cd_backup_planer version 0.9.2
usage : bin/cd_backup_planer [option | file] ...
Usually this command splits a list of directories into volumes
which fit on CD-R media.
It generates a script with backup commands. The lists of files
or directories are determined according to maximum capacities
in terms of total size and number of items.
A result list can contain only directories and datafiles.
Other file types may be included in directories but most CD
writer programs will not process them.
The result lists consist of target and source addresses suitable
for the program mkisofs in its more modern versions:
If fileaddress is a directory: /targetaddress/=fileaddress
If it is a data file: /targetaddress=fileaddress
These lists are appended to the backup command in order to fill
one CD medium. If not given explicitely and not redirected by a
collision, the targetaddress is the same as the fileaddress.
Depending on the -target_escape mode the character "=" either
is escaped or changed to "_".
Other output forms of the result lists are available too.
Since the working directory, the root directory and possibly
the -main_dir all are mapped to the root directory collisions
of targetadresses may occur. A builtin collision test can detect
and avoid them by renaming one of the affected targetadresses.
A filecheck can be made to find problematic files:
- those which are ignored by mkisofs-1.12b3 if not given
explicitely as arguments.
- those which cannot be read (files or directories)
this can mislead the estimation of resulting directory sizes.
The combined options -filecheck -append -filecheck -exclude_unreadable
can help to work around these problems.
Incremental backups are possible with an arbitrary number of levels
-level 0 does a full backup with no time constraints.
-level N backups those files that have changed since the last run
of -level N-1 . -level N-1 has to exist already.
- A backup with -level N makes invalid all levels above N.
The data necessary for keeping track of levels are kept in a
dedicated directory which is addressed by -conf_dir. It also hosts
all parameters of the backup like filelist or maximum sizes.
Incremental backups use the option -younger_than which possibly
also includes some older files in the backup. Regrettably, mere
name changes do not change the file's time stamps in the general
Linux filesystem model (other than e.g.: chmod u+r ). Therefore
lengthy checksum computations often become unavoidable.
Since a large number of single filenames could make problems with
older versions of mkisofs, the incremental backup mainly tries to
exclude directories which contain only very few young files and to
mention only these few exemption files. Also a certain effectivity
is required in terms of byte gain versus number of filenames. This
required quote is high if -max_items is set to a small value.
An additional script can be generated which is able to print
several informations about the backup.
Special options are:
-hilfe, -id, -count, -list, -levelall , -suball, -fast, -help
Other arguments are looked up as file names in the lists.
If a name or a directory above this name is found, then the part
number, the item and the file name are printed. If none of the
files can be found this way, then the script's exit value is 1.
This script is stored on the disk and included in each part of
backup if a name is given with option -info_script (see below).
It is estimated to have 100k size (in case that your scripts
actually exceed this size reduce -max_size accordingly).
If the info script gets included and the output form for mkisofs
(default or -form_script) is used then all possible fileadresses
which may lead to the same name on the CD are excluded.
A number as optional argument tells the backup script to skip all
volumes with lower numbers. To continue a backup which has been
interrupted while volume 3 was ready to be written:
scriptname 3
Then the script skips volumes 1 and 2.
If the script has been generated with -status_file then it can find
the appropriate volume automatically by reading the status file
This automatic is triggered with the argument "AUTO".
Arguments which do not begin with '-' are file addresses.
Fileadresses may be prefixed by target adresses. This is useful
for composing a CD from various scattered sources. The usual
separating character is "=" but another one may be chosen
with the option -target_separator.
Example: -target_separator '=' /images=/home/private/images ...
All fileaddresses together define the "backup area" from which some
sub trees may be excluded by options -no, -not, -xdev, -localdev .
This set of file objects may get examined and reduced further
(e.g. by -younger_than) and so becomes the "backup content".
Options:
Several options (like -max_size) expect size arguments. This may be integer
numbers (like 1000000000) or floating point numbers (like 1.25e9). They may
have appended a binary unit letter out of {b,k,m,g,t,p,e} which multiplies
the number by {1,1024,1024*1024,1024*1024*1024,...,pow(1024.0,6.0)}.
Example: 640m = 640*1024*1024 = 671088640
-help print this text and exit program
-hilfe diesen Text ausgeben und Programm beenden
(like -help , use german language)
-version print version identification on stdout,
if not only argument: print warning message on stderr,
in any case : end program with exit value 4.
-- the next argument is a file, not an option.
-accept_all_filetypes
Is equivalent to : -result_filetypes any
-alldev do not exclude mounted filesystems. (default)
-backup_cmd text the given text is used as command in the script generated
by this program. text is subject to special -bdir
substitution. (default: cd_backup_kommando)
-bd set size paramters suitable for single layer Blu-ray
media of nominal "25 GB" with medium size formatting.
-bdir directory sets the scdbackup installation directory. Currently the
only use of this information is -bdir substitution. It
is performed when reading file specific_arguments of an
incremental backup configuration. If such an argument
starts with @SCDBACKUP_DIR@ then this start piece gets
replaced by the -bdir directory. This works too if a
quotation mark is prepended.
With some very few arguments the start piece is replaced
also if it matches grep expression
.*/scdbackup-[0-9]\.[0-9]\.[0-9]/.*
This is called special -bdir substitution.
-bd_dl set size paramters suitable for double layer Blu-ray
media of nominal "50 GB" with medium size formatting.
-blocklist_fileadr address_prefix
if address_prefix is not empty and not "-" then it will
be used to generate addresses of block checksum lists
which are eventually to be stored. Wether there are such
lists depends on:
-filter_checksum_blocksize with -filter_md5
-scan_for_checksum -blocklist with -search_md5
The effective file address will consist of address_prefix,
record name, date stamp and size value of the checksum
record. For example : "/tmp/x_" may lead to:
/tmp/x_1_1_A41123.194219_229208064
-changetest_options code:code:...
chooses properties to be compared during check
-younger_than. Compared are the file objects on
the disk with those obtained by -load_old_tree.
Multiple codes are separated by ":".
type e.g. directory, data file, softlink
size bytecount
crc CRC-32 or MD5 content checksum
inode internal ID number of filesystem object
off disable comparison with old tree
adr enable comparison with old tree (default)
mistrust avoid a shortcut with CRC depending on
directory timestamps.
dir_trust allows an even faster shortcut with CRC but
is not safe if directories get renamed
in order to replace existing directories.
assume_old assume all timestamps as old enough
ignore_ctime inspect only mtime but not ctime.
ignore_device with inode: do not compare device numbers
no_split_economy no check for uneconomical directory
splittings by small, old files.
