This page is about a specific proposal for automated backups in the future.
For general backup and restore procedures, see BackupAndRecovery.
Proposal for automated Backups
The goal is to provide an easy to use backup and restore solution by default for Servers and Desktops.
Minimal Configuration needed by the user - unless the user wants to change to "expert-mode". Also it should provide easy restore procedure.
Discussion notes
Discussion notes from the Bof of the DebConf 2015 talk "A Vision of Backups in Debian" by Lars Wirzenius [1],[2]
- MVP
- Backup:
file based first
to a server, not USB drive (to avoid having to interact with user about when drive can be unplugged)
- system level backups, run as root from cron, not per-user
Documentation (translation later) - make sure we document everything that is important
Metadata (Readme file, plaintext, machinereadle) should indicate with version of the tool has been used - that later restore is guaranteed
- directory for hooks (pre-/post-), empty for now, but there in case the user wants to run some scripts before/after the backup
- Restore:
- via documented commands
- Backup:
- Later Versions may include:
- provide a set of default profiles for some usecases (e.g. "Desktop User - Backup to Harddrive", "Destop User - my data only", "Mailserver", ...)
- support different storage options (usb, network, ...)
- provide support for databases,
- maybe use hooks for "dumping datatbases": The scripts in the /etc/backup/hooks/ (this the the right place?) directory are used to dump the contents of a database into some file(s) that can be backed up properly (as proposed by Elrond on irc)
More ideas from Lars missing from the above
The story from Lars is really interesting, and also brought up features that are not detailed above, namely:
continuous backups: backup files when they change, not at specific "times" (e.g. "night") when the machine may be offline or the disks unplugged. backup stuff when the backup media is available, basically. the borg community discussed possible approaches to continuous backups in their bugtracker
- automatic detection and configuration of external drive: pop in a backup drive, if it's empty, ask if it should be a backup drive and so on
- GUI: with notification area for start/stop of backups ("it is now safe to remove your USB drive") and restore interface (probably just help the user mount the drive and drag-and-drop for restores)
- backup encryption through asymetric crypto, so the decryption key can be stored offline
- allow restore during operating system install (debian-installer changes!)
- automatic cache exclusions (provided they follow the CACHEDIR.TAG standard)
- efficient large file support
Also keep in mind that there are a *lot* of backup tools in Debian (see BackupAndRecovery) and it would be difficult if not fallacious to try to impose one solution among the many. The best would be for the GUI described above to be a wrapper around existing backup solutions, which could have a default but would allow the user to choose which tool they want based on certain criterias (symetric/asymetric encryption, speed, project maturity, etc).
In other words, maybe what we need here is yet *another* backup software that would only be a wrapper and GUI for the regular user. A bit like what backupninja is doing, but with a GUI and extensible (maybe written in Python instead of bash, for example).
Help
You are welcome to join:
- IRC: #debian-backup on irc.oftc.net
- Wiki: this page
- Mailinglist: not until there's enough documented need for it
People
The following people are helping make this happen:
other resources
https://wiki.debian.org/BackupAndRecovery - List of avaiable backup solutions
https://wiki.debian.org/EvolutionBackup - Backup of Evolution user data
[1] https://summit.debconf.org/debconf15/meeting/180/a-vision-of-backups-in-debian/
backupninja (deb) has dump scripts for various databases and other programs - maybe they can be reused?
CategorySystemAdministration | CategoryRedundant: merge with other ?CategoryBackup pages