dump(8)dump(8)NAME
dump, rdump - Performs incremental file system dumps
SYNOPSIS
/usr/sbin/dump [key[argument...] file system]
/usr/sbin/rdump -fdump_file [otherkey[argument...]file system]
rdumpOPTIONS
Specifies the dump level. All files modified since the last time-stamp
whose names are currently stored in the /etc/dumpdates file for a named
file system at levels less than the one specified are dumped to tape.
When no time-stamp entry is defined for a dump level, the Epoch is
assumed; thus, the value -0 for this key causes the entire file system
to be dumped to the storage medium. Specifies the number of blocks in
1024 bytes to write to the storage medium. The default write block
value is 10 blocks (or 10 kilobytes). In many instances, increasing
the number of blocks per write will increase your dump performance.
For optimal results on AlphaServers, use a value that is greater than
or equal to 32, is a multiple of 4, and is less than 65. The dump
medium is a not a 9-track cartridge tape. Specifies the write density
of the storage medium. The density operand is expressed in bits per
inch (bpi). This information is used in calculating the amount of
medium used per each volume of the storage medium. For the dump com‐
mand, the density will be automatically provided by supported devices.
Otherwise, the default write density for the dump and rdump commands is
1600 bpi. Writes the dump to the dump_file storage device instead of
the default tape drive. When the name of dump_file is - (dash), the
dump process writes to standard output. When the name of the
dump_file is /dev/ntape/tape?*, the dump process will enable the -N
option. When the rdump command is invoked, the dump_file operand must
specify both the remote machine and the storage device in the format
[user@]machine:device, where where user is an optional user identifica‐
tion (account) used to logon to the machine. If you do not specify
user rdump will use the user identification from the current process.
The machine is the name or reference designation of the host machine,
and device is the name or reference designation of the storage device.
If you want to specify an IPv6 address for machine, you must prefix the
address with the \[ (backslash, left bracket) characters and terminate
the address with the \] (backslash, right bracket) characters. Because
the bracket characters are shell metacharacters, your must precede them
with the backslash character. Specifies the label of the dump to be
displayed during restoration. The default label is "No Label". The
label can be a maximum of 16 characters. Notifies, by means of a com‐
mand similar to wall(1), all operators in the group named operator,
which is specified in the /etc/group file whenever dump or rdump
requires operator attention (to change a tape, for example). Specifies
the size of a dump tape. The size operand is expressed in feet. When
the amount of tape specified by size has been written, either process
waits for the current reel to be changed (see the -n option). For the
dump command, the tape size will be automatically provided by supported
devices. Otherwise, the default tape size for the dump and rdump com‐
mands is 2300 feet. Writes the time of the beginning of the dump as
the time-stamp entry in the /etc/dumpdates file for the file system
record when the dump successfully completes. Tells an operator what
file systems must be dumped to the storage device. This information is
obtained from the /etc/dumpdates and /etc/fstab files. The -w key
tells either process to print to the standard output a record for each
file system listed in the /etc/dumpdates file. Specifies a block-mode
device. For the dump command, the estimated calculations will be based
on the device's storage capacity instead of density and size. Prints
the estimated size of the dump file in 1-kilobyte blocks and the esti‐
mated number of volumes that make up the dump file, only. The dump file
will be opened and closed, but nothing will be written to it. The
information will be output through standard error and will have a for‐
mat like the following: 23382 blocks, 0.04 volumes
If the dump file is on disk, the format will be: 23382 blocks
The operand placement will be preserved, but the exact words
"blocks," and "volumes" are not guaranteed. Refer to the EXAM‐
PLES section for further information. Disables the rewinding of
the tape and placing the tape unit off line after completing the
dump session. By default, when the dump command finishes back‐
ing up a file system it rewinds the tape and takes it off line.
For some tape subsystems, this tape will be ejected from the
unit. The -N option is the default when the dump_file operand
is /dev/ntape/tape?. If you use the -N option to dump to a regu‐
lar file that does not have the letters "rm" in its name, dump
will inform you of your error in using -N and terminate. Speci‐
fies output file size in feet. When the -B option is used, the
full_tape_size operand specifies the output file size in number
of 1024-byte blocks. Specifies a tape number, which is used in
the dialog with the operator as the number of the first tape.
