amanda-taperscan - Amanda Taperscan Algorithms
Amanda uses a taperscan algorithm to select volumes on which to store
dumps. Historically Amanda has provided only one, fairly complex
taperscan algorithm, but this algorithm did not suit the needs of all
users. Now Amanda offers a wide array of algorithms to suit varied
needs. Adding new algorithms is not difficult.
The taperscan algorithm is specified with the taperscan parameter,
which has a default value of traditional.
Taperscan algorithms are implemented as perl packages with the prefix
Amanda::Taper::Scan::. See the perl documentation for
Amanda::Taper::Scan for more information.
In general, these volumes will only select reusable volumes. These are
volumes which are listed in the tapelist(5) with the reuse flag, and
which are not among the tapecycle-1 most recent volumes in the list.
Put another way, reusable volumes do not contain data that must be
retained. Note that if fewer than tapecycle-1 volumes have been written
then there are no reusable volumes. Newly labeled volumes (volumes that
have been labeled with amlabel but never used) are considered reusable.
Many of these algorithms look for the oldest reusable volume. In most
cases, this is the best volume to overwrite, as the data it contains is
older than that on any other volume. If there are no reusable volumes,
then there is no oldest reusable volume.
This algorithm duplicates Amanda´s historical behavior, and it operates
in two stages.
First, if there is an oldest reusable volume and if the changer
supports "fast" searches, then the algorithm uses the changer to search
for that volume. Newly labeled volumes are not considered when
calculating the oldest reusable volume. Consequently, this taperscan
algorithm prefers volumes which have been used before to newly-labeled
volumes when a fast-searchable changer is in use.
Second, if there is no oldest reusable volume, or if that volume is not
available in the changer, then the algorithm begins a sequential scan
of the changer, starting at the current slot. It selects the first
suitable volume it finds: a reusable volume (perhaps newly-labeled) or,
if label_new_tapes is true, a blank volume. Even across multiple
invocations (when runtapes > 1), it will not return the same slot
This algorithm shows an undue preference for volumes already
containing data, by omitting newly-labeled volumes from its first
stage. Historically, many Amanda changer scripts were not
fast-searchable (including chg-multi, chg-disk and, if
havereader=0, chg-zd-mtx), and thus skipped the first stage,
allowing new volumes to find their way into the tape cycle. New
changers are almost all fast-searchable, so both stages of the
algorithm are used and new tapes may be unexpectedly excluded. If
this causes an undesirable change in behavior, consider one of the
other taperscan algorithms..sp .5v
amanda(8), amanda.conf(5), tapelist(5), amanda-changers(7)
The Amanda Wiki: : http://wiki.zmanda.com/
Dustin J. Mitchell <firstname.lastname@example.org>
Zmanda, Inc. (http://www.zmanda.com)