nnfs2

Name

nnfs2 -- Synchronise filesystems between Non-Networked hosts

Synopsis

nnfs2 { --no-x11} [--add-first directory-name | --add directory-name | --add-nickname directory-name]

nnfs2 [--help | --licence]

INTRODUCTION

The NNFS goal is to provide to ``one'' user a consistent file system over all the hosts used by the user.

With hosts connected on a LAN you must use NFS instead of NNFS. In the case of hosts not connected by a network or connected by a slow network (modem) NFS is not the answer.

With NNFS, the files are duplicated on all the computers. The up to date files are on the computer on which the user works. The NNFS filesystem is stored on the NNFS medium. Usually the NNFS medium is a floppy disk, it could be whatever you want (easily customisable), if there is much data, several floppies will be asked.

When the user stops his work on a computer, he runs as last command: nnfs2. Then all his modifications (change, creation, deletion) are recorded on the NNFS medium.

When the user starts to work on a computer, he runs as first command: nnfs2. Then all modifications (change, creation, deletion) stored on the NNFS medium are retrieved on current system.

If the user forgot to run an update, nnfs2 will nevertheless optimally work. The only drawback is that some files will not be up to date on the host used by the user.

It is also possible to have a numbered backup of modified/deleted files. So you cannot lose a file when you run nnfs2 (because the file is modified/deleted on another host)

USAGE

nnfs2 has a text and a graphical user interface. If you run the command for the first time or with a NNFS medium formatted or not, the command will provide contextual informations and help and will ask contextual questions. The graphical user interface allows you to edit the configuration.

The interactive help is extensive and guide the user. This documentation is here to give general principles.

By default the nnfs2 works with standards tools: floppy disk and gzip. The floppy disk access is done via system dependent commands (IRIX, AIX, HPUX and mcopy in the other cases). If you want to change these defaults, see the configuration section.

As the default configurations use floppy disk, the nnfs2 default filter does not synchronise big files or ``garbage'' files. If you want to change these defaults, see the configuration section.

It is not recommended to continue to modify files while nnfs2 is running. It is not dangerous but the result is unpredictable.

Be carefull, some applications wrote their data on disk only when you stop them. So you must stop these applications before running nnfs2. For example netscape wrote some folder information only when you quit the program, so you must run nnfs2 only if netscape is not running. To avoid this you can create a shell script killing these applications and cleaning some file before launching nnfs2.

To do before using nnfs2

If the file hierarchy you want to synchronise is yet identical on all the hosts you can go to the next section. The root of the file hierarchy may have a different name on each host. To avoid problems you should use only relative internal symbolic links, so the links are correct on all the hosts.

If the files hierarchy are not identical, you have two solutions:

Initialisations

When you run the nnfs2 command on an host not known by the NNFS filesystem it will explain you how to add the host in the NNFS filesystem.

If the NNFS filesystem does not fit in one medium, you will be asked to introduce medium.

Now, the initialisation for 2 hosts named A and B will be detailed for the 2 user interfaces.

Initialisations with X11 interface

First run on host A.

  • You click on: Continue without reading the medium

  • you enter the local name of the hierarchy to synchronise,

  • you click on: Continue the update,

  • you click on: Medium AAA is ready for writing,

  • you click on: Quit

Second run on host B, assuming medium is ready.

  • You click on: Continue the update

  • you enter the local name of the hierarchy to synchronise,

  • you click on: Continue the update,

  • you click on: Quit

Normal run when initialisations are done, assuming medium is ready.

  • You click on: Continue the update

  • you click on: Continue the update,

  • you click on: Quit

Initialisations with text interface

First run on host A.

  • You type: nnfs2 --add-first dirname

  • You answer: c return to continue the update

  • You hit : return when medium is ready

Second run on host B, assuming medium is ready.

  • You type: nnfs2 --add dirname

  • You answer: c return to continue the update

Normal run when initialisations are done, assuming medium is ready.

  • You type: nnfs2

  • You answer: c return to continue the update

How to recover from NNFS medium read failure

If there is a read error you can rerun nnfs2 a few times with the hope that the error will vanish. If it is not the case, put the bad medium in a trash, take a new one and create immediately a new NNFS filesystem with the current host.

If the read error is before you start to work on an host. Try to not work on non-synchronized file in order to avoid future conflicts. Run nnfs2 as always before stopping to work on the host.

If the read error is before you leave an host. You will want to copy the file modified by your session on the other host. To do so, modify all the file you want to synchronise and rerun nnfs2 on the host. For example by running find . -mtime -1 -print | xargs touch. If you do not do this, these files will be synchronised but only after adding the other host on NNFS filesystem, a run on local the host and a run on the other host.

My experience about floppies is that there is floppies usable only 2 or 3 times and some usable hundreds of time. Even if they are from the same box.

Historised backups

Each time you run nnfs2 the files deleted or modified are moved in a directory. This directory is not mirrored!

