Backing up a living system

No Silver Bullets

Snapshots don't come free of disadvantages. The main disadvantage, as previously discussed, is that using snapshots requires careful planning. If you intend to use snapshots as part of your backup strategy for a folder, you need to store the contents of that folder on snapshot-capable storage to begin with.

Another issue is performance. LVM in particular is known for taking a performance hit when it must keep multiple snapshots taken from the same filesystem [3].

The biggest problem with LVM snapshots, however, is that their life is finite. LVM snapshots track the changes done to the original filesystem (see the box "How Do LVM Snapshots Work?"). If enough changes are done to the original filesystem, the LVM snapshot will run out of space to register the changes. Modern LVM supports dynamic expansion of the snapshot volume if it detects it is running out of room [4], but hard drive space is finite and the snapshot may yet be dropped if it needs to grow past the physical storage medium's capabilities.

How Do LVM Snapshots Work?

An LVM snapshot volume does not actually contain a copy of all the files as they were in the original filesystem when the snapshot was taken.

When a snapshot is taken, the volume manager takes note of the state of the original filesystem at the time of the snapshot. When a file is modified after the snapshot is taken, the old version of the file (the one you would expect to find in a snapshot) is moved to the snapshot volume area. This way, the snapshot volume only contains copies of data that has been modified since the snapshot point.

The snapshot volume can be mounted as a regular filesystem. If the user tries to access a file through the snapshot, LVM checks if the file has been modified since the snapshot point. If not, the file is retrieved from the original filesystem. If the file has been modified, the version from the snapshot area is retrieved instead.

As a result, a snapshot volume does not need to be as large as the original volume because it only contains the changes between the snapshot and the current date. Using a small snapshot volume, however, comes with risks: If the original is modified to the point that the changes are larger than the snapshot volume, the snapshot volume will be dropped and rendered broken. It is, therefore, important to allocate enough room for the snapshot volume.

Not every filesystem is suitable for use in conjunction with LVM snapshots. Filesystems must support freezing in order to guarantee that the resulting snapshots are in a consistent state when they are taken. XFS and ext4 are known to work well, and there are more options [2].

It is worth noticing that many applications with data you might want to back up don't guarantee the consistency of their files while the application is running. For example, a busy database might work asynchronously, keeping many operations in a RAM cache and committing them to disk periodically. Creating a backup that perfectly mirrors the filesystem the database is stored in will then create a copy with inconsistent data, because many database operations may not have been committed to disk. It is important to check the documentation of the programs you are backing up in order to learn of potential pitfalls.

Conclusions

Classic Unix tools are not advisable for making backups for files and folders that are being modified during the process, because backup corruption can occur as a result. This article has covered LVM as a safer alternative for Linux, but there are other tools for the task, including ZFS and BTRFS snapshots.

Many programs, especially services intended for production, suggest their own means for creating backups (Figure 4). Databases are notorious for including their own backup utilities, and if you are indeed using a real database (such as MariaDB or PostgreSQL) you should consider using their tools instead of backup utilities that operate at filesystem or block device level.

Figure 4: Production-ready programs offer instructions for backing up their data – and even provide their own backup functionality. OpenCart, a popular e-commerce platform, has a utility for copying its database on the fly.

Infos

  1. "Is Dump Really Deprecated?" by Antonios Christofides: https://dump.sourceforge.io/isdumpdeprecated.html#canusedump
  2. Freeze support committed into the kernel for a number of filesystems: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=c4be0c1dc4cdc37b175579be1460f15ac6495e9a
  3. "LVM Snapshot Performance" by John Leach: https://johnleach.co.uk/posts/2010/06/18/lvm-snapshot-performance/
  4. "[PATCH] automatic snapshot extension with dmeventd (BZ 427298)": https://listman.redhat.com/archives/lvm-devel/2010-October/msg00010.html

The Author

Rubén Llorente is a mechanical engineer who ensures that the IT security measures for a small clinic are both legally compliant and safe. In addition, he is an OpenBSD enthusiast and a weapons collector.

Buy this article as PDF

Express-Checkout as PDF
Price $2.95
(incl. VAT)

Buy Linux Magazine

SINGLE ISSUES
 
SUBSCRIPTIONS
 
TABLET & SMARTPHONE APPS
Get it on Google Play

US / Canada

Get it on Google Play

UK / Australia

Related content

  • Timeshift

    Timeshift lets users easily create, manage, and restore system snapshots.

  • On the Safe Side

    Data deduplication, encryption, compression, incremental backups, error correction, and support for snapshots and popular cloud storage services: Kopia delivers.

  • Back In Time

    Despite the importance of backups, many users still view the process as too complicated and too inconvenient. Back In Time makes the unloved backup less terrifying.

  • Snapper

    Thanks to Btrfs snapshots, Snapper lets you restore a system to a previous state in case of failure.

  • Duplicati

    The free backup tool Duplicati simplifies the process of backing up data with cloud providers while at the same time protecting backups with strong cryptography.

comments powered by Disqus
Subscribe to our Linux Newsletters
Find Linux and Open Source Jobs
Subscribe to our ADMIN Newsletters

Support Our Work

Linux Magazine content is made possible with support from readers like you. Please consider contributing when you’ve found an article to be beneficial.

Learn More

News