Data backup with btrfs and LVM bash scripts

There have already been a lot of posts about backups, especially a lot for Linux. I also took care of setting up a backup.

It was necessary to create system backups, data from the mounted partition and LVM volumes (virtual machine disks). There were thoughts to use Bacula, because familiar with it, but since at home only 1 computer client-server architecture would only create additional difficulties when restoring in case of damage to the system. It means that we simply copy the system and data, we create the image of the LVM partition using dd. I wanted to make a backup every day (at least data) and store at least 14 days. But the search for ready-made and simple solutions that satisfy all needs were not crowned with success. So take the bash and write your bike. In this article, I share what happened.



Disclaimer: when writing scripts there was no goal to write monsters that do everything. I needed a simple and reliable way to backup. I would be grateful for pointing out inaccuracies and bottlenecks of scripts (those where errors may occur). The article is intended more for newbies in Linux, who are looking for a ready-made solution and for lazy people, :) who are too lazy to write themselves.

Initial conditions:

• OS: Arch Linux
• Root file system (/): btrfs, you need to copy all files.
• Data partition (/ mnt / data /): btrfs.
• LVM volumes (/ dev / virt_image_array / *).
• Section for backups (/ mnt / backup /: etx4, backups will be added here.
• Required utilities (except those included in the base distr): rsync, btrfs-progs (for managing btrfs).

It was decided once a week to make a full copy of everything and create snapshots of btrfs sections every day. You can also create snapshots of LVM volumes, but for me the loss of data for the week is not critical, so weekly copies will suffice.

')

So, script number 1, creates a copy of the root partition files in / mnt / backup / root / "day number" /.







Script No. 2, creates root files snapshots (the script is very similar in logic to the 1st, so I will only comment on the differences). Snapshot is named auto_ "number of the day."





Script No. 3, creates a copy of the LVM volume:







Since the scripts backing up data from / mnt / data are similar to scripts 1 and 2 I think there is no need to write them.



Add to crontab and determine how often to create backups (in my example, backup is created once a week, snapshots once a day).

 20 01 * * 1 /usr/bin/backup_root.sh >> /var/log/backup_root.log 2>&1 50 01 * * 1 /usr/bin/backup_lvm1.sh >> /var/log/backup_lvm.log 2>&1 20 01 * * * /usr/bin/snapshot_root.sh >> /var/log/snapshot_root.log 2>&1 

What else can be screwed:

• If you need to create more backups, but there is no space, then you can install a file system with deduplication (for example, Opendedup), but the reliability of data storage will decrease.
• The storage may be a folder connected via NFS or sshfs.
• If you need to create backups or snapshots more than once a day, you can count not days since 1970, but hours (a crutch, it will be even more incomprehensible for what date the backup).

UPD. According to the advice of onix74, the scripts were corrected.