Prevents any extended attributes from being archived with asso‐
ciated files. Similar to -w, but for any file system listed in
the /etc/dumpdates file, prints an output record and highlights
this record with the > (greater than) character, all files that
must be dumped. When -W is specified, all other options are
ignored and dump exits immediately.
DESCRIPTION
The dump command copies to the default /dev/tape/tape0_d1, or to the
alternate storage device specified with the -f option, all files and
any associated extended attributes (including ACLs, see the acl(4) and
proplist(4) reference pages) changed after a certain date in the speci‐
fied local file system.
The rdump command copies to the dump_file storage device all files and
any associated extended attributes (including ACLs, see the acl(4) and
proplist(4) reference pages) changed after a certain date in the speci‐
fied file system.
These commands cannot be used to archive AdvFS filesets. See vdump(8)
for the operations used to archive AdvFS filesets.
The dump and rdump commands are used to dump local files and any asso‐
ciated local attributes from a single file system defined by the file
system operand to a local or remote storage device, respectively, where
file system contains the files you want to back up.
The dump and rdump commands perform similar functions with respect to
storage of files contained in the named file system. However, the rdump
command requires that the -f option be used with any otherkey and the
special dump_file operand.
Both commands copy all files in file system whose dump level is less
than a specified value, and that have changed after a specified date to
the default storage device or to an alternate storage device. The dump
level and date are specified in the local /etc/dumpdates file. The key
and argument operand specify one or more options that may be used to
write files to the storage medium. Characters permitted by the key op‐
erand are similar to options that consist of any of the characters
0123456789bcdfnsuwBNSTW only, which may be used in any logical combina‐
tion, but must be preceded with the - (dash) character; the argument
operand specifies other options that tell these dump and rdump pro‐
cesses what to do. These options are described in the OPTIONS section.
Not all keys permit argument options to be specified.
The /etc/dumpdates file consists of 3-column record lines that specify
the file system name, a dump level, and a standard time-stamp. These
processes enter a time-stamp into the file system record after each
file in the named file system is successfully backed up. The 3-column
record in the /etc/dumpdates file contains the following information:
Lists the file system device name. This is an integer between 1 and 9
that defines a hierarchy for files in file system. This hierarchy
indicates which files should be written to the storage medium when the
dump or rdump command is executed. Level 0 defines all the files in
file system. When a level is assigned, all files equal to and less
than that level in file system are backed up. The time-stamp tells the
dump or rdump process when file system had its last backup. This time-
stamp is written by the dump or rdump process after the specified file
system backup is completed. When there is no time-stamp, the dump or
rdump process assumes the beginning of time (called the Epoch).
The /etc/dumpdates file is written in ASCII and consists of a single
record per line. This file may be edited to change any record field,
when necessary.
Either process requires operator intervention when any of the following
conditions occur: end-of-tape, end-of-dump, tape-write error, tape-open
error, or when the number of disk-read errors is greater than 32. In
addition to alerting all operators specified by the -n key, these pro‐
cesses interact with an operator at the terminal from which dump or
rdump was invoked when either program can no longer proceed.
All queries written to standard output by the dump or rdump process
must be answered by typing yes or no on the invoking terminal.
Because a dump to any storage medium requires excessive time to
process, each process checks itself at the start of each storage vol‐
ume. In many instances, the default dump performance can be improved
by modifying the number of blocks per write. For additional informa‐
tion, see the description for the -b switch in the OPTIONS section.
When a volume write fails, dump or rdump restarts itself from the last
successful checkpoint, with operator permission, after the currently
written storage medium is properly removed and another (replacement)
storage medium has been mounted.