NOTE: -load_old_tree with nonempty addresses
sets -changetest_options to "type:size:crc:adr"
Set own -changetest_options after -load_old_tree
If both :inode: and :crc: are chosen, and both information
is available, then :inode: will be used for decision.
-collision_test mode controls the collision test for target names
Modes can be (default is -redirect ) :
-off do not perform collision test
-on warn about collisions
-redirect change target names if necessary
-abort abort if a collision is detected
-compare_checksum recordlist searchname old_limit young_limit
the program ends immediately when stdin closes.
It passes stdin to stdout without changes.
Reads from file recordlist a record matching the
other three arguments. The record may have been
written by -filter_crc32 or -filter_md5. In order
to match, searchname has to be "-" or exactly
the same as the name of the record. If old_limit
is not "-" all older records do not match. If
young_limit is not "-" then all younger ones do
not match. (see -younger_than for time formats)
The last matching record in the list is read and
its recorded data amount is read from stdin to
compute a checksum which is compared with the
recorded checksum. If stdin delivers enough
bytes and the checksums are equal, then the
return value of the program is 0. In case of
not enough input or wrong checksum exit is 1.
Any surplus bytes at stdin are ignored.
-complete_list_adr fileaddress
writes a complete file list of the backup area
to the given fileaddress. If fileaddress contains no "/"
and a -level is given then the file is created in the
level_# directory. (See below : -conf_dir) If no -level
is given then such a file goes to -tmp_dir.
If the given fileaddress ends with .gz then the
output gets piped through program gzip.
If a target is given in fileaddress then the
resulting file will be appended to the last
backup volume. (e.g. /complete.gz=complete.gz)
The list is written in a special format to save
disk space while still being human readable. The
first and the last characters of an address determine
its type
/x/x/x/ absolute directory address
/x/x/x absolute file address
./x/ descent into directory x
./x file x in current directory
.. one directory up
../ to start directory
#=# xxxx additional information
The first directory address given is absolute and
sets the start directory for relative addresses.
By ./x/ and /x/x/x the directory cursor can be
moved away from the start directory. As long as
the cursor is in the start directory, addresses
get prepended a target address which also is
valid for all relative addresses below.
Both parts are separated by a "="-character. If the
target address itself contains "=" or "\" then they
are prepended by a "\" each.
The resulting list is usable with -vanish_list_adr
-composition_memory depth width memory_limit
defines size and rules for an acceleration buffer which
may help with dividing very large and deep directory
trees into digestible pieces. Parameter depth controls
how many levels of sub directories shall be recorded for
the case that their sizes is requested again after a
directory split. If they are recorded then there is no
to examine the subtrees on disk again. Thus the buffer
can save a lot of time but can also need lots of memory.
Parameter width controls how many files a directory may
contain to be recorded in the buffer.
The memory consumption per file depends mainly on the
lenght of the filenames. Plus about 20 Bytes. Total
memory consumption is curbed by parameter memory_limit.
If memory_limit is preceeded by "-" then freed buffer
is not regarded as reusable.
Default : 3 1000 4m (0 0 0 would disable the buffer)
-composition_mode criterion
defines the sequence of equal ranked file objects when
backup volumes are composed. All modes try to keep
directory trees together. When arranging objects from the
start list or objects within split directories there
may be used one of several criteria to choose the
next file object :
bestfit the largest object that will fit
target the next higher target address (alphabetically)
source the next higher source address
time the next younger file object
crawler the next one found by tree crawler
default currently like bestfit
alphabetical currently like target
chronological currently like time
fast currently like crawler
With criteria other than bestfit there is an increased
chance to get larger directories split over two adjacent
backup volumes. This probability is mainly controled by
-min_size_percent where a low percentage causes waste of
media capacity and high percentage causes more directory
splits. Waste will be prevented by splitting only if the
number of -max_items is large enough to accept nearly all
Fileobjects in the split directory.
Criterion crawler yields results like criterion source
as long as the directories have a moderate size. At volume
borders it always splits directories. It avoids one or
two passes over the backup area which may need substantial
time if RAM cannot buffer the whole backup area tree
information.
-content_list_adr fileaddress
writes a file list of the backup content. With
the same rules and formats as -complete_list_adr
this option lists only those file objects which
are part of a backup volume. The content lists
and vanish lists of all levels together contain
the same information as a single complete list.
Additionally a content list tells the volume
numbers by marking volume borders.
-conf_dir directoryaddress
sets the location of a backup configuration. It
contains files which are created and used by
incremental backups. (see -level)
specific_arguments : full backup's options list
each line is one argument.
highest_level : highest valid level number
ASKME_level_# : info script of -level #
level_#/ : level directory, contains
last_run_date : time stamp of last run
number_of_volumes : as it says
content_file_list.gz : see -content_list_adr
and -load_old_tree
vanished_file_list.gz: see -vanish_list_adr
VANISHED.sh.gz : see -vanish_script_adr
-cut_file source_address start_byte byte_count target_address
The program ends immediately after this option is
done. It copies an interval of bytes out of a file into
a new one. Intended to be used by -split_large_files .
If source_address is '-' then data are read from stdin
and a number of start_byte bytes is skipped, else the
source_address is opened as file and the read pointer is
positioned at start_byte. target_address may be '-' for
stdout or a suitable address for a new file. The
intermediate messages are controlled by -pacifier .
-debug_filelist list_adr directory_prefix
The program ends immediately after this option is done.
It prints the list lines together with eventual resulting
absolute file addresses. directory_prefix is prepended to
any of these addresses. List lines get prepended three
blanks and the count of already listed file items.
So they may get filtered away by grep -v '^ '
-debug_stderr_log The program ends immediately after stdin closes.
It changes carriage returns into newlines and tries to
spare empty lines. Intended for making more readable
the results of scdbackup ... 2>&1 | tee -i ...
-debug_targetlist list_adr directory_prefix
like -debug_filelist but showing the target addresses
of the files.
-delete_filelist list_adr directory_prefix
NOTE: for removing obsolete files during the restore of
a backup you should better use -delete_targetlist.
The program ends immediately after this option is
done. It reads file addresses from list_adr
together with type, size, checksum properties.
These informations get compared with the current
state of the file whose address is composed from
directory_prefix and the address read from the
list.
The file list has to be generated with
-vanish_list_adr and must contain type, size
and checksum informations.
If the addressed file exists and if the other
properties match, then the file gets deleted.
Directories get deleted if they are empty.
If everything existed and matched, the program
exits with value 0. Else this value is 1.
-delete_targetlist list_adr directory_prefix
like -delete_filelist but processing the target addresses
of the vanished files. If there have been target=source
redirections when a backup was made, then this can make
a big difference.