Example 1. Archiving modified files in default directory

Current date: Sun Sep 16 14:57:47 CEST 2001

Modified file: foo/bar/Makefile

Historised file: .nnfs/history/2001_09/16_14:57.47/foo/bar/Makefile

So you can easily remove old historised files by date.

Conflicts

If you don't run nnfs2 before and after working, some update conflicts may be raised if you work on the same files.

Example 2. A simple conflict

  • You run nnfs2 on host A

  • You modify foo on host A

  • You run nnfs2 on host A

  • You modify foo on host B

  • You run nnfs2 on host B (you should have done this before modifying the file)

The conflict solving method is straightforward, the remote conflicting file is always copied on local host. If you don't agree, you can retrieve the local file content from the .nnfs/history directory.

This method is not used to solve conflict when adding a new host. In this case, the most recent file is copied over the old ones

ADVANCED USAGE

Alias

If you run nnfs2 on several hosts using NFS, you must indicate to nnfs2 that the hosts use the same file system. The mirrored directory must have the same name on all aliased host

Assuming that the medium is ready, adding an alias with X11 interface.

Assuming that the medium is ready, adding an alias with text interface.

Configuration

The configuration file is .nnfs/nnfsrc, it is created the first time you run nnfs2. It is a hugely commented shell script you can edit. It is a shell script to allow some default values to be computed from the system.

If you find this configuration file too complex to edit, you can use the nnfs2 X11 interface to edit the file. There is many tips to help you configure.

The more useful options are :

The option hard to configure is OUTPUT_FILTER. The default value is fine for me but certainly not for you. The minimal filter must stop the copy of nnfs2 history and all backup files.

Example 3. Required minimal filter

-regex=~$
-regex=^\.nnfs/history/

But, to run nnfs2 you must be connected, and the connection modify some file you don't want to copy on other hosts because it is a security breach or it is a non-sense. For example, any dot file in connection directory ending by history or authority

Example 4. Filter files modified by the connection

-regex=^\.[^/]*(history|authority)$

If the medium is small, you need to filter the files that are garbage as core, the executable files or the results of compilation as .o files. I assume here that executables biggers than 50k are not scripts shell or perl.

Example 5. Filter garbage files

-type=f size=+50k perm=+111
-regex=(\.(o|a|so|sl|aux|log|dvi|summary|old)|/(core|a\.out))$

The big shell or perl script are filtered, if you don't want this. Tell nnfs2 to not filter them.

Example 6. Do not filter scripts (notice the + on line start)

+regex=\.(pl|sh|tcl)$
+regex=/configure$

Reference of the filter description

Each line starts by - or +, to remove or add to the file set the files verifying all the conditions in the line.

The conditions are:

  • perm=+xxxx: The condition is verified if any bit indicated in octal are in the file mode. If you indicate 0111 the file mode 0100, 0111, 0750 are verified.

  • perm=-xxxx: The condition is verified if all the bits indicated in octal are in the file mode. If you indicate 0111 the file mode 0755 is verified but not 0750.

  • perm=xxxx: The condition is verified if the file mode is exactly equal to the value indicated in octal.

  • size=+xxxx: The condition is verified if the file size if bigger than the size specified. A size in kilo-bytes may be specified as 56k and in mega-bytes as 2m

  • size=-xxxx: same as above, but the file size must be smaller. The two conditions may appear on the same line.

  • type=dflp: The type is regular file (f), directory (d), FIFO (p) or symbolic link (l)

  • regex=aregularexpression: If you want the expression to match the full name, add ^ at the begin and $ at the end. This condition must be at the end of the line.

Synchronisation of UID and GID

If you want to synchonise the files with there UID and GID you need to be root and to run the rootnnfs2.

FILES

~/.nnfs/nnfsrc

nnfs2 configuration

~/.nnfs/gtkrc

nnfs2 GTK configuration

~/.nnfs/history

Directory where modified or deleted files are saved

~/.nnfs/log~

Some log about nnfs2 last work

BUGS

Limitations:

Bugs:

NO WARRANTY

This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.

It is nearly impossible that nnfs2 do an irrecoverable error. In the worst case, nnfs2 will stop. The returned value (error) of all the system calls are verified and the program tries to analyse the error.

The parameters of the more stressing test on nnfs2 are the following:

Incredibly, nnfs2 works in this case thousands of time on a Linux 386 Debian host.

nnfs2 had been compiled and tested on HPPA 1.1 HPUX 10.20 and MIPS IP30 IRIX 6.5

On Mac OS X it does not work optimaly with symbolic links (see BUGS).

SEE ALSO

gzip(1)
mcopy(1)

AUTHOR

Author: Thierry EXCOFFIER, Author Home Page: http://www710.univ-lyon1.fr/~exco/

nnfs2 home page: http://www710.univ-lyon1.fr/~exco/nnfs.html