amanda - Advanced Maryland Automatic Network Disk Archiver
SYNOPSIS
amdump config
amflush [ -f ] config
amcleanup config
amrecover [ [ -C ] config ] [ options ]
amrestore [ options ] tapedevice [ hostname [ diskname ]]
amlabel config label [ slot slot ]
amcheck [ options ] config
amadmin config command [ options ]
amtape config command [ options ]
amverify config
amrmtape [ options ] config label
amstatus config [ options ]
DESCRIPTION
Amanda is the "Advanced Maryland Automatic Network Disk
Archiver". This manual page gives an overview of the
Amanda commands and configuration files for quick refer-
ence.
Here are all the Amanda commands. Each one has its own
manual page. See them for all the gritty details.
amdump Take care of automatic Amanda backups. It is nor-
mally executed by cron on a computer called the
tape server host and requests backups of file sys-
tems located on backup clients. Amdump backs up
all disks in the disklist file (discussed below) to
tape or, if there is a problem, to a special hold-
ing disk. After all backups are done, amdump sends
mail reporting failures and successes.
amflush
Flush backups from the holding disk to tape.
Amflush is used after amdump has reported it could
not write backups to tape for some reason. When
this happens, backups stay in the holding disk.
After the tape problem is corrected, run amflush to
write backups from the holding disk to tape.
amcleanup
Clean up after an interrupted amdump. This command
is only needed if amdump was unable to complete for
some reason, usually because the tape server host
crashed while amdump was running.
amrecover
Provides an interactive interface to browse the
Amanda index files and select which tapes to
recover files from. Can also run amrestore and the
Read an Amanda tape, searching for requested back-
ups. Amrestore is suitable for everything from
interactive restores of single files to a full
restore of all partitions on a failed disk.
amlabel
Write an Amanda format label onto a tape. All
Amanda tapes must be labeled with amlabel. Amdump
and amflush will not write to an unlabeled tape
(see TAPE MANAGEMENT below).
amcheck
Verify the correct tape is in the tape drive and
all file systems on all backup client systems are
ready to be backed up. Can optionally be run by
cron before amdump so someone will get mail warning
that backups will fail unless corrective action is
taken.
amadmin
Take care of administrative tasks, like finding out
which tapes are needed to restore a filesystem,
forcing hosts to do full backups of selected disks,
and looking at schedule balance information.
amtape Take care of tape changer control operations, like
loading particular tapes, ejecting tapes, and scan-
ning the tape rack.
amverify
Check Amanda backup tapes for errors (GNU tar for-
mat backups only).
amrmtape
Delete a tape from the tapelist and Amanda
database.
amstatus
Give the status of a running amdump.
CONFIGURATION
There are three user-editable files that control the
behavior of Amanda. The first is amanda.conf, the main
configuration file. It contains parameters to customize
Amanda for your site. Second is the disklist file, which
lists hosts and disk partitions to back up. Third is the
tapelist file, which lists tapes that are currently
active. These files are described in more detail in the
following sections.
All files are stored in a config directory under
/usr/amanda/etc/amanda. A site will often have more than
configuration for infrequent full archival backups. The
configuration files would be stored under
/usr/amanda/etc/amanda/normal/ and
/usr/amanda/etc/amanda/archive/, respectively.
All log and database files generated by Amanda go in cor-
responding directories somewhere. An example location is
under /usr/adm/amanda. In our example, the files would go
in /usr/adm/amanda/normal and /usr/adm/amanda/archive.
CONFIG FILE PARAMETERS
There are a number of configuration parameters that con-
trol the behavior of the Amanda programs. All have
default values, so you need not specify the parameter in
your amanda.conf if the default is suitable.
Lines starting with # are ignored, as are blank lines.
Keywords are case insensitive, i.e. mailto and MailTo are
treated the same.
Integer arguments may have one of the following (case
insensitive) suffixes, some of which have a multiplier
effect:
b byte bytes
Some number of bytes.
g gb gbytes gigabytes
Some number of gigabytes
(bytes*1024*1024*1024).
k kb kbytes kilobytes
Some number of kilobytes (bytes*1024).
kps kbps
Some number of kilobytes per second
(bytes*1024).
m mb meg mbytes megabytes
Some number of megabytes (bytes*1024*1024).
mps mbps
Some number of megabytes per second
(bytes*1024*1024).
tape tapes
Some number of tapes.
day days
Some number of days.
week weeks
expected to mean an infinite amount.
