Table of Contents
There are times when data needs to be excluded from a backup. When these times arise be confident that Amanda has this capability. (Actually it's not Amanda, it's tar.) There are three ways of excluding data in an Amanda backup:
Exclude an individual item explicitly in the dumptype
Utilize an "Exclude List"
Do not include the data in the disklist
This document is based on Amanda 2.4.2 and some of this might not work with older versions. This was compiled from my personal experience and with help from the members of the amanda-users mailing list (mailto://[email protected]) when I was originally setting this up, to whom I wish to thank for all of their support.
As far as I am able to tell the only way to exclude files or directories with Amanda is to use GNU-tar as the dump program (others?). The file system dump programs provided with unix systems (e.g. dump, ufsdump) get data at a raw drive level and generally do not allow exclusion of specific files or directories.
The GNU version of tar, (GNU-tar or gtar), reads its data at a file system, (or higher), level and does include the option to exclude specific files and/or directories. It should be mentioned here that tar will change the access times on files. Tar has the ability to preserve the access times however, doing so effectively disables incremental backups since resetting the access time alters the inode change time, which in turn causes the file to look like it needs to be archived again.
The only exception that I am aware of is to just not include the data in question in the disklist. This option may not be suitable for everyone's needs and can confuse the issue some, so I have elected to include this mechanism in its own section named Do not include the data in the disklist.
For the purpose of this document an Amanda backup configuration named "exclude-test" will be used. The machine that contains the tape drive which receives data to be archived will be referred to as "SERVER". The machine that data is being archived from will be referred to as "CLIENT". These two systems are usually different machines but are not required to be, and may be the same machine. Parts of this setup are on the server and some are on the client.
When Amanda attempts to exclude a file or directory it does so
relative to the area being archived. For example if /var is in
your disklist and you want to exclude /var/log/somefile, then your
exclude file would contain ./log/somefile. You may use one
exclude file in multiple dumptypes without any restriction.
The first step that should be taken is to verify that backups are currently working. Connect to SERVER and run amcheck as your Amanda user, to verify that there are no errors in the current setup.
$ amcheck -cl CLIENT
Output should look something like below for success:
Amanda Tape Server Host Check ----------------------------- /path/to/holding-disk: 4771300 KB disk space available, that's plenty. Amanda Backup Client Hosts Check -------------------------------- Client check: 1 host checked in 0.084 seconds, 0 problems found.
Next make sure that GNU-tar is the dump program currently in use. The easiest way to tell if your dumptype is using gnutar is to run the following:
$ amadmin exclude-test disklist CLIENT
Among all the output is the "program" value currently in use. This value is also specified with the "program" option in the dumptype. If the dumptype has the line "program GNUTAR" your setup should be ready to exclude data.
If GNU-tar is not in use add the line "program GNUTAR" to the dumptype, and then run amcheck again to verify that backups should work. The capitalization of GNUTAR is required.
The dumptype should look something like:
define dumptype exclude-test {
comment "test dumptype for documentation"
priority high
program "GNUTAR"
}If the need is to exclude only one file or directory then the easiest way to accomplish this is to exclude an individual item explicitly in the dumptype. If the need is to exclude multiple files or directories then use an Exclude List.
The easiest way to exclude a file or directory is to specify it with the "exclude" option in the dumptype. This option accepts an argument of the file or directory to be excluded. Amanda allows only one exclude option in any dumptype at a time.
Any path specified to be excluded must be encapsulated with quotes. Continuing
with our example from above /var/log/somefile and using the same
dumptype as above, the dumptype would now look like:
define dumptype exclude-test {
comment "test dumptype for documentation"
priority high
program "GNUTAR"
exclude "./log/somefile"
}Next run amcheck again to verify that there are no problems with the revised Amanda configuration. If the data is not being excluded as expected please see the Troubleshooting section below. This completes the setup of excluding an individual item in the dumptype.
An exclude list is a file that resides on the CLIENT machine and
contains paths to be excluded, one per line. This file can be in
any location on the CLIENT so long as the same path is specified
in the dumptype. Some find /usr/local/etc/amanda an appropriate
location, but it is up to you. I personally like to have a
subdirectory for exclude files but it is up to you where you place
this file.
The exclude file may also be placed in the area being archived. This is an easy way to have a different exclusion file for each disklist entry without needing separate dumptype definitions. To use this technique, enter a path relative to the area being archived as the exclude file below instead of an absolute path.
