amfetchdump — extract backup images from multiple Amanda tapes.
amfetchdump [-pcClawns] [-d device] [-O directory] [-i logfile] [-b blocksize] config hostname [
disk
[
date
[
level
[
hostname
[...]
]
]
]
] [ -o configoption]...
Amfetchdump pulls one or more matching dumps from tape or from the holding disk, handling the reassembly of multi-tape split dump files as well as any tape autochanger operations.
It will automatically use the logs created by amdump(8) to locate available dumps on tape, in the same way that the find feature of amadmin(8) lists available dumps. If these logs are unavailable, it can search tape-by-tape to find what it needs, and can generate new logs to serve as an emergency tape inventory.
The hostname, diskname, datestamp, and level dump pattern-matching works as in amrestore(8), with the added requirement that at minimum a hostname must be specified when not in inventory mode.
Unless
-p
is used, backup images are extracted to files in the current directory named:
hostname.diskname.datestamp.dumplevel
-pPipe exactly one complete dump file to stdout, instead of writing the file to disk. This will restore only the first matching dumpfile (where "first" is determined by the dump log search facility).
-d deviceRestore from this tape device instead of the default.
-O
directoryOutput restored files to this directory, instead of to the current working directory.
-cCompress output, fastest method available.
-CCompress output, smallest file size method available.
-lLeave dumps in the compressed/uncompressed state in which they were found on tape. By default, amfetchdump will automatically uncompress when restoring.
-aAssume that all tapes are already available, via tape changer or otherwise, instead of prompting the operator to ensure that all tapes are loaded.
-i filenameGenerate an inventory of all dumps "seen" on the tapes we search, for later use as a log.
-wWait to put split dumps together until all chunks have been restored. Normally, amfetchdump will attempt to read pieces of a split file from tape in order, so that it can assemble them simply by appending each file to the first. This option disables the appending behavior, and instead restores each piece as an individual file and reassembles them only after all have been restored.
This requires at least double the size of your dump in free disk space, in order to build the final assembled dumpfile.
This behavior is implicitly invoked in circumstances where knowing the location of all dumps on tape in advance is not possible, such as when you are restoring without log files.
-nDo not reassemble split dump files at all, just restore each piece as an individual file.
-sDo not fast-forward straight to needed files on tape. This will slow down most restores substantially. Only use this option if your tape drive does not properly support the fast-forward operation.
-b
blocksizeForce a particular block size when reading from tapes. This value will usually be autodetected, and should not normally need to be set.
configoptionSee the "CONFIGURATION OVERRIDE" section in amanda(8).
All the examples here assume your configuration is called
SetA.
Here's a simple case, restoring all known dumps of the host vanya to the current working directory.
$ amfetchdump SetA vanya
A more likely scenario involves restoring a particular dump from a particular date. We'll pipe this one to GNU-tar as well, to automatically extract the dump.
$ amfetchdump -p SetA vanya /home 20051020 | gtar -xvpf -
In a situation where all of our dump logs have been wiped out, we could also use amfetchdump to inventory our tapes and recreate an imitation of those logs, which we'll send to stdout for casual perusal.
$ amfetchdump -i - SetA
Note that you can specify a restore while in inventory mode, and amfetchdump will continue searching for more dumps from this host even after successfully restoring a dump, inventorying all the while. If your backup searcher has been trashed, this is a handy way to recover what you have.
$ amfetchdump -i /var/amanda/log SetA backupserver
Amfetchdump is dependent on accessing your server's config, tape changer, and (normally) dump logs. As such, it's not necessarily the most useful tool when those have all been wiped out and you desperately need to pull things from your tape. Pains have been taken to make it as capable as possible, but for seriously minimialist restores, look to amrestore(8) or dd(8) instead.
John Stange, <[email protected]>, National Academies Press
Ian Turner, <[email protected]>: XML-conversion