Most likely you’d need a privileged container as unprivileged containers have all capabilities enabled from the start, but constrainted to their user namespace.
1 Like
Ok, thanks for the steer. I’ll go with privileged or live without.
That should save me a lot of failed attempts!
It’s pretty bad from a security point of view, so not something we’d really recommend.
FWIW, I’ve just been ignoring the errors from the GPU monitoring stuff in my environment. It provides slightly less data in the dashboard but doesn’t impact actual usage of Frigate.
1 Like
thanks Stéphane. I agree on the security front and from what I can tell it’s only cosmetic for frigate.
The main reason GPU stats are “nice to have” in container is simply confirmation that gpu hardware accelleration is working correctly. I have no need to monitor for day to day use.
For now I’m running unprivileged and doing without. Everything seems to be working well.
Thanks for all the support and the work with getting gpu firmware up and running.
Yeah, an alternative could be us adding a incus admin os system gpu show or something that could show some of those stats.
1 Like