Connect to CLIENT and create the exclude directory as root. For example:
$ mkdir -p /usr/local/etc/amanda/exclude $ cd /usr/local/etc/amanda/exclude
Next create the exclude list for Amanda to use. You can name the
exclude file anything you wish it to be. Create a file, and in
this file place all paths to files and directories that are to be
excluded. Keeping with the /var example, assume that
/var/log/XFree86.0.log, and /var/log/maillog need to be excluded.
Remember that all paths are relative. The exclude list would look
like:
./log/XFree86.0.log ./log/maillog
Make sure that permissions are restricted on this file. Run the following as root, where exclude-filename is the name of the file you just created. For example:
$ chmod 644 /usr/local/etc/amanda/exclude/exclude-filename
This concludes the necessary configuration on the client.
Connect to SERVER and cd to the exclude-test Amanda configuration
directory. Edit the Amanda configuration file e.g. amanda.conf.
Add an entry similar to the following line, to the dumptype for
the client in question, where the exclude-filename is the file that
was created on CLIENT in the step above including the quotes.
For example:
exclude list "/usr/local/etc/amanda/exclude/exclude-filename"
The new dumptype should look something like:
define dumptype exclude-test{
comment "test dumptype for documentation"
priority high
program "GNUTAR"
exclude list "/usr/local/etc/amanda/exclude/exclude-filename"
}Save the file. Run amcheck again to verify that there are no problems with the revised Amanda configuration. If amcheck succeeds then run amdump to verify the data is being excluded correctly. If the data is not being excluded as expected please see the Troubleshooting section below. This completes the setup of an exclude list.
Amanda uses disklist entries to define which directories or partitions
should be archived. This allows us to exclude data by just not
placing the data in question in the disklist. Assume that there
is a disk mounted on /example. The directory
/example has five subdirectories
"a", "b", "c", "d", and
"e".
The directories "a", "b", and "c" need
to be archived, while "d" and "e" should not.
This can be accomplished by not specifying "d" and "e" in the
disklist. Using the same dumptype and host in the above examples
the disklist would contain:
CLIENT /examples/a exclude-test CLIENT /examples/b exclude-test CLIENT /examples/c exclude-test
Run amcheck to verify that Amanda is working correctly. If the data is not being excluded as expected please see the Troubleshooting section below. This completes the setup of using a disklist to exclude data.
Quiz: what is the difference between the following entries in an exclude list?
./foo ./foo/ ./foo/*
case 1 : directory ./foo won't be in the backup image (that's what you want) case 2 : matches nothing (don't use it) case 3 : directory ./foo will be in the backup image but nothing below it.
Amanda has the ability to use wildcard expansion while excluding data as implemented by tar(1). The only places that wildcard expansion is allowed is in the "exclude" option in the dumptype, or in the exclude list. Some simple examples:
Exclude any file or directory that ends in ".log" e.g. ppp.log, XFree86.0.log
./*.log
Exclude any file or directory with the string "log" e.g. logfile, maillog, syslog, ppp.log, XFree86.0.log
*/*log*
Exclude any file or directory that starts with string "cron" and ends in ".gz" e.g. cron.1.gz, cron.2.gz, log/cron.1.gz
./*cron*.gz
The question mark can be used to specify a single character. e.g. log.1, log.2, etc
./log.?
If you find that you are having trouble getting the exclude patterns to match correctly, check out this really cool script written by John R. Jackson.
ftp://gandalf.cc.purdue.edu/pub/amanda/gtartest-exclude
This script allows you to test your patterns before placing them in an exclude list or in the dumptype. Instructions on how to run the script are included in the script.
There are versions of GNU-tar that do not correctly exclude data. Version 1.12 (plus the Amanda patches from http://www.amanda.org) are known to work correctly, as does version 1.13.19 (and later). Anything else is questionable.
Connect to CLIENT and run a
ps ax | grep tar(or
ps ef | grep taron Solaris)
to see exactly how the tar command is running. Look in the output for the --exclude or --exclude-from options in the running tar process. For example:
$ ps ax | grep tar ? R 0:37 /bin/tar --create --directory /var --listed-incremental /var/lib/amanda/gnutar-lists/CLIENTvar_0.new --sparse --one-file-system --ignore-failed-read --totals --file /dev/null --exclude-from=/usr/local/etc/amanda/exclude-test/exclude.var .
In the above output notice the string "--exclude-from=". The string following the "=" is the exclude file currently in use. If the string was "--exclude" then the string following the "=" is the file or directory that is currently set to be excluded.
Contact the amanda-users mailing list: mailto://[email protected]. Subscription information is available at http://www.amanda.org.