Xfiles file tree synchronization and cross-validation
Contents
What it is
Screenshot
Requirements
License
Download
Preface to installation and running
Binary installation
Installation from Source
Setting up to run
Usage
Scripting
FAQ: How does Xfiles differ from Rsync?
Appendix: Version log
Appendix: Sample xfiles.py script
Appendix: Link/Alias detection
Appendix: Troubleshooting
Appendix: Installation from Source using native link detection
What it is
Xfiles is an interactive utility for comparing and merging
one file tree with another over a network. It supports freeform
work on several machines (no need to keep track of what files
are changed on which machine). Xfiles can also be used as a
cross-validating disk<->disk backup strategy
(portions of a disk may go bad at any time, with no simple indication of which files were affected. Cross-validate against a second disk before backup to make sure you aren't backing up bad data).
A client/server program (GUI on the client)
traverses a file tree and reports any files that
are missing on the server machine, missing on the client machine,
or different.
For each such file, the file size/sizes and modification date(s)
are shown, and a comparison (using Unix diff) can be obtained.
For files that are missing from one tree,
`similarly named' files in that tree are reported.
Inconsistent files can then be copied in either direction
or deleted on either machine.
The file trees do not need to be accessible via nfs.
Files checksums are computed in parallel, so largely similar trees can
be compared over a slow network link. The client and server
processes can also be run on the same machine.
See the topic 'How does Xfiles differ from Rsync' below for
more information.
Xfiles1.4 requires Java jdk1.2 or above.
To see if you have java installed, type java -version,
which should report something like
> java -version
java version "1.3.0"
There are various free implementations of jdk1.2 and above.
www.javalobby.org has links to some
implementations.
The Sun java versions are at java.sun.com under "Products and APIs".
www.blackdown.org has versions that are better tuned to Linux,
though their improvements are also integrated into recent Sun versions.
IBM also has a free JDK1.3 implementation with an alternate
high-performance virtual machine.
Read the LICENSE.txt file first. Xfiles is released under the GPL,
but it optionally uses jpython for scripting;
jpython has its own open source
license, see LICENSE.txt.
Download the latest
xfilesBinary or
xfilesSource or
archive.
Optionally download
jpython.jar
if you want to do scripting (see scripting section below).
[Note: in netscape download this by right-clicking on the link,
or by shift-button 1. Simply clicking on the link
causes the file to be read into the netscape window,
and sometimes causes the browser to hang if you have java turned on]
jpython.jar is also needed to recompile the source,
though small changes to the source will let it recompile without it.
jpython.jar is not needed if you do not want to script.
Download the nativeFile archive only if you want native Unix
link detection (see discussion at end). You probably do not need this.
Java tends to be poorly integrated with the underlying operating system.
If you're new to java take care to read the directions below --
there are not yet conventions for where and how java programs (nor
Java itself) should be installed, so the installation outlined below is
not as streamlined as a Linux rpm or similar. The individual steps are
easy however.
If you don't already have Java you'll
need to select some install location such as /usr/local/java;
any directory will work.
Similarly, the Xfiles program can live anywhere. If you follow the
installation below you will need to launch the program from the directory
where it resides (this does not restrict its function). Changes to the
shell files to make it run from any directory are evident.
The program and this documentation refer to client and server machines
and directories. These are interchangable -- the server merely refers to
the machine that the server is running on (see below).
Binary Installation
For both client and server machines, do the following steps:
- Put the Java bin directory on your path.
-
If using scripting put the jpython.jar file in the same
directory as xfiles.jar and the
xfilesClient, xfilesServer scripts.
Installation from Source
(The following instructions are for Unix/Linux)
- Put the Java bin directory on your path.
The programs javac, java, and rmic will be needed.
- To compile the source you must download jpython.jar
and place it in the source directory.