Boolean arguments may have any of the values y, yes, t,
true or on to indicate a true state, or n, no, f, false or
off to indicate a false state. If no argument is given,
true is assumed.
org "string"
Default: DailySet1. A descriptive name for the
configuration. This string is used for mail
reports. Each Amanda configuration should have a
different value.
mailto "string"
Default: operators. A comma separated list of
recipients for mail reports.
dumpcycle int
Default: 10 days. The number of days in the backup
cycle. Each disk will get a full backup no less
often than this many days.
runspercycle int
Default: same as dumpcycle. The number of amdump
runs in dumpcycle days. A value of 0 means the
same value as dumpcycle. A value of -1 means guess
the number of runs from the tapelist file, which is
the number of tape used in the last dumpcycle days
/ runtapes.
tapecycle int
Default: 15 tapes. The number of tapes in your
active tape cycle. This must be at least one
larger than the number of Amanda runs done during a
dump cycle (see the dumpcycle parameter) times the
number of tapes used per run (see the runtapes
parameter).
For instance, if you have dumpcycle set to 14 days,
do one Amanda run per day and have runtapes set to
one, tapecycle must be at least 15 (14 days * one
run/day * one tape/run + one tape).
In practice you should have five or more extra
tapes to allow for schedule adjustments or disaster
recovery.
dumpuser "string"
Default: daemon. The login name Amanda uses to run
the backups. The backup client hosts must allow
access from the tape server host as this user via
tapedev "string"
Default: /dev/null. The path name of the tape
device. If you have configured a tape changer (see
the tpchanger option), this option is not used and
should be commented out of your configuration file.
rawtapedev "string"
Default: /dev/null. The path name of the raw tape
device. It is only used if Amanda is compiled for
Linux machines with floppy tapes. It is needed for
QIC volume table operations.
tpchanger "string"
Default: none. The name of the tape changer. If
you do not have a tape changer, this option is not
used and should be commented out of your configura-
tion file.
changerdev "string"
Default: /dev/null. A tape change configuration
parameter. Usage depends on the particular changer
defined with the tpchanger option.
changerfile "string"
Default: /usr/adm/amanda/log/changer-status. A
tape change configuration parameter. Usage depends
on the particular changer defined with the
tpchanger option.
runtapes int
Default: 1. The maximum number of tapes used in a
single run. If you do not have a tape changer,
this option is not used and should be commented out
of your configuration file.
labelstr "string"
Default: .*. The tape label constraint regular
expression. All tape labels generated and used by
this configuration must match the regular expres-
sion. If multiple configurations are run from the
same tape server host, it is helpful to constrain
their labels to different strings (for example,
"DAILY[0-9]*" vs. "ARCHIVE[0-9]*") to avoid over-
writing each other's tapes.
tapetype "string"
Default: EXABYTE. The type of tape drive associ-
ated with tapedev or tpchanger. This refers to one
of the defined tapetypes in the config file, which
specify various tape parameters, like the length,
filemark size, and speed of the tape media and
Default: 300 seconds. Amount of time per disk on a
given client that planner will wait to get the dump
size estimates. For instance, with the default of
300 seconds and four disks on client A, planner
will wait up to 20 minutes for that machine. A
negative value will be interpreted as a total
amount of time, instead of a per-disk value.
netusage int
Default: 300 Kbps. The maximum network bandwidth
allocated to Amanda, in Kbytes per second. See
also the interface section.
inparallel int
Default: 10. The maximum number of backups that
Amanda will attempt to run in parallel. Amanda
will stay within the constraints of network band-
width and holding disk space available, so it
doesn't hurt to set this number a bit high. Some
contention can occur with larger numbers of back-
ups, but this effect is relatively small on most
systems.
maxdumps int
Default: 1. The maximum number of backups from a
single host that Amanda will attempt to run in par-
allel. See also the dumptype section and inparal-
lel option.
bumpsize int
Default: 10 Mbytes. The minimum savings required
to trigger an automatic bump from one incremental
level to the next. If Amanda determines that the
next higher backup level will be this much smaller
than the current level, it will do the next level.
