Lvm – Why is zfs/btrfs recommended for an lxd host

btrfslvmlxcperformancezfs

Note that for maximum performance, it's recommended to have
/var/lib/lxd on a btrfs filesystem or setup ZFS or LVM and configure
LXD to use a ZFS pool or LVM VG.
Source

Why is that? I.e. How will those file systems (on the host) equate to better container performance?

Is it computational performance, is it just reduced hard drive consumption because of copy-on-write, both, neither, or something else?

Edit 1

Thanks to @Sven for pointing out that "LVM VG" refers to Logical Volume Manager Volume Group.

Best Answer

In addition to what @Sven said; ZFS, btrfs and LVM all provide copy-on-write clone/snapshot features. This makes it very cheap storage and time-wise to spin up new containers. With an image stored on regular ext2-4 filesystem, LXD will have to copy all the data itself, which takes more time and storage.

Related Solutions

How to create a LXC container on LVM from an OpenVZ template

I do this all the time using LXC and the OpenVZ CentOS templates.

The two guides I followed initially are here and here.

I usually place my containers under /srv/ or /srv/lxc/. I don't bother with LVM, but if you want to, just mount your new LVM volume under /srv/lxc/container_name per-container. That makes sense, right?

In this example, I have a ZFS-on-Linux mount at the place where I want the container to reside under /srv:

[root@Lancaster_Mirror1 ~]# df -h
Filesystem            Size  Used Avail Use% Mounted on
/dev/cciss/c0d0p2      12G  1.9G  9.4G  17% /
tmpfs                 7.8G     0  7.8G   0% /dev/shm
/dev/cciss/c0d0p1     291M   59M  218M  22% /boot
/dev/cciss/c0d0p7     2.0G  119M  1.8G   7% /tmp
/dev/cciss/c0d0p3     9.9G  1.9G  7.5G  20% /usr
/dev/cciss/c0d0p6     6.0G  321M  5.3G   6% /var
vol2/images            98G   43G   56G  44% /images
vol3/Lancaster_Test    98G   22G   77G  22% /srv/Lancaster_Test <<--container

As for the LXC installation, I take the prebuilt OpenVZ image and put it in the destination container directory, then unpack it - tar -ztvf centos-6-x86_64-devel.tar.gz. Nothing needs to be modified on the template to make this work.

  426  wget http://downloads.sourceforge.net/project/lxc/lxc/lxc-0.7.3/lxc-0.7.3.tar.gz
  427  rpmbuild -ta lxc-0.7.3.tar.gz
  429  rpmbuild -ta lxc-0.7.3.tar.gz
  434  yum --nogpg install lxc-0.7.3-1.x86_64.rpm libvirt
  437  lxc-create -f /etc/lxc/Lancaster_Test.conf -n Lancaster_Test
  438  lxc-checkconfig
  441  screen -dmS init-Lancaster_Test /usr/bin/lxc-start -n Lancaster_Test
  442  screen -dmS console-Lancaster_Test /usr/bin/lxc-console -n Lancaster_Test

Maybe I'm off, but I choose to use the containers directly on a filesystem. Are you doing something different?

I can provide excerpts of the lxc config files, but you don't seem to have an issue with that.

LVM snapshot of btrfs volume changes mounted device

I have not looked massively into this but btrfs as a filesystem works on groups of disks, not individual devices.

I suspect that there is no way for btrfs to distinguish between the mounted snapshot and the real mounted filesystem when a mount occurs. It may as a matter of fact see the UUID of the underlying subvolume, assume its a mirror of the original volume and write to both volumes at the same time.

I would be surprised if this ever gets fixed seeing as for most intents and purposes btrfs snapshots supersede LVM snapshots.

Best Answer

Related Solutions

How to create a LXC container on LVM from an OpenVZ template

LVM snapshot of btrfs volume changes mounted device

Related Topic