A file system snapshot contains metadata about and data blocks from an original volume that have changed since the snapshot was taken. When you access data via the snapshot, you see a point-in-time copy the original volume. There is no need to restore data from backup media or to overwrite the changed data.

During the snapshot’s lifetime, the snapshot must be mounted before its original volume can be mounted.

In a Xen host environment, the virtual machine must be using an LVM logical volume as its storage back-end, as opposed to using a virtual disk file.

Linux snapshots allow you to create a backup from a point-in-time view of the file system. The snapshot is created instantly and persists until you delete it. You can backup the file system from the snapshot while the volume itself continues to be available for users. The snapshot initially contains some metadata about the snapshot, but no actual data from the original volume. Snapshot uses copy-on-write technology to detect when data changes in an original data block. It copies the value it held when the snapshot was taken to a block in the snapshot volume, then allows the new data to be stored in the original block. As blocks change from their original value, the snapshot size grows.

When you are sizing the snapshot, consider how much data is expected to change on the original volume and how long you plan to keep the snapshot. The amount of space that you allocate for a snapshot volume can vary, depending on the size of the original volume, how long you plan to keep the snapshot, and the number of data blocks that are expected to change during the snapshot’s lifetime. The snapshot volume cannot be resized after it is created. As a guide, create a snapshot volume that is about 10% of the size of the original logical volume. If you anticipate that every block in the original volume will change at least one time before you delete the snapshot, then the snapshot volume should be at least as large as the original volume plus some additional space for metadata about the snapshot volume. Less space is required if the data changes infrequently or if the expected lifetime is sufficiently brief.

Beginning in SLES 11 SP3, an LVM logical volume snapshots can be thinly provisioned. Thin provisioning is assumed if you to create a snapshot without a specified size. The snapshot is created as a thin volume that uses space as needed from a thin pool. A thin snapshot volume has the same characteristics as any other thin volume. You can independently activate the volume, extend the volume, rename the volume, remove the volume, and even snapshot the volume.

To use thinly provisioned snapshots in a cluster, the original volume and its snapshots must be managed in a single cluster resource. This allows the volume and its snapshots to always be mounted exclusively on the same node.

When you are done with the snapshot, it is important to remove it from the system. A snapshot eventually fills up completely as data blocks change on the original volume. When the snapshot is full, it is disabled, which prevents you from remounting the original volume.

Remove snapshots in a last created, first deleted order.

The Logical Volume Manager (LVM) can be used for creating snapshots of your file system.

Snapshots can be useful if you need to roll back or restore data on a volume to a previous state. For example, you might need to revert data changes that resulted from an administrator error or a failed or undesirable package installation or upgrade.

You can use the lvconvert --merge command to revert the changes made to an LVM logical volume. The merging begins as follows:

After a merge begins, the merge continues automatically after server restarts until it is complete. A new snapshot cannot be created for the original volume while a merge is in progress.

While the merge is in progress, reads or writes to the original volume are transparently redirected to the snapshot that is being merged. This allows users to immediately view and access the data as it was when the snapshot was created. They do not need to wait for the merge to complete.

When the merge is complete, the original volume contains the same data as it did when the snapshot was taken, plus any data changes made after the merge began. The resulting logical volume has the original volume’s name, minor number, and UUID. The original volume is automatically remounted, and the snapshot volume is removed.