In the context of self-hosting, backups are an important element to compensate for unexpected events (fire, database corruption, loss of access to the server, compromised server...). The backup policy to implement depends on the importance of the services and data you manage. For example, backing up a test server will be of little interest, while you will want to be very careful if you are managing critical data for an association or a company - and in such cases, you will want to store the backups in a different location or locations.
A good backup consists of at least 3 copies of the data (including the original data), on at least 2 separate storages, in at least 2 separate locations (far enough apart) and ideally with 2 separate methods. If your backups are encrypted these rules also apply to the decryption phrase/key.
A good backup is also in many cases, a recent backup, so it takes either a lot of rigor or to automate the process.
A good backup is checked regularly to ensure the effectiveness and integrity of the data.
Finally, a good backup is one that is restorable within an acceptable timeframe for you. Remember to document your restoration method and to estimate the transfer time of a copy, especially if the Internet connections involved are not symmetrical.
Example of a robust and comfortable combination:
Below, a list of risks sorted from the most to the least probable, whose probability remains to be adapted according to your situation (location of the server, quality of the installations, user profiles, etc.). It is up to you to put the cursor where it should be, especially considering the consequences of a data loss.
Keep in mind that real accidents are linked to the occurrence of 2 events simultaneously.
A method that allows a partial backup is to backup files and emails via synchronization software like Nextcloud client or ThunderBird. This way, you avoid the risk of hardware failure.
If this method is easy to set up, it is not without risk because of the synchronization itself. For example, if you are on Windows or Mac, you increase the risk of data loss following the encryption of files by a cryptolocker type virus. On any type of system, a false manipulation can delete all your copies on the server and on the equipment that synchronizes. This concern is aggravated by the fact that the deletion synchronization is usually rather instantaneous.
While the risk of false manipulation can be mitigated by desktop backup software such as TimeShift, the risk of false manipulation can only be mitigated by a backup on a hard drive. Only a backup on a disconnected external hard drive really protects you from ransomware.
Found errors? Think you can improve this documentation? Simply click the Edit link at the top of the page, and then the icon on Github to suggest changes.