Docker is a great containerization technology for running applications. It keeps multiple applications completely isolated from each other, only allowing connections exactly when you tell them to. But what if you’re on a hypervisor? You want your host OS to be as lean as possible (else it defeats the point),…
NFS is a great protocol for sharing files quickly and simply over the network. Whilst it’s not designed for end user use, it’s great for mounting directories from remote machines, and having them be performant. NFS’ lack of authentication is in a way a feature, honest. Not only does it remove the additional application and protocol overheads and complexities, but it also forces you to think properly. Either firewall it off, or run it over a VPN.
For my latest server, I intend to have the majority of my data storage handled by a
FreeNAS TrueNAS VM which handles ZFS, and simply expose the data to each VM / LXC container through NFS. TrueNAS has built-in support for functioning as an NFS server, and a web UI to manage shares.
Unfortunately however, if you try and mount an NFS share from an LXC container, you’ll receive some form of “Permission denied” error, even if you’re
nfs-client:~# mount x.x.x.x:/mnt/test-pool/test /mnt/nfs mount: permission denied (are you root?) nfs-client:~# whoami root
So, what gives?
LXC is a container technology. Whilst it’s similar to Docker, it’s also incredibly different and fills a different need. For example, it’s very common to run Docker inside an LXC container, but doing the inverse or attempting to run Docker inside docker (which is possible) may get you some weird looks.
LXCs isolation comes in part from apparmor, and it’s apparmor which is preventing LXC from mounting NFS shares. On Linux, NFS is implemented in the kernel. There are a few reasonably popular and maintained userspace implementations, such as
nfs-ganesha, although they don’t have much traction (and I couldn’t make them work).
#What is “privileged” mode?
LXC has the concept of privileged mode, which defaults to false. By default, the user IDs of LXC containers are different to those on the host. So processes running as
root (id 0) in LXC won’t be id 0 to the host. This provides an additional layer of security on the off chance an attacker is able to break out of the LXC container isolation, however unlikely that may be.
Anyone familiar with Docker will note that docker doesn’t work like this: The id inside the container is the same as on the outside. This makes volume maps much easier to work with, but makes it more important to not run docker processes as root.
Personally, I think this isn’t the best name for the setting. It implies more of LXCs security issues are removed than actually are. I still wouldn’t recommend running privileged unless you absolutely have to, but it doesn’t remove the isolation quite to the level Docker’s
privileged: true does.
To support NFS under LXC, some of the apparmor protections need to be lifted. Looking online, there are a number of guides which claim to make the required modifications for LXC to mount NFS. I tried every single one, and nothing appeared to work correctly. That said I’ve not worked much with apparmor. They all hinged around creating a new profile for LXC, and forcing the container to use it. If you know the secret ingredient to make this work for unprivileged containers, please tell me! There are however several ways to make it work for privileged containers, which like I say isn’t the end of the world, and is needed for some of my use cases anyway.
One that’s done, restart the container, and you should be able to mount your NFS shares as normal. To ensure they’re mounted correctly on boot, configure your mounts in
192.168.1.1:/data /mnt/data nfs defaults 0 0
Or the quick and dirty way using the command line:
$ mount -t nfs 192.168.1.1:/data /mnt/data
How do you mount NFS shares inside an LXC container?
- Create a privileged LXC container, using any guest distribution of your choosing
- Once created, modify the config file (
/etc/pve/lxc/<id>.confon Proxmox) and add
- Restart the container
- Mount your data (e.g.
mount -t nfs 192.168.1.1:/data /mnt/data)
Share this page
View all →
You should back up your data, properly! If you’re not, you’re playing a dangerous game with fate. Computers are pretty reliable, but they also go wrong, often. You should always backup your files, but backing up a containerized application isn’t quite as simple. A container is 3 things:ConfigurationVolumesNetworking The point…
GPU Passthrough has become a great way to run a Linux host, but still run games under Windows. By having 2 GPUs, 1 for the Linux host and the other for a Windows VM, you give Windows its own full GPU for games, but without having to run Windows as…