kidab

Ok reporting back that everything seems to work! My container is running Ubuntu 22 so I modified the rc.local to this: #!/bin/bash mount --make-rshared / At first it wasn't being applied. To solve that I followed an online tutorial to create a rc-local.service to actually run the contents of that file^. After a reboot the errors about shared mounts went away. However I did proceed to get the following error from the Nvidia gpu operator (formatting modified slightly for clarity): level=info msg="creating symlinks under /dev/char that correspond to NVIDIA character devices" time="2024-05-14T21:31:51Z" level=info msg="Error: error validating driver installation: error creating symlink creator: failed to load NVIDIA kernel modules: failed to load module nvidia: exit status 1; output=modprobe: FATAL: Module nvidia not found in directory /lib/modules/6.1.79-Unraid Failed to create symlinks under /dev/char that point to all possible NVIDIA character devices. The existence of these symlinks is required to address the following bug: https://github.com/NVIDIA/gpu-operator/issues/430\n\nThis bug impacts container runtimes configured with systemd cgroup management enabled. To disable the symlink creation, set the following envvar in ClusterPolicy: validator: driver: env: - name: DISABLE_DEV_CHAR_SYMLINK_CREATION value: "true" I followed the instructions in the error and update the helm install options. As of now, my node was properly tagged as having two Nvidia GPUs and all seems well! I have yet to schedule any actual workflows but am sure they will run. Thanks!!!

Awesome I will test this out and report back later. I believe the contents of /etc/r.local are going to be what solves my issues. Specifically the mounting of / with —make-rshared Everything else regarding k3s installation is pretty much exactly same for me. Thanks!

Not sure if this is the right place to ask but am at my wit's end. I've installed the plugin and am running all kinds of LXC containers. Everything works great! However im currently facing an issue trying to run a k3's cluster. I am trying to install the NVIDA GPU operator which performs all kinds of bind mounts to view and confirm what GPU resources are available. Installation fails however with several errors complaining that the root directory (and others) are not mounted in shared or slave mode. I've done all kinds of reading and have also tried adding "lxc.rootfs.options = rshared" to the lxc conf. No change. I've also tried "slave" and see neither option in "/proc/mounts" for the root directory. (shown below). No change if I use an XFS or btrfs formatted disk. /dev/md1p1 / xfs rw,noatime,nouuid,attr2,inode64,logbufs=8,logbsize=32k,noquota 0 0 Really hoping you could point me in the right direction, thanks!