UNIX System Administration Handbook - Evi Nemeth [120]
Consult your manuals for an exact list of commands supported by mt.
10.8 AMANDA
The Advanced Maryland Automatic Network Disk Archiver, Amanda, is a sophisticated network backup system that can replace the home-grown scripts used by many sites. It can back up all the machines on a LAN to a single server’s tape drive. It supports most flavors of UNIX and many different kinds of backup media.
Amanda was originally written by James da Silva of the University of Maryland in 1991. It’s now supported by a team of sysadmins from around the world. The latest information and source code are available from www.amanda.org.
Amanda is not itself a backup program, but rather a wrapper that manages other backup software. Most sites will have Amanda use their systems’ native dump and restore commands as the underlying backup software, but Amanda can also drive gnutar and even Samba’s smbtar for those pesky NT machines.
Amanda supports a wide variety of tape drives, and it also takes advantage of jukeboxes and stackers. Amanda can use your tape drives’ hardware compression facilities, or it can compress backup images with compress or gzip on client machines before the data goes over the network.
Another great feature of Amanda is tape management. Amanda writes a header on each tape it uses and so will never overwrite the wrong tape. In addition, Amanda manages dump levels based on configuration parameters and the fullness of tapes. Amanda keeps records of which backups are on which tapes, and it can print out sticky labels showing the contents of each tape. (These labels become very useful when the disk that holds Amanda’s database crashes.)
Amanda is one of the most popular free backup solutions in common use. It boasts implementation at about 1,500 sites worldwide. It seems to scale well and is constantly being enhanced to support the latest and greatest backup devices.
The architecture of Amanda
In the Amanda model, tape drives and holding disks are attached to a central server. The server also hosts all the Amanda config files, log files, and databases. Amanda can only write one backup image to tape at a time, but it can spool multiple dumps to its holding disks simultaneously and then stream them out to tape.
Amanda supports multiple “configurations” on the same server machine. For example, one configuration might perform only level 0s of clients, while another does only incrementals. Each configuration generates its own log files and databases.
The Amanda server looks at its configuration files to determine which filesystems need to be backed up, which tape devices are available, and how much of the system’s resources (network bandwidth, tapes, CPU load, etc.) it is allowed to use. It then contacts client machines and asks them to estimate the size of their backup files. Using this information, Amanda schedules the backups.
The Amanda server is actually a collection of programs that implement the various parts of the system. It is best to run the server programs on a fast machine that is not generally busy. If you are backing up large amounts of data, the server should have the best connectivity your network architecture permits. Since the holding disk is used to spool dump images from the network to the tape drive, it should be at least as large as the largest partition you want to back up. Amanda also needs some disk space (< 75MB) for its own logs and databases.
As of this writing, the latest stable release of Amanda is 2.4.1p1; the examples, configuration files, and commands shown in this chapter are taken from that version.
Amanda setup
When you download Amanda, take the latest stable version, not the current development snapshot. We are talking about backups, so you want something stable!
After unpacking the source code, read the README, docs/SYSTEM.NOTES, and docs/INSTALL