-dir_list_adr fileaddress
writes a complete directory list of the backup area to
the given fileaddress. The output format is similar to
-complete_list_adr but the following fields are fixely
enabled: "type:chmod:chown:chgrp:date:size:crc:adr" with
chmod:chown:chgrp:date defining access permissions,
ownership and time stamp. If MD5 checksums have been
enabled by -list_output_format then they are used here
rather than CRC32 checksums.
-dir_script_adr fileaddress
writes a shell script for restoring the access permissions
of the directories of the backup area. The script will use
target addresses relative to the CDs' root directory.
-dir_size_cmd text the given text is used as command for determination of
a directory's size. The command has to expect a single
argument (the adress of the directory) and give back as
first word on standard output the size in terms of
kilobytes (like du -sk does).
Default: "-internal" does not run an external command
but recursivly sums up the sizes of file objects and the
length of their addresses. Added is an overhead of 256
bytes per file. It may be changed by "-internal+$num" .
For example "-internal+10" adds 10 extra bytes.
"-internal-$num" disables adding of address length.
-disable_safe_leveling is equivalent to
-load_old_tree "" ""
-content_list_adr "" -complete_list_adr ""
-vanish_script_adr "" "" -vanish_list_adr "" ""
-dvd set size paramters suitable for single layer DVD media
of nominal "4.7 GB". Not suitable for DVD-RAM.
-dvd_dl set size paramters suitable for double layer DVD media
of nominal "8.5 GB".
-dvd_ram set size paramters suitable for DVD-RAM media
of nominal "4.7 GB" with medium size formatting.
-exclude_size sizelimit exclude any single file which contains at least
sizelimit bytes. A sizelimit of 0 disables this test.
If sizelimit begins with "-" then with symbolic links
the size of the link target gets tested.
'k' 'm' and 'g' are recognized like with -max_size .
-extract_multiplied source skip
the program ends immediately when stdin closes. It
reads source sequentially and expects its data to stem
from option -multiply_volumes. The given number of copies
get skipped and the next one gets written to stdout.
Eventually a defunct marker tag is appended.
-extract_tag_tail maximum_size
the program ends immediately when stdin closes. It
consumes stdin until the header of a potentially valid
checksum tag is encountered, which is preceeded by at
least one byte. Beginning with this tag header, the rest
of the stdin stream is passed to stdout. Different from
-search_md5 even damaged tags will be accepted.
The passed amount of data will consist of the header plus
at most the amount of data given by maximum_size. Use a
generous maximum_size to catch all interesting data.
The exit value is 0 if data have been extracted. It is
non 0 else.
-extract_tag_tail_dislocated maximum_size
like -extract_tag_tail but rather accepting the first
occurence of a checksum tag identifier text. It is not
necessary that the position information of the tag
matches the position of the identifier text within the
data stream. Therefore it is possible to retrieve tags
from data streams where bytes are missing. On the other
hand this command accepts checksum tags which are part of
the payload data (which is not too uncommon). Only tags
which start at position 0 in the data stream get skipped.
-facl_list_adr fileaddress
applies program getfacl to the backup area and writes
its output to the given fileaddress. This will record
standard file permissions as well as eventual ACL
settings. The format is suitable for program setfacl
gunzip <facl_list.gz | setfacl --restore=-
The adresses recorded within the facl file are changed
to the target adresses of the backup items. Example :
-facl_list_adr /extra_lists/facl_list.gz=facl_list.gz
See man getfacl, setfacl, mount (option "acl").
-level, -tmp_dir and gzip compression apply like
described with -complete_list_adr.
-faclsrc_list_adr fileaddress
same as -facl_list_adr but recording the original source
addresses as on disk. Both options may be used (with
different fileaddresses) to produce both lists in the
same backup.
-fattr_list_adr fileaddress
applies program getfattr to the backup area and writes
its output to the given fileaddress. This will record
XFS style extended file attributes as far as they are
enabled and present. Recorded addresses get changed to
the according target addresses.
See man getfattr, setfattr, mount (option "user_xattr")
-level, -tmp_dir and gzip compression apply like
described with -complete_list_adr.
-fattrsrc_list_adr fileaddress
same as -fattr_list_adr but recording the original source
addresses as on disk. Both options may be used (with
different fileaddresses) to produce both lists in the
same backup.
-filecheck mode controls the examination of all files.
Modes can be (default is -off ) :
-off no examination (saves much time)
-on warn about -implicite_ignored files
and unreadable files
-append like -on. Append -implicite_ignored
files explicitely to the backup list.
-append_hexcode like -append.
The problem causing characters are
replaced by +XY+ (XY=ASCII code in hex)
-exclude same as -off (due to a long existent bug)
-exclude_unreadable like -on. Exclude unreadable
files explicitely from the backup list.
-abort like -on. Abort program if warnings
have been issued.
-only like -on. End program after check.
(exit 16 = warning issued , 0 = OK)
The modes may be combined.
-filter_checksum_blocksize number
sets the currently completest attachment of checksum
informations with options -filter_md5 and -filter_crc32 .
A checksum tag for the whole stream is appended as well
as a list of block checksums. The size of a block is set
by the given number. A small number will result in many
blocks and therefore will attach a large number of 16 byte
MD5 checksums.
This option sets -filter_checksum_padsize to 1 and
-pre_checksum_size to 0. It will eventually get disabled
by any of these two.
-filter_checksum_padsize number
with -filter_crc32 or -filter_md5 add at least the given
number of bytes in form of checksum tags at the end of
the filtered byte stream. Such tags may be found by
-search_crc32 resp. -search_md5 if -scan_for_checksum is
enabled. Thus avoiding the need of a record list.
If number is not 1 then -filter_checksum_blocksize gets
disabled.
-filter_crc32 recordlist recordname
the program ends immediately when stdin closes.
Passes stdin to stdout without changes. Computes
a CRC32 checksum and counts the bytes. Together
with the recordname and a time mark they get
appended as a record to file recordlist.
The recordname maybe a text without spaces with
up to 80 characters.
Such records are used by -compare_checksum .
If -pre_checksum_size is greater than 0 there will also
be written a preview record for -search_crc32 .
CRC32 got 32 bits and may fail in 1 of 4e9 tests.
The program will return 0 if any data were read.
If no bytes were seen at stdin, no record is
written and exit value is 1.
-filter_md5 recordlist recordname
the program ends immediately when stdin closes.
Like -filter_crc32 but with "RSA Data Security,
MD5 Message-Digest Algorithm" checksum.
MD5 got 128 bits and may fail in 1 of 3e38 tests.
-filter_pipe shellkommando
Given before -filter_md5 , -filter_crc32 or
-filter_through this defines a shell command
which gets started as soon as the first input
byte arrives on stdin. The filter's output is
piped into that command. If no bytes are received
then the shell command is not started at all.
