backup

Manage Backups on a Hetzner Storagebox

Index

• Usage
• Background

Install

You can install cloudbackup by checking out the repository and installing the requirements. The prefered mode of use however is via docker container.

• examples/local-backup - example script using a checked out repository
• examples/docker-backup - example script using the docker container
• examples/kubernetes - example manifests for use in a kubernetes cluster
  • examples/kubernetes/metrics - Example deploy of prometheus metrics
  • examples/kubernetes/mysql - Example backup cronjobs for mysql backups
  • examples/metrics/files - Example backup cornjob for file backups

Concept

Setting up backups has always been a pain even when you did it once per server and everything on it magically got backed up by the existing scripts.

In a cloud environment you want as little single points of failure as possible - so generally each app has its own mysql database server and files. However that means you also need to reinvent the backup wheel for every service and generally at least twice - one for the general files of an app and one for its database system.

This script does not try to replace existing backup solutions but instead enrich them:

• abstract the storage interface so it can be used against sftp or s3(reference implementation is for the hetzner storage box)
• Provide access to all backups from a single place
• make it easy to add another repository for another app

Used backup software

Mariabackup

Mariabackup is a fork of perconas xtrabackup patched to work with mariadb. It uses a genius combination of copying the actual files from the filesystem and reading the incoming changes to them through the binary log to create lightning fast backups - as long as you're using the InnoDB storage engine.

Borgbackup

Borgbackup is a deduplicating backup solution. This means it will cut your files into chunks, save the chunks and calculate their hashes.
This has the benefit of a traditional incremental file backup that only files are actually written again which have changed but improves it by the fact that even when large files change only the chunks which actually changed are written again.