This chapter describes the process of backing up a network. Backup is only half the story, though; in order to do any good, a backup must allow you to restore data. There are two types of restore that you must consider:
Full restores are difficult to handle because you must find some way to get the restore onto a computer that has no working software. One common approach to solving this problem is to prepare an emergency restore system on a floppy disk, bootable CD-ROM, bootable Zip disk, or the like. For network backup clients, this disk should include network configuration tools and whatever network backup clients or servers you used to create the backup in the first place.
Sometimes, you may want to use a different method for a full restore than you used for a backup. For instance, a client-initiated Samba backup using a backup share might be more easily restored by using either a client-initiated direct tar restore via rshd or a server-initiated restore using either NFS or Samba.
In some cases, and especially if you haven't adequately planned an emergency restore procedure, you may have to reinstall the base OS in order to restore normal backup client functionality. You can then use this configuration to restore the rest of the data to the system, as if it were a partial restore.
AMANDA is unusual among the tools described here in that it includes client-based restore tools. The most powerful of these is amrecover , which in turn calls other tools such as amrestore . When you type amrecover as root on the backup client to enter the recovery utility, the program presents you with its own prompt at which you can type commands like setdate (to set the date of a backup from which you want to recover), cd (to change into a directory in the backup), add (to add a file to a restore set), and extract (to restore files from a restore set). After you type extract , amrecover prompts you for the appropriate tapes to restore data.
No matter what methods you use for restoration, it's critical that you test them before they become necessary. Ideally, you should set up a test system and try backing it up and performing both partial and full restores on that test system. If your network hosts several different OSs, repeat these tests with each of the OSs. Fully document your restoration procedures, and periodically retest them, particularly if your network changes in any important way. Keep backups of your emergency restore disks ”you don't want to be surprised by a bad floppy, which Murphy's Law guarantees you'd discover at exactly the wrong time.