-filter_through the program ends immediately when stdin closes.
Passes stdin to stdout without changes. Useful
together with -filter_pipe .
-find directory prints the directory's name and the names of all files
and directories below. Afterwards the program ends.
Like program 'find' but without any interpretation of
the given name.
-find_path_list filename
reads fileaddresses line by line from given file, applies
option -find to each of it and finally ends the program.
-follow_symbolic_links mode
In normal mode "off" symbolic links are counted only
with their own (small) file size. In mode "on" they will
be accounted with the size of their link target. In case
of directories this means that the whole tree below
the target is measured and eventually gets split over
several volumes. The formatter program will then have
to follow symbolic links in order to put the target data
into the backup. With mkisofs this is achieved by option
"-f" and with afio the appropriate option is "-h".
In case of ill (cyclical) link situations there may be
multiple copies of the affected directories included in
the backup. Watch for error messages like :
"Too many levels of symbolic links"
-form_listing the result list lines are shown in the form
Type Size [Partprefix]Address
the backup_cmd is not used. Partprefix may result from
-split_large_files and eventually has the form
**partnumber/total:start_byte+byte_count**
-form_path_list_afio listfilename safe_number_of_arguments
a special version of -form_script_raw. The file
addresses do not become arguments but are written
to listfilename (see -form_path_list_mkisofs) and
this listfilename is handed by option -path-list
to the writing command. It may then use option
-find_path_list to get a full file list for afio.
-form_path_list_afio_old listfilename safe_number_of_arguments
like -form_path_list_afio but without writing splitted
directories via option -dir-path-list of the writing
command.
-form_path_list_mkisofs listfilename safe_number_of_arguments
a special version of -form_script suitable for
versions of mkisofs which know option -path-list
With this form a very large setting of -max_times
is permittable (like 100000) since it avoids to
use them as arguments of shell commands.
listfilename will be created to store the list of
items. Therefore write permission is needed for
its hosting directory. If listfilename does not contain
a "/" then it gets prepended -tmp_dir.
safe_number_of_arguments should be small enough
to comfort your shell (e.g. 100).
This option sets -target_escape to "on".
-form_quoted_list_xorriso listfilename safe_number_of_arguments
a special version of -form_path_list_mkisofs suitable for
versions of xorriso which in its -as mkisofs emulation
knows option --quoted_path_list. This format can process
filenames which contain NewLine-characters.
-form_raw the result list lines only show the addresses
(with eventual partprefix like with -form_listing).
The volume ends are marked by empty lines.
-form_script the result list lines are appended as arguments
to the command in the form
/address[/]=address
this is the default output format.
This option sets -target_escape to "off".
-form_script_here the result list lines are appended as standard
input lines to the command in the form
address
this is done inline by a here-document with
delimiter "+".
-form_script_raw the result list lines are appended as arguments
to the command in the form
address
-implicite_ignored text addresses a problem with mkisofs-1.12b3
The text tells which characters cause the
backup command to ignore filenames that are
not explicitely given as arguments.
For mkisofs-1.12b3 text should be: "~#"
-info_script name set the fileaddress for the separate information
script. Be aware that two files are created.
One with the name given and one with name
extension .pnr (e.g. ASKME and ASKME.pnr).
Special names are:
-off don't generate info script
-short info script dummy without item lists
-full fully usable info script
-short_on_level With -level -create_configuration
default becomes -info_script -short
other names are used to store the script and
included in each part of the backup.
Different from file addresses the separating
character for the -info_script name is fixed
"=".
Example: /ASKME=/home/cd_backup_planer_dir/ASKME
-info_exclude don't include the info script in all backup
parts automatically despite a file name was set
with -info_script previously.
-level code incremental backup with the configuration that
was selected by -conf_dir. The code determines
the further action of the backup :
-create_configuration creates -conf_dir ,
records the other parameters of the backup
and finally starts a -level 0 run.
-update_configuration records new parameters
in an existing -conf_dir and starts -level 0.
-off disables the incremental mode.
-last chooses the level of the most recent run.
-next chooses the next unused level (-last+1).
0,1,2,... direct level choice by number.
By -last, -next or level numbers, many options other
than -conf_dir and -level are ignored and replaced
by those from -conf_dir/specific_arguments. A few
command line options will be able to override the
recorded ones :
-backup_cmd -composition_* -dir_size_cmd
-form_* -info_* -max_* -min_*
-prompt_script -scattering -status_file
-tmp_dir -shell_to_use -verify_source_filter
(I.e. you cannot change the backup area that way)
If created or updated by this program, specific_arguments
begins with lines equivalent to these options:
-working_directory ...address...
-load_old_tree content_file_list
vanished_file_list
-changetest_options type:size:inode:adr
-list_output_format type:size:inode:adr
-content_list_adr content_file_list.gz
-vanish_script_adr content_file_list
/tmp/VANISHED.sh.gz=VANISHED.sh.gz
-vanish_list_adr content_file_list
vanished_file_list.gz
If you do not want a vanish script in incremental
backups then you may disable this feature by :
option -disable_safe_leveling within your own
options to this program.
-changetest_options and -list_output_format are set
according to the start arguments. Above example assumes
the default settings.
-level_timestamp datestring_or_fileaddress
sets the own timestamp of an incremental backup run to
a point in time which is older than the program's start.
This timestamp will be later used by the next higher
level as criterion for -younger_than .
If datestring_or_fileaddress begins with '.' or '/' then
it is a file address of which the younger one of ctime
or mtime will be used. Else it is a time_definition as
expected by -younger_than .
Usually this setting is not needed. But if the incremental
backup is done from a filesystem snapshot then the
timestamp MUST be set to the TIMEPOINT OF THE SNAPSHOT.
It is harmless, if the timestamp is a bit too old.
-list_checksum recordlist searchname old_limit young_limit
the program ends immediately after listing those
records from recordlist which match the other
three arguments. The last one listed would be the
one used by -compare_checksum (see there).
The program will return 0 if any records were
listed. Otherwise its exit value is 1.
-list_output_format code:code:...
defines which information shall be written to the
lines resulting from -*_list_adr options.