These processes also tell an operator what is going on at periodic
intervals when writing to the storage medium. This information con‐
sists of somewhat conservative estimates for the number of blocks to
write, the number of storage media that must be used for the dump, the
time to complete the dump, and the time until the storage medium must
be replaced with another one to complete the dump. Output is verbose,
so that others know that the terminal controlling dump is busy. When
processing takes place, the following conditions apply:
Fewer than 32 read errors during a dump or rdump tape-dump process are
ignored. Each renewal of the storage medium requires a new dump
process, so that parent processes for storage media already written are
in effect until the entire storage medium is written. When the dump
command has the W or w key set, no records are written to the standard
output for a file system that has no current record in the /etc/dump‐
dates file, even when listed in the /etc/fstab file. When no argument
is specified, the key operand is assumed to be -9u so that the default
file system is dumped to a default storage medium named
/dev/tape/tape0_d1 (usually a tape).
dump
The dump command copies to the default /dev/tape/tape0_d1, or to the
alternate storage device specified with the -f option, all files
changed after a certain date in the specified local file system.
rdump
The rdump command copies to the dump_file storage device all files
changed after a certain date in the specified file system. This com‐
mand is similar in operation to dump, except that the -f option is
always specified (see the OPTIONS section) together with any otherkey
you may wish to specify. The dump_file operand should always be speci‐
fied by machine name and device name as machine:device name. If you
want to specify an IPv6 address for machine, you must prefix the
address with the \[ (backslash, left bracket) characters and terminate
the address with the \] (backslash, right bracket) characters.
The rdump command starts remote server /usr/sbin/rmt or /etc/rmt on the
client machine to access the storage medium.
Another vendor's rdump command may fail because rmt is not located in
/etc. To avoid this problem, it may be necessary to provide a symbolic
link on the machine pointing to /usr/sbin/rmt, as shown in the follow‐
ing example: ln -s /usr/sbin/rmt /etc/rmt
Although the rdump command can detect magnetic tape on remote ULTRIX
and Tru64 UNIX systems, it cannot detect magnetic tape on other remote
systems. Instead, it defaults to treating the output medium as a disk
file and aborts the operation should it encounter overflow or I/O error
cases.
Remote systems must be able to run the uname command if you are to use
the rdump command. If a remote system cannot run the uname command, you
can set the environment variable OSF_RDUMP_SIMP_RCMD before you attempt
the operation.
NOTES
Estimates for dump and rdump.
The information in this note is specific to supported tape
devices and the densities and sizes they use under rdump. The
size and density information is used to estimate the number of
volumes required for the current dump or rdump request. Some of
the factors that will effect the estimate calculation include
the following: Track format Compression Interrecord gap size
Writing optimization technologies Appending to a tape
The estimate calculation does not take these factors into
account and can result in very large (100-500%) miscalculations.
The estimates can be customized by adjusting the size in feet
(-s) or size in 1024-byte block (-BS) variables to the desired
results. The default values for /dev/tape/tape?_d0 devices used
in the dump estimate calculation are shown in the following ta‐
ble:
Tape Device Density Size Size
(bpi) (feet) (1024-byte
blocks)
TA78 6250 1925 141056
TA79 6250 1925 141056
TA81 6250 1925 141056
TA90 38000 436 194560
TA91 38000 436 194560
TF30 6667 1182 92416
TF70 10000 2457 287948
TF70L 10000 2457 287948
TK50 6667 1182 92416
TK70 10000 2457 287948
TKZ09 54000 7456 4718592
TLZ04 61000 1584 1132646
TLZ06 61000 2640 1887436
TLZ07 61000 2640 1887436
TS05 1600 2075 38912
TU77 1600 2075 38912
TU78 1600 2075 38912
TU80 1600 2075 38912
TU81 1600 2075 38912
TU81E 1600 2075 38912
TZ05 1600 2075 38912
TZ07 1600 2075 38912
TZ30 6667 1182 92416
TZ85 42500 4925 2453299
TZ857 42500 4925 2453299
TZK08 54000 3276 2073600
TZK10 16000 2580 483840
The rdump command starts the remote server /usr/sbin/rmt on the
client machine to access the storage medium. If the rdump com‐
mand cannot find /usr/sbin/rmt, it will try /etc/rmt and rmt.
