source: gpfs_3.1_ker2.6.20/lpp/mmfs/samples/util/README @ 152

Disclaimers
-----------

The files in this directory are provided by IBM on an "AS IS" basis
without warranty of any kind.


Introduction
------------

The files in this directory contain generally useful utilities that use
GPFS specific APIs to scan filesystem metadata.  The source code is provided
to give examples of how to use some of the APIs.  Execution of these commands
requires root authority.


Utilities
---------
tsgetusage   - Fast scan of all files in a filesystem and calculate how
               much disk space is used by each unique user and/or group.

tsinode      - Fast scan of all files in a filesystem and display one line
               of ls information about each one.

tslistall    - Fast scan of all files in a filesystem and display one line
               of inodenumber pathname type ctime mtime

tsfindinode  - Fast parallel scan of all directories in a filesystem
               listing inodenumber type pathname for each inode specified

tsreaddir    - Fast parallel scan of all directories in a filesystem
               listing inodenumber type pathname for each file
               (used by tslistall)

tstimes      - Fast scan of all files in a filesystem and display one line
               of inode mtime ctime for each file
               (used by tslistall)

tsbackup     - Fast parallel backup utility using multiple TSM clients
               and a single TSM server.  Parallelism is achieved via
               the use of multi-threading and the use of multiple TSM
               client processes.  This utility makes use of the GPFS
               inode scan APIs.


------------------------------------------------------------------------------

tsgetusage   - Fast scan of all files in a filesystem and calculate how
               much disk space is used by each unique user and/or group.

  Usage:  tsgetusage [-g][-h][-q][-v][-u][-H hashTableSize] fsPath

          where:  fsPath  path to file system
                  -u      outputs data in diskusg format for acctdisk
                  -g      collect stats for groups
                  -H      sets the hash table size (default 1027)
                  -h      prints this message
                  -q      outputs only the resulting stats
                  -v      print information about each file

    Using the -u option will produce a file with a format like
    /usr/sbin/acct/diskusg that is used for other types of filesystems.
    This can be used as input to the /usr/sbin/acct/acctdisk program, and then
    merged with the other disk accounting data using /usr/sbin/acct/acctmerg.

    For more information about the AIX Accounting System, the preparation
    of daily and monthly reports, and the accounting files, see the AIX
    Accounting Overview.


------------------------------------------------------------------------------

tsinode      - Fast scan of all files in a filesystem and display one line
               of information about each one.

  Usage:  tsinode fsPath [startinode [endinode]]

          where:  fsPath      path to file system
                  startinode  first inode number to be listed
                  endinode    last inode number to be listed

  Hint:
    For the best performance on a filesystem with millions of inodes,
    run several copies of tsinode with different inode ranges and
    concatenate the output files.


------------------------------------------------------------------------------

tsfindinode  - Fast parallel scan of all directories in a filesystem
               listing: inodenumber pathname

  Usage:  tsfindinode {-i {inum | inumfile}} [-t nThreads] [--] fsPath

          where:  fsPath       path to file system
                  -i inum      adds one inode number to the list
                               (this can be used multiple times)
                  -i inumfile  names a file containing inode numbers
                               to be added to the list
                               (this can be used multiple times)
                  -t           set the number of threads (default = 16)
                  --           marker so that names starting with "-"
                               are not treated like options.

This program recursively traverses the directory structure to list all
accessible files and directories in the file system whose inode numbers
match those specified in the list of inodes.  Each file prints one line
on stdout containing the file's inode number, and the pathname to the
file.  The number and names are separated by a tab character.
The starting directory is listed with the pathname given as the argument.
All other lines have pathnames relative to the starting directory.

The program uses the GPFS directory scan interface to quickly read the
directory entries without calling "stat" on each file.  The directories
are scanned in parallel by multiple threads which results in an unsorted
output file.  Inode numbers that are not found in the tree are printed to
stderr with the name "(notfound)".

Example:

>tsfindinode -i 23612 -i 1452 -i 47666 /gpfs
     23612  include/stdio.h
      1452  include/curses.h
     47666  (notfound)
>


------------------------------------------------------------------------------

tslistall    - Fast scan of all files in a filesystem and display one line
               per file with the following information:
                 inodenumber/generation mtime ctime type pathname

  Usage:  tslistall fsPath  [tmpdirname]

          where:  fsPath      path to file system
                  tmpdirname  is a directory where intermediate files for sort
                              can be generated. The default directory id /tmp.
                              This is necessary for GPFS filesystem that have
                              large numbers of files.