-
make
Setting up to run
- The following files comprise the program, verify that these
exist in the current directory:
xfiles.jar java archive containing the program
xfilesClient shell program to launch the client gui
xfilesServer shell program to launch the server
jpython.jar optional, needed for scripting
- Networking must be configured and turned on even if both
client and server are running on the same machine. I believe that
Java RMI needs the system service that converts a hostname into an IP address,
and uses this even if both the client and server are running on the same
machine. There is discussion of this point on some of the java mailing lists.
On my Redhat system, all I do is run
netcfg and turn on one
of the configured interfaces. Doing /sbin/ifconfig <interface>
up would probably work too.
You may need to enable the hosts in the Unix .rhosts file.
If you know networking please send me more authoritative(spell?) instructions
for what's needed here.
-
Launch the Xfiles server. On the server machine, run:
xfilesServer
This should print out the server hostname,
then "XfilesServer is running".
If you get an error that says "java.lang.ClassNotFoundException: <name
of class>", one of the paths in the xfilesServer file is not set
correctly.
-
Launch the Xfiles client on the client machine,
giving the server hostname (spelled as printed by xfilesServer)
and the roots of the client and server file trees:
xfilesClient duckpond /usr/jfk/devl /home/jfk/Devl
The command above will compare the file tree starting at /usr/jfk/devl
on the client machine with the file tree
/home/jfk/Devl on the machine 'duckpond'.
Usage
When the GUI comes up, select a directory and press the start button.
The client displays the client file tree at startup to allow
you to select a sub-tree of the specified root if desired.
If you select nothing the entire tree will compared when
you hit the start button.
To save you time Xfiles first scans the whole (sub)tree
before reporting any differences (this may take a while);
all differences are then reported consecutively.
After synchronizing one directory, you can select another in the GUI
and press start again.
Currently the GUI file tree does not update to reflect deletions in earlier
runs, however (see the TODO section).
The following are optional command-line arguments:
- -fast
This causes Xfiles to ignore files whose size and date are identical,
so the scan is much faster. This will not detect disk corruption however.
- -exclude substring1 substring2 ...
Arguments up to the next flag argument indicate files to be excluded.
Any file that contains one of these arguments as a substring will
be skipped.
- -clientonly
Focus on the client: files that are older on the server or are missing
on the server are not reported.
- -serveronly
Focus on the server: files that are older on the client or are missing
on the client are not reported.
- -scriptargs
Arguments following this are not examined; all arguments are passed
to the optional xfiles.py setup(args) function,
so scripting arguments should be passed following this flag.
Xfiles writes a file XFILES.LOG listing the
selected actions.
Scripting
File selection and interaction with a revision control system such
as RCS can be handled by scripting using jpython.
To enable scripting, download the file jpython.jar
and place it in the client and server launch directories.
Then create a file xfiles.py, which also must be copied into both the
client and server launch directories.
xfiles.py can define the following functions:
- setup(args)
This function, if defined, will be called with the command line arguments.
Arguments following -scriptargs are ignored by the application,
but all arguments (not just those following -scriptargs)
are passed to the xfiles.py script.
- pathFilter(path)
This function should return 1 if Xfiles should process path,
0 if Xfiles should ignore it.
- preCopy(path)
this function is called before Xfiles writes this file.
It can be used (for example) to check out the file from RCS.
- postCopy(path)
this function is called after Xfiles writes the file.
It can be used to check the file into RCS.
It is not necessary to define all of these functions, however,
if a function is defined it should be correct -- if the
function call generates an error Xfiles will quit.
A sample xfiles.py file is contained with the distribution
and listed at the end of this document.
FAQ: How does Xfiles differ from Rsync?
Rsync works automatically, and assumes the source directory is correct.
If that disk has corruption, good files on
in the backup directory may be overwritten with corrupted
files from the source disk. A similar scenario is when you
save a file that was accidentally edited somehow.
Again, the newer file is not the desired one.
Xfiles is also allows relatively unstructured work
in several places, without needing to think of one
location as the 'master' -- in this case each directory may have files
that are obsolete, as well as ones that should be mirrored.
