LVM block based snapshotter

This is a containerd snapshotter plugin which stores snapshots in xfs or an user preferred based filesystem format in a LVM thin pool creating sparse images.

Requirements

LVM snapshotter requires lvm2 set of tools which also support thin volume provisioning to be installed on the system. On ubuntu, it can be installed using apt install lvm2 thin-provisioning-tools command. On fedora/centos, it can be installed using yum install lvm2 command.

Setup

LVM snapshotter by default will format snapshots as xfs and depends on LVM thin provisioning tools. Make sure that the necessary components, lvm, thin-provisioning-tools(in case of Ubuntu) and mkfs.xfs are installed.

To make this snapshotter work you will need to prepare a volume-group and a thin-volume-pool prior to start containerd and update the configuration file, Eg:

[plugins]
  ...
  [plugins.lvm]
    vol_group = "vgcontainerd"
    thin_pool = "lvthincontainerd"
  ...

These commands are examples on how to create a volume group and a LVM thin pool. We will assume the disk is /dev/sdc

vgcreate vgcontainerd /dev/sdc
lvcreate --thinpool lvthincontainerd -l 90%FREE vgcontainerd

The following configuration options are honored:

• vol_group - a Logical volume group created using lvm tools. This is a mandatory argument.
• thin_pool - a Logical thin pool created using lvm tools. This is a mandatory argument.
• root_path - a directory where the metadata holding volume will be mounted (if empty, default location of containerd plugin will be used. If that does not exist /mnt is the final fallback).
• img_size - size of the thin image created within the thin pool (if empty, default size if 10G)
• fs_type - filesystem type to format the image with (If empty, xfs filesystem will be used).

Run

You can use this snapshotter with the below commands:

ctr images pull --snapshotter lvm docker.io/library/hello-world:latest
ctr run --snapshotter lvm docker.io/library/hello-world:latest