Multiple codes are separated by ":".
type the first letter indicates the file type :
d = directory , l = softlink , b = block device
- = datafile , p = pipe , 0 = unknown
Extra information may be appended as a comma
separated list:
cZahl ... ctime in seconds since 1970
dZahl ... device number
iZahl ... inode number
tZahl ... mtime in seconds since 1970
pStart:Fsiz ... indicates splitfile part address
tells start byte and full size
PStart:Ssiz ... indicates original address
tells start byte and split part size
size bytecount (-1 = unknown)
crc CRC32 content checksum (-1 = unknown)
md5 MD5 (128 bit) rather than CRC32 (32 bit)
inode append inode-number to type. With directories
or absolute addresses the device number is
appended, too. E.g.: -,i10201 or d,d772,i4013
mtime append mtime to type. E.g.: -,t1151999973
ctime append ctime to type. E.g.: -,c1151999973
adr file address (always on, code ignored)
on_level_crc like crc but only if -level is not -off
on_level_md5 like md5 if -level is not -off
on_level_inode like inode if -level is not -off
The sequence is always type size crc adr . Eg.:
d 91286 A929B47C /home/thomas/
-load_old_tree old_list_adr vanish_list_adr
loads a complete file list or a layered set of
content and vanish lists in order to form a tree
of file objects. This tree represents an old
state of the backup area (typically from the time
when the next lower level's backup was made).
This tree is used by -younger_than and by
-vanish_list_adr to determine files which have
been moved or deleted since then.
If a -level is given and the addresses contain
no "/" then the files are read from the lower
level_# directories. Thus no action at level 0.
If old_list_adr points to a complete list, then
no vanish lists need to be read and therefore
vanish_list_adr may be any nonempty text.
If a list file itself is missing but exists
with name extension .gz then that file used. If
the effective filename ends with .gz it gets read
via a pipe through program gunzip.
CAUTION: a loaded tree might need much memory.
NOTE : old_list_adr from this option is identical
with old_list_adr from -vanish_list_adr. If both
options are given, the last one will finally set
this address.
NOTE: -load_old_tree with nonempty addresses
sets -changetest_options to "type:size:crc:adr"
-localdev (in effect only with NFS on SGI-IRIX systems)
exclude all remote filesystems.
Opposite of -alldev. See also -xdev
-main_dir text name of a directory which shall be mapped to the
root directory of the CD. Especially useful is
the text -auto . With this, the common directory
of the backup list (if any) is used. Text -off
disables this mapping (default).
-max_items number the maximum number of items in a result list
of directories and files appended to the
backup command. (default: 400)
-max_size number the maximum number of bytes per volume.
(default: 600m)
-max_size_stack size count
registers a number of volumes (i.e. media) with a
capacity different from -max_size. This may happen several
times. The volume group registered last will be used for
planning first. Watch the size announcement of the
volume to choose the appropriate media size. Example :
If you got 25 CDs with 700 MB and the rest stores 650 MB :
-max_size_stack 685m 25 -max_size 640m
-max_size_stack_save -on|off
It may happen that larger volumes cannot be filled with
more than -max_size bytes. In that case in default mode
"-on" the -max_size_stack volume is not counted as
consumed and will be retried with the next backup volume.
This can cause large media to be required surprisingly
late. In mode "-off" -max_size_stack volumes are
consumed one by one unconditionally.
-max_tree_depth number
If very many directories are nested then a stack overflow
can occur. An overflow is dangerous and thus is prevented
by a depth limit and a controlled program abort. The
default number is 128. It can be reduced to 16 or raised
to an arbitrary high number.
The stack size can be set under Linux by ulimit -s.
It should be at least 20 kB for each depth step.
-merge_trees source target dump options
The program ends immediately after this option has been
executed. It moves source and all file objects underneath
to target and analog addresses underneath. The move may
cross filesystem borders. target and the existing
directories underneath get the access permissions of
source and its subdirectories.
Files which get overwritten may be saved in a directory
tree underneath dump . If dump is a single "-" then
eventually existing files underneath of target get lost
when overwritten.
If prior to this option a directory list is given with
option -load_old_tree then the properties recorded there
will be copied to the target directories. (Such a list
may be generated via -dir_list_adr.) Without the list
directory properties will be copied from source.
There are the following options :
overwrite overwrite existing files
no_overwrite do not overwrite such files
do not change existing
directory permission
overwrite_older overwrite only if target file
is older than source file
move move files (delete source)
copy copy files (preserve source)
owner_plus_w give w-permission to owner in any case
chmod_original copy permissions unchanged
tree_prefix=<path> the prefix directory for loading
a directory list with -load_old_tree.
Only needed if <path> is not source.
tree_source with -load_old_tree read source part
of addresses.
tree_target with -load_old_tree read target part.
Options may be concatenated by a colon. The empty option
list "-" is the same as :
"overwrite:move:chmod_original:tree_target" .
-min_size_percent number percentage of -max_size which has to be
used of a volume before a new one may be opened.
If this percentage isn't reached then a large
directory may be split to fill the volume.
(default is 80.0)
-multiply_volume source repeat_faktor
the program ends immediately when stdin closes. It
reads source several times and puts out the copies to
stdout. Markers are inserted between the copies in order
to enable option -extract_multiplied to fetch single
copies out of the resulting data.
-no all following fileaddresses are to be omitted
from the result lists. This option can cause
the result lists to fragment since omitting a
address within a directory requires to put to
the list all other addresses within this
directory.
-not same as -no
-not_inode works like -no, but additionally tries to recognize the
excluded file objects by their device and inode numbers.
This test is more expensive and may put extra workload
on the disk.
-not_pattern shellpattern adds shellpattern to the list of search
expressions which get tested against file leafnames and in
case of a match cause the exclusion of the file.
shellpattern may use the wildcards *,?,[...].
-not_regex regex adds the regular expression regex to the list of search
expressions which get tested against file leafnames and in
case of a match cause the exclusion of the file.
regex is an expression as known from programs grep,
sed or vi.
-pacifier mode|rythm some options print intermediate messages about
their progress to stderr. If you are logging stderr
this may be annoying.
-off disables these message
-on enables a rythm of 1 message per second
A number sets the rythm to the a value in seconds
-timestamp prepends the messages by the number
of elapsed seconds.
-composition_memory displays the bufferstate of option
-composition_memory
-log_mode non-truncated messages in distinct lines
All modes besides -off are combinable. E.g.:
-pacifier 3 -pacifier -timestamp -pacifier -log_mode
-permute_blocks key source blocksize blockcount datastart
the program ends immediately when stdin closes. It
reads file source via random access and writes the
result to stdout. The output data are permuted blockwise.
The output data are preceded by a header part which tells
how the data have been swapped and moved :
scdbackup block permutation header v0.1
blocksize=... Number of bytes per block
complete_blocks=... Number of complete blocks
remainder=... Number of bytes in last block
permutation_key=... Permutation scheme (0 to 7)
datastart=... Number of first data block
key defines the permutation scheme :
0= blocks stay in sequence
1= swap both halves of the file
2= mirror sequence of blocks
3= swap 1st and 2nd quarter, sawp 3rd and 4th
plus 4 = additionally interlace 1st and 2nd half
(e.g. 7 = 3+4 = swap quarters, then interlace halves)
key "restore" will revert the permution and put out
the original file state.