The rdump program can detect remote tape support on Tru64 UNIX
and ULTRIX systems. However, due to the lack of a standard for
UNIX magnetic tape functions, it cannot utilize remote tape sup‐
port on other systems. This means that multivolume dumpsets can
only be created when the remote system is Tru64 UNIX, DEC OSF/1
(the former name of Tru64 UNIX), or ULTRIX, or if there is
embedded multivolume support in the remote system (such as is
the case with VMS, where support is in the Magtape ACP). For
rare cases where the remote system is non-UNIX, compatibility
may require that rdump not use UNIX-like commands. In order to
obtain this behavior, the user or system manager should use the
following command: setenv OSF_RDUMP_SIMP_RCMD
The previous command can be used on a system wide (global) or
per rdump command basis. For proper operation, the server's
/.rhosts file must contain the name or reference designation of
the client's machine. The rdump and the dump commands do not
handle MFS or AdvFS filesystems. After encountering tape write
errors, dump or rdump queries the operator about performing a
rewrite. If the operator requests a rewrite, a rewind is per‐
formed, followed by an attempt to rewrite the data. In the event
the no-rewind device is used, the user should always load a new
tape to avoid the possibility of overwriting previously written
archives.
EXIT STATUS
The dump and rdump commands exit with 0 status on success. Startup
errors are indicated with an exit code of 1; abnormal termination is
indicated with an exit code of 3.
EXAMPLES
To perform a full level 0 dump, enter a command similar to the follow‐
ing: dump -0un -f /dev/tape/tape1_d0 -b 32 /fs1
In this example, 0 specifies that all files in the file system
fs1 will be dumped to /dev/tape/tape1_d0; u specifies dump to
update the /etc/dumpdates file after a successful dump of the
file system; and n specifies that operators will be notified.
The estimate calculation will be based upon the tape device
defaults. The write block size is set to 32 kilobytes. To dump
a local file system to a remote storage tape, enter a command
similar to the following: rdump-3u-f
tape_server:/dev/tape/tape1_d0 /fs1
In this example, 3 specifies the dump level of all files in the
file system /fs1 that will be dumped to tape /dev/tape/tape1_d0
on system tape_server, and u specifies the dump to update the
/etc/dumpdates file after a successful dump of the file system.
The estimated calculation will be based on the rdump defaults.
The write block size will be the default. To obtain an esti‐
mated number of blocks and volumes for the current dump session,
enter the following: dump -0Ef /dev/tape/tape1_d0 /usr
The system displays output similar to the following: 358696
blocks, 0.19 volumes The following examples show alternative
ways of scheduling backups. Select the backup schedule that best
fits your needs. 7-day incremental schedule
The following schedule is a 7-day incremental schedule. This
schedule is useful for small-to-medium storage systems. The
basic algorithm is n, n+1, ... The number of dump files to per‐
form a full restore can vary from 1 to 7. The following is an
example of a 28-day schedule: 0 1 2 3 4 5 6
0 1 2 3 4 5 6
0 1 2 3 4 5 6
0 1 2 3 4 5 6 7-day alternative differential
schedule
The following schedule is a 7-day alternative differential
schedule. This schedule is useful for small-to-medium storage
systems. The basic algorithm is n-1, n+3, ... The number of
dump files to perform a full restore can vary from 1 to 4. The
following is an example of a 28-day schedule:
0 5 4 7 6 9 8
0 5 4 7 6 9 8
0 5 4 7 6 9 8
0 5 4 7 6 9 8 28-day alternative differential
schedule
The following schedule is a 28-day alternative differential
schedule. This schedule is useful for small-to-large storage
systems. The algorithm combines the 7-day incremental and the
7-day alternative differential schedules. The number of dump
files to perform a full restore can vary from 1 to 7. The fol‐
lowing is an example of a 28-day schedule:
0 5 4 7 6 9 8
1 5 4 7 6 9 8
2 5 4 7 6 9 8
3 5 4 7 6 9 8
This schedule limits full dumps to once a month. Therefore, it
is possible to lose a month's worth of data. Alternative
approaches to address this problem might include duplicating the
full tape or doing full backups twice a month instead of once a
month.
FILES
The dump command path in single user mode. The dump command path in
multiuser mode. Used by the rdump remote tape access program. Con‐
tains a list of file systems that were backed up, the date that each
file system was backed up, and the backup level.
SEE ALSO
Commands: restore(8), rrestore(8), rmt(8)
Files: acl(4), proplist(4)dump(8)