This is a shell script that invokes tsreaddir and tstimes
then merges the resulting files.  The output is printed to stdout
with 1 line per file and is sorted by inode number.

By saving the output of this command, you can compare it with a later list
to find out what has changed in the filesystem since the old list was made.

Example:

>tslistall /gpfs
0000000003/0000000001 1012390401.913049600  1012390401.913049600  DIR /gpfs
0000001452/0000000003 1024702059.694025000  1024702059.694025000  LNK include/curses.h
0000001812/0000000000 1024702015.294047000  1024702015.294047000  DIR include
0000019578/0000000000 1026797433.109149000  1026794560.584238000  REG mail.rc
0000023612/0000000017 1024702090.705801000  1024702090.705801000  REG include/stdio.h
0000023874/0000000000 1024702024.982645000  1024702024.982645000  REG include/string.h
>


------------------------------------------------------------------------------

tsreaddir    - Fast parallel scan of all directories in a filesystem
               listing inodenumber type pathname

  Usage:  tsreaddir [-d] [-n] [-t nThreads] [--] fsPath

          where:  fsPath  path to file system
                  -d      list only directories.
                  -n      do not list directories.
                  -t      set the number of threads (default = 16)
                  --      marker so that names starting with "-"
                          are not treated like options.

This program recursively traverses the directory structure
to list all accessible files and directories in the file system.
Each file prints one line on stdout containing the file's inode number,
the generation number, the file's type and the pathname
to the file.  The fields are separated by tab characters.
If directories are listed, the starting directory is listed with
the pathname given as the argument.  All other lines have pathnames
relative to the starting directory.

The file types are "DIR" for directories, "REG" for regular
files, "LNK" for symbolic links and "UNK" for all others.
If the -d option is specified only directories are listed,
and if the -n option is specified directories are not included in
the output list.

The program uses the GPFS directory scan interface to quickly read the
directory entries without calling "stat" on each file.  The directories
are scanned in parallel by multiple threads which results in an unsorted
output file.

Example:

>tsreaddir /gpfs
0000000003/0000000001 DIR /gpfs
0000001812/0000000000 DIR include
0000023612/0000000017 REG include/stdio.h
0000023874/0000000000 REG include/string.h
0000001452/0000000003 LNK include/curses.h
0000019578/0000000000 REG mail.rc
>


------------------------------------------------------------------------------

tstimes      - Fast scan of all files in a filesystem and display one line
               of inode mtime ctime for each file

  Usage:  tstimes fsPath

          where:  fsPath  path to file system

This program performs a gpfs inode scan on the specified file system
to obtain the mtime and ctime for each file.  The results are output
on stdout, with one line per file containg the inode number, the
generation number, the mtime seconds and nanoseconds GMT, and the ctime seconds
and nanoseconds GMT.  The output file is sorted by the inode numbers.
The fields are separated by tab characters.

Example:

>tstimes /gpfs
0000000003/0000000001   1012390401.913049600    1012390401.913049600
0000001452/0000000003 1024702059.694025000  1024702059.694025000
0000001812/0000000000 1024702015.294047000  1024702015.294047000
0000019578/0000000000 1026797433.109149000  1026794560.584238000
0000023612/0000000017 1024702090.705801000  1024702090.705801000
0000023874/0000000000 1024702024.982645000  1024702024.982645000
>


------------------------------------------------------------------------------

mmbackup/tsbackup (see documentation for mmbackup in the GPFS Admin Guide)

tsbackup     - Fast parallel backup utility using multiple TSM clients
               and a single TSM server.  Parallelism is achieved via
               the use of multi-threading and the use of multiple TSM
               client processes.  This utility makes use of the GPFS
               inode scan APIs.

  Usage:
    mmbackup Device -n ControlFile [-t {full | incremental}] [-r IOrate]
      or
    mmbackup Device -R [-r IOrate]

Parts:

tsbackup.C       (sample code in this directory)
tsbackup.h       (sample code in this directory)
mmbackup.sh      (script in /usr/lpp/mmfs/bin)
mmexectsmcmd.sh  (script in /usr/lpp/mmfs/bin)