See also the bumpmult option.
bumpmult float
Default: 1.5. The bump multiplier. Amanda multi-
plies bumpsize by this factor for each level. This
prevents active filesystems from bumping too much
by making it harder to bump to the next level. For
example, with the default bumpsize and bumpmult set
to 2.0, the bump threshold will be 10 Mbytes for
level one, 20 Mbytes for level two, 40 Mbytes for
level three, and so on.
bumpdays int
Default: 2. To insure redundancy in the dumps,
Amanda will keep filesystems at the same incremen-
tal level for at least bumpdays days, even if the
bump threshold criteria are met.
Default: disklist. The file name for the disklist
file.
infofile "string"
Default: /usr/adm/amanda/curinfo. The file or
directory name for the historical information
database.
logdir "string"
Default: /usr/adm/amanda. The directory for the
amdump and log files.
indexdir "string"
Default /usr/adm/amanda/index. The directory where
index files are stored. Index files are only gen-
erated for filesystems whose dumptype has the index
option enabled.
tapelist "string"
Default: tapelist. The file name for the active
tapelist file.
reserve number
Default: 100(percent) The amount of holding-disk
space that should not be used for full backups if
no tape is available. By default, when there is no
tape to write to, degraded mode (incremental) back-
ups will be performed to the holding disk. If you
wish to also allow full backups in this case, you
may reduce the amount of holding disk space
reserved for incrementals.
holdingdisk
Define parameters for a holding disk. The syntax
is:
holdingdisk name {
holdingdisk-option holdingdisk-value
...
}
The options and values are:
comment "string"
Default: none. A comment string
describing this holding disk.
directory "disk"
Default: /dumps/amanda. The path to
this holding area.
Default: 10 Mbytes. Amount of space
that can be used in this holding
disk area. If the value is nega-
tive, Amanda will use all available
space minus that value.
chunksize int
Default: -1. Holding disk chunk
size. Dumps larger than the speci-
fied size will be stored in multiple
holding disk files. The size of
each chunk will not exceed the spec-
ified value. However, even though
dump images are split in the holding
disk, they are concatenated as they
are written to tape, so each dump
image still corresponds to a single
continuous tape section.
If you specify 0, Amanda will create
holding disk chunks as large as
((INT_MAX/1024)-64) Kbytes.
A negative value causes filesystems
larger than the absolute value to be
dumped directly to tape, without
using the holding disk. The default
value, -1, is interpreted as
-((INT_MAX/1024-64) Kbytes. which
causes Amanda to dump filesystems
larger than slightly less than 2
Gbytes (assuming 32 bit integers)
directly to tape. The default is
provided mainly for backward compat-
ibility. You are encouraged to
specify a positive value smaller
than the maximum file size in the
holding disk filesystem.
Each holding disk chunk includes a
32 Kbyte header, so the minimum
chunk size is 64 Kbytes (but that
would be really silly).
Operating systems that are limited
to a maximum file size of 2 Gbytes
actually cannot handle files that
large. They must be at least one
byte less than 2 Gbytes. Since
Amanda works with 32 Kbyte blocks,
and to handle the final read at the
end of the chunk, the chunk size
file size, e.g. 2047 Mbytes.
DUMPTYPE SECTION
The config file may define multiple sets of backup options
and refer to them by name from the disklist file. For
instance, you might define one set of options for file
systems that can benefit from high compression, another
set that does not compress well, another set for file sys-
tems that should always get a full backup and so on.
A set of backup options are entered in a dumptype section,
which looks like this:
define dumptype name {
dumptype-option dumptype-value
...
}
Name is the name of this set of backup options. It is
referenced from the disklist file.
Some of the options in a dumptype section are the same as
global options. The global option value is used to set
the default for all dumptype sections. For instance, set-
ting dumpcycle to 50 in the main part of the config file
causes all the dumptype sections to start with that value,
but you may change it on a section by section basis.
