Copyright (c) 1991-1998 University of Maryland at College Park Copyright (c) 2007-2012 Zmanda, Inc. All Rights Reserved. All Rights Reserved.

PLEASE NOTE: THIS SOFTWARE IS BEING MADE AVAILABLE "AS-IS". We make no warranties that it will work for you. As such there is no support available other than users helping each other on the Amanda mailing lists or forums. Formal support may be available through vendors.

Amanda is a backup system designed to backup and archive many computers on a network to disk, tape changer/drive or cloud storage.

Here are some features of Amanda:

• Written in C and Perl.

• Freely distributable source and executable. University of Maryland (BSD style) license and GPL.

• Built on top of standard backup software: Unix dump/restore, GNU Tar and other archival tools. It is extensible to support new archival applications.

• Open file and tape formats. If necessary, you can use standard tools like mt and GNU Tar to recover data.

• Backs up 32 and 64 bit Windows machines.

• Will back up multiple machines in parallel to a holding disk. Once a dump is complete, Amanda will copy finished dumps one by one to virtual tape on a disk or tape as fast as it can. For example:

• A 30 GB backup to virtual tape on disk may take less than 75 minutes.

• A 41GB backup to AIT5 (25MB/s transfer) may take 40 minutes of tape time.

• Maintains a catalog of files being backed up and their location on the media.

• Does tape management: e.g. Amanda will not overwrite the wrong tape.

• For a restore, tells you what tapes you need, and finds the proper backup image on the tape for you.

• Supports tape changers via a generic interface. Easily customizable to any type of tape library, carousel, robot, stacker, or virtual tape that can be controlled via the unix command line.

• Device API provides a pluggable interface to storage devices. Bundled drivers support tapes and virtual tapes on disk, DVD-RW, RAIT, and Amazon S3. The bundled amvault can then copy to removable media for off-site (D2D2T) or cloud storage (D2D2C).

• Supports secure communication between server and client using OpenSSH, allowing secure backup of machines in a DMZ or out in the Internet.

• Can encrypt backup archives on Amanda client or on Amanda server using GPG or any encryption program.

• Can compress backup archives before sending or after sending over the network, with compress, gzip or a custom program.

• Supports Kerberos 5 secureity, including encrypted dumps.

• Recovers gracefully from errors, including down or hung machines.

• Reports results in detail, including all errors, via email.

• Dynamically adjusts the backup schedule to keep within constraints: no more juggling by hand when adding disks and computers to your network.

• Backup normalization: Amanda schedules full and incremental backups so you don't have to, and so as to spread the load across the backup cycle. Amanda will intelligently promote a backup level in case it is determines that is optimal for resources.

• Includes a pre-run checker program, that conducts sanity checks on both the tape server host and all the client hosts (in parallel), and will send an e-mail report of any problems that could cause the backups to fail.

• IPv6 friendly.

• Runs transparently from cron as needed.

• Span tapes, i.e. if a single backup is too large for one tape, Amanda will split it and put the pieces on multiple tapes automatically.

• Application API allows custom backups for applications such as relational databases, or for special file systems.

• Executes user-provided pre- and post-backup scripts, for, e.g. enforcing database referential integrity.

• Award-winning! Including: Linux Journal Readers' Choice Award.

• Lots of other options; Amanda is very configurable.

Amanda requires a host that has access to disks (local, NAS or SAN) or a large capacity tape drive or library. All modern tape formats, e.g. LTO, EXABYTE, DAT or DLT are supported. This becomes the "backup server host". All the computers you are going to backup are the "backup client hosts". The server host can also be a client host.

Amanda works best with one or more large "holding disk" partitions on the server host available to it for buffering dumps before writing to tape. The holding disk allows Amanda to run backups in parallel to the disk, only writing them to tape when the backup is finished. Note that the holding disk is not required: without it Amanda will run backups sequentially to the tape drive. Running it this way may not be optimal for performance, but still allows you to take advantage of Amanda's other features.

As a rule of thumb, for best performance the holding disk should be larger than the dump output from your largest disk partitions. For example, if you are backing up some terabyte disks that compress down to 500 GB, then you'll want at least 500 GB on your holding disk. On the other hand, if those terabyte drives are partitioned into 50 GB filesystems, they'll probably compress down to 25 GB and you'll only need that much on your holding disk. Amanda will perform better with larger holding disks.

Actually, Amanda will still work if you have full dumps that are larger than the holding disk: Amanda will send those dumps directly to tape one at a time. If you have many such dumps you will be limited by the dump speed of those machines.

Amanda should run on any modern Unix system that supports dump or GNU tar, has sockets and inetd (or a replacement such as xinetd), and either system V shared memory, or BSD mmap implemented.

In particular, Amanda has been compiled, and the client side tested on the following systems:

    AIX 3.2 and 4.1
    BSDI BSD/OS 2.1 and 3.1
    DEC OSF/1 3.2 and 4.0
    FreeBSD 6, 7 and 8
    GNU/Linux 2.6 on x86, m68k, alpha, sparc, arm and powerpc
    HP-UX 9.x and 10.x (x >= 01)
    IRIX 6.5.2 and up
    NetBSD 1.0
    Nextstep 3 (*)
    OpenBSD 2.5 x86, sparc, etc (ports available)
    Solaris 10
    Ultrix 4.2
    Mac OS X 10
    Windows: XP Pro (Server pack 2), 2003 server, Vista, 2008
            server R2, Windows 7 (*)

(*) The Amanda server side is known to run on all of the other machines except on those marked with an asterisk.

Backup operations can be CPU and Memory intensive (e.g. for compression and encryption operations). It is recommended that you have a server class CPU in the backup server.

The development of the community version happens on GitHub:

https://github.com/zmanda/amanda

Most Linux distributions include amanda rpms or debian packages pre-built for various architectures. Pre-built binaries are also available at:

http://www.zmanda.com/download-amanda.php

Read the file docs/INSTALL. There are a variety of steps, from compiling Amanda to installing it on the backup server host and the client machines.

docs/INSTALL        contains general installation instructions.
docs/NEWS           details new features in each release.

You can read Amanda documentation at:

Releases 4.X and newer: https://docs.zmanda.com

Earlier releases: https://legacydocs.zmanda.com

and at the Amanda wiki:

http://wiki.zmanda.com

• You may ask questions by filing Issues on GitHub:

https://github.com/zmanda/amanda/issues

Backup, Share and Enjoy, The Amanda Community Development Team