An automated algorithm will duplicate the files that need to be deleted,
resulting in extra work later. Xfiles ask you if you want
to copy (and in which direction), and gives the option to delete.
In the case of text files it also gives a "diff" (On unix)
to help in the decision.
In the case where both files have desirable parts, the file must
be manually merged; Xfiles doesn't help here (except to identify the
situation). Adding something like the Unix xdiff visual merge
would be a nice addition.
In rsync's favor, rsync is much faster. (Although, if you add -c
to force checksums on all files, rsync gets a lot slower).
Author
Primary author: J.P.Lewis scribblethink.org
Contributors: Peter Gadjokov, Wolfgang Lugmayr, Dan Schmitt
Please e-mail problems, successes, fixes, and fears
to: zilla@computer.org
Send email with subject line XFILES to be notified of updates.
To do
- It should be possible to rename files.
- The client side gui should update to reflect the changed file tree
- Portable replacement for diff.
- If both client, server files are ascii it would be better
to transmit a diff and do a patch on the other end.
Appendix: Version log
- 1.4 jul00
Added -clientonly, -serveronly, -fast flags;
preserve dates when copying (assumes jdk1.2),
update files to assume jdk1.2 (JFC no longer required on classpath),
xfiles.py has setup() call.
- 1.3 28mar99
Added JPython scripting to control what files
are visited and to script interaction with a revision control
system such as RCS. 1.3.1: improved documentation.
- 1.2 22mar99
External rmiregistry no longer needed
(simplified usage and troubleshooting,
also helps if running over a firewall),
paths are transmitted from client to server as java.io.Files
rather than Strings so
that path separator translation is done at unserialization time --
should help if client and server are running on different O/Ss.
- 1.1 7jan99
Rewritten to not require native (non-java) link detection;
program is more portable to non-Unix environments;
installation is simplified; other improvements.
- 1.0.2
improved definition of 'similarly named' files,
warn if deleting on the wrong side, copy direction is color coded.
- 1.0.1 5jan99
fixes: stop did not work always, bad window sizing,
prohibit copying in the wrong direction
by accident - e.g. client to server
if the server file exists and the client does not.
- 1.0 3jan99
initial version
Appendix: Sample xfiles.py script
The following example script implements these functions:
- setup(args)
- Just prints out the command-line arguments
- pathFilter(path)
- Ignores all object files, libraries, .so files, RCS files, java class files
- preCopy(path)
- Checks out a file from RCS if needed before you copy over it.
- postCopy(path)
- Checks a newly copied file back into RCS with a message
saying that Xfiles created the revision.
Although jpython is quite compatible with regular python,
this file is written using a mixture of python and java calls
(for example, java string functions are used rather than
the python string.split routine).
This is done to limit the dependence on jpython to the
one file jpython.jar.
import java
from java.io import File
import java.lang.Runtime
runtime = java.lang.Runtime.getRuntime()
# ignore files that end with these strings
skipextensions = ['RCS', ',v', '.o', '.so', '.a', '.class', '.jar']
# grab the command line arguments.
# You could pass in the skipextensions here for example.
def setup(args):
print 'setup got args:'
for arg in args:
print arg
# return 1 if xfiles should visit this path, else 0
#
def pathFilter(path):
print 'pathFilter(%s)' % path
if path[len(path)-1] == '~': # emacs backup file
return 0
if path == 'so_locations':
return 0
spath = java.lang.String(path)
for ext in skipextensions:
if spath.endsWith(ext):
return 0
return 1
# called before copying over a file
# (check out from RCS if appropriate)
#
def preCopy(path):
name = filename(path)
spath = filedir(path)
spath = spath + '/RCS/'
print 'name = %s' % name
if exists(spath): # RCS/ exists
spath = spath + name + ',v'
print 'spath = %s' % spath
if exists(spath): # RCS/file,v exists
docmd('co -l -f %s' % path)
# called after copying over a file
# (check in to RCS if appropriate)
#
def postCopy(path):
name = filename(path)
spath = filedir(path)
spath = spath + '/RCS/'
print 'name = %s' % name
if exists(spath): # RCS/ exists
spath = spath + name + ',v'
print 'spath = %s' % spath
if exists(spath): # RCS/file,v exists
docmd('ci -u -f -mXfiles_copy_checkin %s' % path)
# helper commands
def docmd(cmd):
if 1:
print cmd
pid = runtime.exec_(cmd)
pid.waitFor()
def filedir(path):
result = File(path).getParent()
if not result:
if isabs(path):
result = path # Must be root
else:
result = ""
return result
def filename(path):
return File(path).getName()
def exists(path):
return File(path).exists()
def isabs(path):
return File(path).isAbsolute()
Appendix: Link/Alias detection
This section is largely obsolete because Java link detection appears to
work adequately.