If key is preceded by "-" then reversion is performed,
too. But in this case, the fileheader is skipped an the
parameters are taken from arguments : key, blocksize,
blockcount, and datastart.
When converting to permuted state, blockcount can be
used to set the size of a block device. Else the size is
taken from file size or by counting block device bytes.
Under normal circumstances, blockcount and datastart
should be 0, though.
-pre_checksum_size number sets the amount of data from the start
of the stream which is used to compute a preview
checksum. If this number greater than 0 then -filter_md5
and -filter_crc32 write two records. The first
with name appendix "_pre" may be used by option
-search_md5 and -search_crc32 to locate the
second one which describes the full data stream.
Number must be greater or equal -1. Default is 0.
If number is larger than 0 then
-filter_checksum_blocksize gets disabled.
-prevent_doubles on|off if set to "on" the emerging backup plan is
checked wether a newly appended item already is
and eventually the new item is discarded. This test may
need lots of time and is therefore disabled by default.
-problemlist_limit number
Set the maximum item number for eventual lists of
problem files from -filecheck, -exclude_size, -xdev.
If the number is exceeded with on of these lists then
the planning run will finally abort. (Default: 10000)
-process_blocklist action listfile skip source target
provides an overview over eventual data damages and also
recovery facilities. The program ends after this option
is done. It depends on block checksum lists.
See -filter_checksum_blocksize .
The block checksum list is read from listfile where it
has to be appended to a valid checksum tag. That may be
a backup volume written with -filter_checksum_blocksize
or a -blocklist_fileadr result.
The skip value is the number of tags to be skipped until
finally one gets read and used for processing. Normally
0 is ok.
source is a file address ("-"=stdin) of a potentially
damaged backup volume.
target will be a file address ("-"=stdin/stdout)
where undamaged blocks get assembled.
Depending on action both may have to be random access
files or may be allowed to be single pass byte streams.
The following keywords for action are defined:
"report" reads a stream from source and prints
-one per line- a list of block numbers to stdout. The
end of this bad block list is marked by the line "@".
target is not used.
"collect" reads a stream from source and writes the
undamaged blocks to target via random access.
"mend" reads source with random access and eventually
reads copies of damaged target blocks from source
by random access. If this succeeds then the undamaged
block copy gets written into target.
mend supports an optional start block number and an
even more optional end block number. Numbering starts
with 0. Keyword "end" depicts the last block.
If start block number is larger than end block number
then the blocks are processed in that reversed order.
Examples: mend mend:1200 mend:10k:20k mend:end:0
"merge" reads source and target sequentially an writes
either an undamaged block, or the damaged one from
target or even a block of zeros to stdout.
actions which write to target do also append the block
checksum list to target's payload data.
A short description of the list parameters and a
summary of the result are printed to stderr.
If stderr is to be interpreted by a program then you
should set option -pacifier -off before option
-process_blocklist .
The exit value is 0 if the resulting data are supposed to
be completely undamaged. It is non 0 else.
-result_filetypes typelist
defines what file types may appear in the result list
resp result script. typelist consists of one or more of
the following words separated by ":".
data ordinary data files
dir directories
symlink symbolic links
block_device device files of hard disks, CD, ...
unclassified anything different from above list
any everything together
arg_redirect redirect explicitely given symlinks
File types which are not mentioned in typelist get
discarded when they are to be added to the list of a
backup piece.
If pseudo-type arg_redirect is given, then the arguments
of this program get checked for symbolic links. Those are
replaced by their link targets (data sources).
-scan_for_checksum mode
controls -search_crc32 and -search_md5 scanning for
checksum tags (appended by -filter_checksum_padsize).
Following modes are defined :
-on Search enabled. Search properties unchanged.
-off Search disabled. Search properties unchanged.
-first Search on. End after first valid tag.
-whole Search on. Search up to end of byte stream.
-standalone Search on. Do not read recordlist.
-withlist Search on. Read and use recordlist.
-print Search on. List valid tags at end of run.
-noprint Search on. Do not list.
-anytag Accept any valid checksum tag.
-blocklist Accept only tags with block checksum list.
-pacifier Print progress messages to stderr.
-nopacifier Do not print progress messages.
-source:text Use file address standard input
-skip:zahl Number of bytes which got skipped at input.
With -source: this skipping is done after
opening the file address.
Combined modes are permitted. Default :
-scan_for_checksum -withlist -scan_for_checksum -off
-scan_for_checksum -anytag -scan_for_checksum -nopacifier
-scattering -on|-off mode -on allows moving large files to CDs
different than those which host their directory
siblings. This helps to fulfill -min_size_percent
but can lead to scattered directory contents. This option
is in effect only with -composition_mode default .
-search_crc32 recordlist searchname old_limit young_limit
like -compare_checksum but with automatic volume
recognition and only for records with CRC32
checksums. The search method depends on the
setting of -pre_checksum_size . If it is 0 then
all records are read, sorted by size and compared
with intermediate checksums of the data stream.
The stream is read up to its end and more than
one match may occur if streams with identical
data were recorded. With -pre_checksum_size -1
reading ends at the first match.
If -pre_checksum_size is >0 then at first a
matching preview record is searched to determine
the record with the full stream's checksum.
The preview checksums are computed by -filter_md5
and -filter_crc32 if pre_checksum_size was not
set to 0. The sizes of record and search have
to match in this case.
-search_md5 recordlist searchname old_limit young_limit
like -search_crc32 but with all records which
have MD5 checksums.
-set_shell_in_scdbackup_scripts shellname filename ... filename
a very special option which is used only for
configuration. The given files get tested wether
they start with "#!". If true, the rest of the
first line gets replaced by shellname. No other
option may follow since the rest of arguments is
used as a list of filenames.
-shell_to_use progname the progname is written after "#!"
into the first line of scripts generated by
this program. Default: /bin/sh
-special_file fileaddress size type
defines a file as a special data source. Currently only
type "block_device" is permitted. Additionally option
-split_large_files has to be given with a split_directory
and a split_command. The content of the block device will
be included in the backup in form of one or more parts.
If size is <=0, then the device is read up to EOF and
the resulting count is used. Fileaddress may be of form
target=source.
-split_file_prefix text
defines a text which is put in front of the names of file
pieces which emerge by -split_large_files.
-split_file_suffix text
like -split_file_prefix . The text is put between name
and counting numbers. See also -split_list_adr.
-split_large_files split_directory split_size split_command sc_option
sets the size limit for regular files together with a
directory address and a shell command which is able
to cut a piece out of a file. This program itself can
act as such a command with sc_option -cut_file . Argument
split_command is subject to special -bdir substitution.