The dumptype options and values are:
auth "string"
Default: bsd. Type of authorization to perform
between tape server and backup client hosts. May
be krb4 to use Kerberos-IV authorization.
comment "string"
Default: none. A comment string describing this
set of backup options.
comprate float [, float ]
Default: 0.50, 0.50. The expected full and incre-
mental compression factor for dumps. Normally you
should not set this because Amanda keeps the actual
full and partial dump compression rates in the his-
torical database.
compress [client|server] "string"
Default: client fast. If Amanda does compression
of the backup images, it can do so either on the
backup client host before it goes to the holding
disk or on the tape server host as it goes from the
dump image usually compresses, the speed and load
on the client or server, network capacity, holding
disk capacity, tape hardware compression, etc.
For either type of compression, Amanda also allows
you to select one of two styles of compression.
Best is the best compression available, often at
the expense of CPU overhead. Fast is often not as
good a compression as best, but usually less CPU
overhead.
So the compress options line may be one of:
compress none
compress [client] fast
compress [client] best
compress server fast
compress server best
Note that some tape devices do compression and this
option has nothing to do with whether that is
enabled, although you may want to disable Amanda
compression if the drive is also going to do it.
The tapedev path usually determines whether hard-
ware compression is enabled.
dumpcycle int
Default: 10 days. The number of days in the backup
cycle. Each disk using this set of options will
get a full backup no less often than this many
days. Setting this to zero forces a full backup
each run.
exclude [list] "string"
Default: none. The string is passed as a value to
the GNU tar --exclude or --exclude-list option. In
the case of exclude list, the file must be avail-
able on the backup client host.
holdingdisk "boolean"
Default: yes. Whether a holding disk should be
used for these backups or whether they should go
directly to tape.
ignore "boolean"
Default: no. Whether disks associated with this
backup type should be backed up or not. This
option is useful when you share a disklist file
among several configurations, some of which should
not back up all the listed file systems.
index "boolean"
kencrypt "boolean"
Default: no. Whether the backup image should be
encrypted by Kerberos as it is sent across the net-
work from the backup client host to the tape server
host.
maxdumps "int"
Default: 1. The maximum number of backups from a
single host that Amanda will attempt to run in par-
allel. See also the inparallel option.
priority "string"
Default: medium. When there is no tape to write
to, Amanda will do incremental backups in priority
order to the holding disk. The priority may be
high, medium or low.
program "string"
Default: DUMP. The type of backup to perform.
Valid values are DUMP for the native operating sys-
tem backup program, and GNUTAR to use GNU tar.
record "boolean"
Default: yes. Whether to ask the backup program to
update its database (e.g. /etc/dumpdates) of time
stamps. You normally want this enabled for daily
backups and turned off for periodic archival runs.
skip-full "boolean"
Default: no. If true and planner has scheduled a
full backup, these disks will be skipped.
skip-incr "boolean"
Default: no. If true and planner has scheduled an
incremental backup, these disks will be skipped.
starttime "int"
Default: none. Backups will not start until after
this time of day. The value should be hh*100+mm,
e.g. 6:30PM (18:30) would be entered as 1830.
strategy "string"
Default: standard. Strategy to use when planning
what level of backup to run next. Values are:
standard
The standard Amanda schedule.
nofull Never do full backups, only level 1 incre-
mentals.
dumps.
skip Never do backups (useful when sharing the
disklist file).
The following dumptype entries are predefined by Amanda:
define dumptype no-compress {
compress none
}
define dumptype compress-fast {
compress client fast
}
define dumptype compress-best {
compress client best
}
define dumptype srvcompress {
compress server fast
}
define dumptype bsd-auth {
auth bsd
}
define dumptype krb4-auth {
auth krb4
}
define dumptype no-record {
record no
}
define dumptype no-hold {
holdingdisk no
}
define dumptype no-full {
skip-full yes
}
In addition to options in a disktype section, you may also
put one or more other disktype names, which make this
disktype inherit options from other disktypes. For
instance, you might want two sections that are the same
except for the record option:
define dumptype normal {
comment "Normal backup, no compression, do indexing"
no-compress
index yes
maxdumps 2
}
define dumptype testing {
comment "Test backup, no compression, do indexing, no recording"
normal
no-record
}
The config file may define multiple types of tape media
and devices. The information is entered in a tapetype
section, which looks like this in the config file:
define tapetype name {
tapetype-option tapetype-value
...
