙂
b
they ecosystem has matured a lot since the rke1 days, my experience with 1 was fine generally