Because Xfiles traverses a directory tree, it needs to be able to distinguish
between "real" files and links (aliases) so as to avoid an infinite
loop in the case where a link points to a directory above itself.
There are two approaches to this, and you need to select
which one you will use:
- A built in method
- An approach that relies on a
native (non-java) library nativeFile.
Xfiles will call the native function if it exists in the launch directory.
Installation with the native function is a bit more work, and it
does not exist for non-Unix operating systems yet.
For most purposes it will probably be fine to use the built-in code.
Read the appendix
Links/Aliases/Shortcuts in Java
for more details on this issue.
Troubleshooting Guide
- Connection refused to host: 127.0.0.1
The error
java.rmi.ConnectException: Connection refused to host: 127.0.0.1
this often has something to do with the /etc/hosts file.
On one Redhat6.1 machine there was a /etc/hosts file that looked like
127.0.0.1 localhost.localdomain localhost [Qserver]
xx.xx.xx.xx Qserver [localhost]
where [] means optional.
When "Qserver" is present on the first line and the rmi server
is running on this host, the client gets the `rmi connection' refused error.
When Qserver is removed from from the first line, the rmi connection
works, but now the machine hangs on startup at "starting sendmail"
for several minutes (but then continues ok).
The sendmail hang issue is discussed in one of the /usr/doc/HOWTO
documents, but I do not understand that discussion.
In another case this seemed to happen because the unix DISPLAY
variable was not set before running the server, or because
"host +" was issued before starting the X server.
- java.rmi.UnknownHostException: Unknown host: scruff
I think this happened on linux when /etc/HOSTNAME or
something else set the name "scruff",
but scurff was not found in /etc/passwd.
- 'import exceptions' failed; using string-based exceptions
This jpython warning is not a problem.
- java.lang.NoClassDefFoundError: org/python/core/PyObject
Means that jpython.jar was not found on the CLASSPATH at runtime.
This is not a problem if you are not scripting xfiles.
- XfilesScript.java:22: cannot resolve symbol
symbol : class PythonInterpreter
This means that jpython.jar is not found during compilation.
Download jpython.jar, put it in the compile directory, and
check that the Makefile adds it to the CLASSPATH.
Installation from Source, using native link detection
- Put the Java bin directory on your path.
The programs java, javac, and rmic will be needed.
-
To compile the source you must download jpython.jar
and place it in the source directory.
-
Download and extract the nativeFile archive.
-
Build the nativeFile library - in the Nativefile subdirectory
edit the Makefile CFLAGS to point
to the location of your java installation's include files and the JNI include
files. The latter are in a subdirectory of the java include, probably called
either "genunix" or the name of your OS.
-
Build the nativeFile library - build the java binding to the nativeFile:
make nativeFile.class
-
Build the nativeFile library - compile the .so: make nativeFile.so
-
Copy nativeFile.class and libnativeFile.so
into the directory containing Xfiles,
go to that directory.
-
Examine the LD_LIBRARY_PATH variable in xfilesClient
and xfilesServer and modify if needed.
-
make
A Serbo-Croatian translation of this page by Jovana Milutinovich from Geeks Education.
back to home