}
Name is the name of this type of tape medium/device. It
is referenced from the tapetype option in the main part of
the config file.
The tapetype options and values are:
comment "string"
Default: none. A comment string describing this
set of tape information.
filemark "int"
Default: 1000 bytes. How large a file mark (tape
mark) is, measured in bytes. If you only know the
size in some linear measurement (e.g. inches), con-
vert it to bytes using the device density.
length "int"
Default: 2000 kbytes. How much data will fit on a
tape.
speed "int"
Default: 200. How fast the drive will accept data,
in bytes per second.
In addition to options, you may also put another tapetype
name, which makes this tapetype inherit options from
another tapetype. For instance, the only difference
between a DLT4000 tape drive using Compact-III tapes and
one using Compact-IV tapes is the length of the tape. So
they could be entered as:
define tapetype DLT4000-III {
comment "DLT4000 tape drives with Compact-III tapes"
length 12500 mbytes # 10 Gig tapes with some compression
filemark 2000 kbytes
speed 1536 kps
}
define tapetype DLT4000-IV {
comment "DLT4000 tape drives with Compact-IV tapes"
DLT4000-III
length 25000 mbytes # 20 Gig tapes with some compression
}
The config file may define multiple types of network
interfaces. The information is entered in an interface
section, which looks like this:
define interface name {
interface-option interface-value
...
}
Name is the name of this type of network interface. It is
referenced from the disklist file.
Note that these sections define network interface charac-
teristics, not the actual interface that will be used.
Nor do they impose limits on the bandwidth that will actu-
ally be taken up by Amanda. Amanda computes the bandwidth
each file system backup will take based on the estimated
size and time, then compares that plus any other running
backups with the limit as another of the criteria when
deciding whether to start the backup. Once a backup
starts, it is free to use as much of the network as it
can.
The interface options and values are:
comment "string"
Default: none. A comment string describing this
set of network information.
use "int"
Default: 300 Kbps. The speed of the interface in
Kbytes per second.
In addition to options, you may also put another interface
name, which makes this interface inherit options from
another interface. At the moment, this is of little use.
DISKLIST FILE
The disklist file determines which disks will be backed up
by Amanda. The file contains one line per disk:
hostname diskdevice dumptype [ spindle [ interface
] ]
Lines starting with # are ignored, as are blank lines.
The fields have the following meanings:
hostname
The name of the host to be backed up.
diskdevice
The name of the disk device to be backed up. It
as /usr.
dumptype
Refers to a dumptype defined in the amanda.conf
file. Dumptypes specify backup related parameters,
such as whether to compress the backups, whether to
record backup results in /etc/dumpdates, the disk's
relative priority, etc.
spindle
Default: -1. A number used to balance backup load
on a host. Amanda will try not to run multiple
backups on a given spindle.
interface
Default: local. The name of a network interface
definition in the config file, used to balance net-
work load.
TAPE MANAGEMENT
The tapelist file contains the list of tapes in active
use. This file is maintained entirely by Amanda and
should not be created or edited for normal operation. It
contains lines of the form:
YYYYMMDD label
Where YYYYMMDD is the date the tape was written, and label
is a label for the tape as written by amlabel.
Amdump and amflush will refuse to write to an unlabeled
tape, or to a labeled tape that is considered active.
There must be more tapes in active rotation (see the
tapecycle option) than there are runs in the backup cycle
(see the dumpcycle option) to prevent overwriting a needed
backup file when a problem occurs.
AUTHOR
James da Silva <jds@cs.umd.edu>
University of Maryland, College Park
SEE ALSO
amadmin(8), amcheck(8), amcheckdb(8), amcleanup(8),
amdump(8), amflush(8), amlabel(8), amoverview(8), amre-
cover(8), amrestore(8), amstatus(8), amtoc(8)