Files which are larger than split_size are processed as
several separate parts. Shell scripts resulting from this
program will use split_command to temporarily store the
parts in split_directory. All partial files of a volume
have to fit into this directory. The directory has to
contain a file named scdbackup_split_file_directory_mark .
If split_directory is '-' then this feature is disabled.
split_directory '-xorriso_cut_out' uses option -cut_out
of ISO 9660 formatter program xorriso. In this case no
directory is needed, split_command and sc_option get
ignored then.
If split_size is '-' then the effective -max_size - 1m
is used. Keep in mind that -exclude_size can prevent
large files from being split and processed at all.
If split_command is '-' then the previously set command
stays in effect. Use '-' as sc_option in this case.
-split_list_adr fileaddress
writes a list of file parts caused by -split_large_files.
The information in this file may be helpful if a script
output format is chosen which does not contain target
address (e.g. -form_path_list_afio). When restoring such
a backup, the file parts will appear in split_directory
and need to be put to their original positions.
General list format, working directory, compression and
backup inclusion are as described at -complete_list_adr.
Each data line describes a file part by the form
partprefix part_address=buffer_address
partprefix has the form
**partnumber/total:start_byte+byte_count**
part_address consists of
original_dir"/"marked_name"_"partno"_"total
marked_name consists of
prefix_text""original_name""suffix_text
buffer_address is composed of
split_dir"/"basename_of_part_address"-"hexcount
where hexcount is unique for the whole backup.
-status_file filename If the filename is not empty and not "-"
then the result script will write the current
volume number into this file before it calls
the -backup_cmd . Also the script will be able
to resume at the start of an interrupted volume
when given the argument "AUTO" rather than
a volume number.
-tmp_dir directory_address
sets a directory for storing temporary files. One of its
uses is with list file addresses without '/' which
would go to the level directories if there was any -level
set (see -complete_list_adr). directory_address is
subject to special -bdir substitution.
-target_escape mode controls the handling of "="-characters in target
addresses. mode "on" causes "=" and "\" to be
preceeded by an additional "\"-character. Such target
addresses are recognized by modern versions of mkisofs
which know option -graft-points.
mode "off" exchanges in target addresses each "="
by "_". "\" is not changed. Use with mkisofs 1.12b3.
Note: Option -form_path_list sets mode to "on"
-form_script sets it to "off". Afterwards it may
be set explicitely by -target_escape.
-target_separator text the first character in text is used as
separating character between target address
and file address. This does only apply to input but
not to output where eventually "=" is used.
An empty text disables explicit target addresses.
Default setting is "=". Within the target address
eventual "=" and "\" characters have to be escaped
by preceeding "\" characters.
Example: map /home/office to the / directory of the CD:
-target_separator ':' /:/home/office -target_separator ''
Example: map /home/test/=\ to /=\
-target_separator '=' '/\=\\'='/home/test/=\'
-tree_hashsize number option for balancing speed of file search in the
directories of old_tree against the memory consumption of
-old_tree. Each directory in old_tree occupies by its
hashvector number*sizeof(int) bytes. The search in large
directories is accelerated by at about the factor number.
The default of 20 is well suited for normal sized
directories of 20 to 1000 files. Directories with several
ten thousand files may slow down the program substantially
whereas trees with very many directories may consume too
much memory.
-vanish_list_adr old_list_adr result_address
checks the file objects which have been recorded
some time ago in old_list_adr. Those objects
which have vanished since then are reported to
file result_address. If a -level is given and
result_address contains no "/" then the file is
created in the level_# directory. (See -conf_dir)
If result_address ends with .gz then the output
gets piped through program gzip.
If a target is given in result_address then the
resulting file will be appended to the last
backup volume. (e.g. /vanished.gz=vanished.gz)
old_list_adr may be output of -complete_list_adr
Alternatively a plain list of file addresses may
be used.
If a tree gets loaded by -load_old_tree then it
is used for comparison. If there is no old tree
then the old_list_adr gets read directly but it
has to be a complete list in a single file.
NOTE : old_list_adr from this option is identical
with old_list_adr from -load_old_tree. If both
options are given, the last one will finally set
the effective address.
CAUTION: If -load_old_tree adresses are empty or
-changetest_options is "off" then the result
list may contain old file adresses that have been
moved but the backup plan will not contain the
new addresses. Best is to also test for CRCs
(see -changetest_options)
-vanish_script_adr old_content_list_adr result_address
CAUTION: Running the result script is dangerous !
Especially if generated without CRC test
Same as -vanish_list_adr but reports shell
commands to form a script that removes all
vanished objects. Such a removal is necessary
before applying an incremental level during
re-install of a multilevel backup. But this
script does not check against cruel accidents.
NOTE: the script contains target addresses only.
-verify_filelist list_adr directory_prefix
NOTE: If you want to verify the backuped files on CD or
during restore then you should better use
-verify_targetlist.
The program ends immediately after this option is
done. It reads file addresses from list_adr
together with eventual informations about type,
size and checksum. The informations get compared
with the current state of the file. Its address
is composed from directory_prefix and the address
read from the list.
If the addressed file exists and if the other
properties match, then the file's address is
printed on stdout. Else there will be an error
message at stderr.
If everything existed and matched, the program
exits with value 0. Else this value is 1.
Currently this is mainly a harmless version of
-delete_filelist . But that might change.
-verify_sources listfile
the program ends immediately after this option is
done. This option is similar to -verify_filelist
but it expects addresses of form target=source
and only checks the source parts. No directory
prefix gets applied.
-verify_source_filter shell_filter
With output format -form_path_list_mkisofs the
filelist can be cleaned from undesired addresses
immediately before the prompt to insert a CD.
The decision is made by shell_filter which has to
put out any valid file to stdout and omit any
invalid file. Use this filter in order to omit
all files which have been removed since the
backup was planned :
..../cd_backup_planer -verify_sources -
An empty text as shell_filter disables filtering.
shell_filter is subject to special -bdir substitution.
-verify_targetlist list_adr directory_prefix
like -verify_filelist but processing the target addresses
of the listed files. If there have been target=source
redirections when a backup was made, then this can make
a big difference.
-volume_info pass additional arguments to the backup command
(before file names) : -v volume -t total_volumes
-volume_offset number
Adds number to all output of volume numbers and subtracts
it from input of volume numbers.
-working_directory directory_address
set the working directory to the given address.
This option should be used only once and only
before any fileaddress is given.
-write_random target amount randomseed
the program ends immediately after this option is done.
