Backup strategy
From Sysadmin
In designing a backup strategy an organisation will establish what will be backed up, when and how it will be done.
Backup strategies are broadly divided in to three categories:
- Backups
- File recovery
- Disaster recovery
Backups are common (often nightly), file recovery is less common and disaster recovery is (hopefully) very rare.
What to Backup
Important data is often more wide spread that users and even sysadmins think. All too often organisations only back up their important data and only discover following a disaster that very important data was being held in storage areas that were not being backed up.
In general all data should be backed up. This includes both system and user data. Exceptions to this rule should be made very carefully, and only if there is fairly compelling evidence that they do not need to be backed up or should not be backed up. The onus of proof is on the position that data should not be backed up.
The application of quotas is very useful for force users to keep their data usage at reasonable levels. Even if disk space is plentiful this will keep usage down so that there is less data to back up.
System Data
Back up your system data. This includes /bin, /sbin as well as /etc. A lot of work goes into setting up systems correctly. If the config is lost then the sysadmin may be left reinventing the wheel to get the system back to its former state. Even with diligent maintenance documentation tends to lag behind actual system configuration. As a result simply reconfiguring based on the documentation may leave a system that is not fully usable. All of this can be resolved by the restoration of system files from backups.
Some sysadmins argue that system data should not be backed up but there are very good reasons to backup system data.
Restoration from a system backup ensures that the previously running system is fully restored. If the system is restored from installation media or from an automated build system then the system is different than it was before and may have unknown bugs or behavioral characteristics.
The ratio of user data to system data has been increasing for decades. 15 years ago the OS would take up as much as 80% of the available disk space. Today an OS installation typically takes about 2% of available disk.
There are edge cases where system backups may not be required. A large organisation that has thousands of servers and builds them from precanned config and patch levels may not need system backups since a system emerging from their build system will be correct by definition as it has gone through a rigerous testing process. Most organisations cannot assert such a rigorous build process and are better off maintaining backups of system data.
Personal Data
Backup all personal data. If it is considered valuable enough to be held as personal data then it is valuable enough to backup. Any data that is not worth backing up can be held in a temporary area such as /tmp on a Unix system.
Backup any data that cannot be easy recovered from the Internet. If data (such as an ISO9660 image of a Linux distribution) can be recovered easily from the Internet then it doesn't necessarily need to be backed up.
This page is a stub. Robert will add additional information to the page later.