It writes into the target file ("-" = stdout) the given
amount of random bytes or a periodical byte pattern of
up to 16 bytes. If randomseed begins with "=" then the
rest of it is the periodical pattern. If the "=" is
followed by "~" then that pattern is inverted into its
binary complement. Without "=", randomseed is the key
for a pseudo random number generator based on MD5 and
EXOR. If this key starts with "-" then it is used for a
slow method which is free of EXOR. The number sequence of
a particular randomseed is reproducible. Large amounts
like 100g or 2t are permitted if the OS can handle them.
-write_to_file targetfile byteaddress bytecount bytecode [... bytecode]
-write_to_file targetfile byteaddress "string" piece_of_text
-write_to_file targetfile byteaddress "file" sourcefile
the program ends immediately after this option is done.
It opens the existing targetfile for writing and positions
the write pointer at byteaddress (file start = 0).
If the third argument is a number between 1 and 4096 then
it is the count of the following bytecode arguments.
Permissible are decimal bytecodes between 0 and 255.
With "string" the forth argument is written into
targetfile. With "file" the forth argument gives the
path of a file of which the content will be copied into
targetfile.
targetfile "-" means stdout. It is only permitted with
byteaddress 0. sourcefile "-" means stdin.
-xdev exclude all filesystems mounted below the items
of the backuplist. Opposite of -alldev .
-yes all following fileaddresses are appended to the
backuplist (i.e. these files are to be backuped).
This option ends the range of option "-no".
It is the default mode for fileaddresses.
If the option -yes is given at least once, then
the option -no does not apply to the backuplist.
( -yes x -not x : file x is in backup
x -not x : file x is not in backup)
-younger_than time_definition try to backup only files where the
content or inode changed since the given time.
Moves of a file within the same file system
cannot be detected, regrettably.
Time definitions can have the following forms:
Absolute timepoints can be defined by the input
format MMDDhhmm[[CC]YY][.ss]] like with UNIX
command 'date'. For example:
020616122000 is 6 Feb 2000 16:12:00 local time.
Also the format YYMMDD[.hhmm[ss]] is recognized:
11 May 2001 23:45 = A10511.2345
1 Mar 2010 = B00301 , 14 Jul 1998 = 980714
Relative times are preceded by '-' or '+'. They
are added to the current time. '-' therefore
points to the past, '+' points to the future. The
basic unit is the second. It can be modified by
a letter at the end of the number:
seconds 1s=1 , hours 1h=3600s , days 1d=24h ,
weeks 1w=7d , months 1m=31d ,
years y : Xy=X*365.25d+1d
Please note : 'm' and 'y' are possibly slightly
larger than the actual calendar time range.
Example:
-3w is 3 weeks in the past = -1814400s .
Examples:
Divide the content of the working directory into volumes of
625 MB and maximum 200 independend items. Create info script ASKME
in working directory and include it in each backup part.
bin/cd_backup_planer \
-max_items 200 -max_size 625m -info_script ASKME . >bscript
chmod u+x bscript
To actually perfom the backup, execute :
bscript
With this option the ASKME script is generated at a separate place
and mapped to the root directory :
bin/cd_backup_planer \
-info_script /ASKME=/home/backup_planer_dir/ASKME
Divide the directores "office" and "images" into volumes of
default size. Do not include subdirectories "office/private" and
"images/girlfriends"
Replace the omitted "images/girlfriends" by "images/cats"
bin/cd_backup_planer \
office images -target_separator = /images/girlfriends=images/cats \
-not office/private images/girlfriends \
>bskript
Fill DC-6250 tapes with compressed tar archives.
bin/cd_backup_planer \
-backup_cmd 'read dummy ; tar cvzf /dev/rmt0' \
-form_script_raw -max_size 245m -max_items 100 \
-info_script ASKME \
. \
>bskript
Create a configuration for incremental backups and perform planning
of the first -level 0 run. By -level -create_configuration a new
directory gets created which will be called $HOME/my_backup_conf .
bin/cd_backup_planer \
-conf_dir $HOME/my_backup_conf -level -create_configuration \
... all arguments defining the full backup ... \
>bskript
The configuration may be used by simply giving this directory and
the desired level. All other parameters of the backup will be read
from $HOME/my_backup_conf/specific_arguments (1 line = 1 argument).
One should use a fixed backup scheme. Let us assume that we made
the configuration and the first -level 0 backup at sunday of week 0.
A backup scheme which from monday to saturday records only the
daily changes and makes a full backup each sunday.
Monday : -conf_dir $HOME/my_backup_conf -level -next
...
Saturday: -conf_dir $HOME/my_backup_conf -level -next
Sunday : -conf_dir $HOME/my_backup_conf -level 0
Now the media of monday to saturday may be recycled.
A scheme which makes quarterly (13 weeks) full backups, four-weekly
and weekly updates (-conf_dir same as above). This scheme lets you
retrieve old file versions for quite a long time.
Week 1 : -level 1
Week 2 : -level 2
Week 3 : -level 2
Week 4 : -level 2
Week 5 : -level 1
Week 6 : -level 2
Week 7 : -level 2
Week 8 : -level 2
Week 9 : -level 1
Week 10 : -level 2
Week 11 : -level 2
Week 12 : -level 2
Week 13 : -level 0
For week 14 hop to week 1.
The backup of week 12 also covers weeks 11 and 10.
The backup of week 9 also covers week 8 to 1.
It is advised to keep levels 0 and 1 for a longer time.
To combine both schemes shown, simply replace the sunday run of
scheme 1 by the weekly run of scheme 2.
If you ever need to re-install an incremental backup with multiple
levels you will begin with copying level 0 first. Then you need to
remove the files which have vanished in the week between the level 0
backup and the next level.
For that purpose you will find on the last backup volume of level 1
a file named /tmp/VANISHED.sh.gz . It does not have x-permission to
avoid ugly accidents. You can only trust it if it has been produced
reading a content list with CRCs or inode numbers. This is true by
default but may be changed with options like -content_list_adr ,
-list_output_format , -load_old_tree , -changetest_options ,
-vanish_script_adr .
If you want to be very cautious, you may use option "md5" rather
than "crc" with -list_output_format.
You will have to apply chmod u+x and gunzip after copying or you
will have to execute it on-the-fly via a pipe like :
gunzip </cdrom/tmp/VANISHED.sh.gz | /bin/sh
Better have a look at its content to decide wether you think the
condemned files are really obsolete. Last chance to save old trash.
The vanish script of level 1 may be executed at any time after
level 0 was installed and before level 2 gets installed. None of
the condemned files is contained in the level 1 backup.
If disk space is short, better execute it before installing level 1.
Otherwise it is more convenient to execute /tmp/VANISHED.sh after
it was copied to the disk.
When installing the following levels 2,3,4,... do not forget to
execute their vanish scripts in time and to remove the last one
from the /tmp directory after all work is done.