====== Dokuwiki & rsync ======

A method to synchronize two DokuWiki installations. 

For more elaborate concepts on this topic see [[dokupubsub]].

===== About =====

The idea is simple: For each wiki-write-transaction a small DokuWiki action plugin launches rsync to publish/update the underlying data storage. 

Synchronization happens in the background: a small perl script prevents multiple simultaneous [[wp>rsync]] processes and re-runs if changes occur during an active transfer.

===== Notes =====

Note: The index of the receiving side is not updated automatically.

While ''rsync'' is capable of bi-directionally synchronizing multiple wikis. The script below is in use on a (read/write) master to (mostly read-only) slave wiki.  Use rsync's ''-u'' option, but with care: page-locks, and edit-drafts are not distributed.  - While suitable for a small group of editors, race-conditions may occur on larger public wikis.

One workaround is to make a receiving namespace read-only; optionally in combination with the [[http://www.dokuwiki.org/plugin:include|include plugin]] to make parts locally writable content.

Alternatively one can make use of redirecting or proxying write-requests (HTTP-POSTs in general) to the central //master//. There can be different master servers per namespace; but depends on secure cross-site requests and authentication token passing.

Last but not least NFS-sharing the ''data/lock'' folder may be an option. see [[dokupubsub]].


[[http://dokuwiki.org/tips:farm]] and [[http://dokuwiki.org/tips:farm2]] have hints about patching DokuWiki to set up master server (aka. //farmer//).

===== Source =====

Configure and test the perl script first; then save this plugin as ''lib/plugins/rsync/action.php'' and edit a page to test it. check ''/tmp/rwikisync.log''. 

<code php>
<?php
/**
 * @license    GPL 2 (http://www.gnu.org/licenses/gpl.html)
 * @author     Robin Gareus <robin@gareus.org>
 */

// must be run within Dokuwiki
if(!defined('DOKU_INC')) die();

if(!defined('DOKU_PLUGIN')) define('DOKU_PLUGIN',DOKU_INC.'lib/plugins/');
  require_once (DOKU_PLUGIN.'action.php');

class action_plugin_rsync extends DokuWiki_Action_Plugin {

  function getInfo(){
    return array(
      'author' => 'Robin Gareus',
      'email'  => 'robin@gareus.org',
      'date'   => '2007-05-23',
      'name'   => 'rsync plugin',
      'desc'   => 'rsync dokuwiki data',
      'url'    => '',
    );
  }

  function register($contr){
    $contr->register_hook('IO_WIKIPAGE_WRITE',
                          'AFTER',
                          $this,
                          'handle_wikipage_write',
                          array());
  }

  function handle_wikipage_write($event, $param) {
    // if an old revision is saved or the file is empty -> run away
    if ( $event->data[3] || !$event->data[0][1] ) return false;
    //$ns=$event->data[1];
    //$path=$event->data[0][0];
  
    system ("/usr/local/bin/dokusync.pl &>/tmp/rwikisync.log &");
    return true;
  }

}
</code>

The Perl script to rsync the actual data. Save as ''/usr/local/bin/dokusync.pl'' or change the path in the above plugin code accordingly.

You need to change ''username'' and provide a ssh-key (just generate one with ''ssh-keygen''). There's many tutorials on [[http://www.google.com/search?q=rsync+ssh+key|ssh keys and rsync]]. 

<code perl>
#!/usr/bin/perl

### CONFIG
$rsyncopts="-azv --delete --no-o --no-g --chmod=Dg+wx,ug+rw";
$myssh="ssh -i /var/opt/id_rsa -l username";
@excludelist  = ( "_darcs", "cache/", "attic/", "private/", "index/" );
$src="/var/www/data/";
$dstpath="/site/docroot/data/";
$dsthost="example.org";

$dst=$dsthost.":".$dstpath;

###prevent multiple instances
use File::Pid;
$pidfile = File::Pid->new({ file => "/var/lock/rwikisync.pid"});
if ( my $num = $pidfile->running ) {
        # touch restart-file
        system("touch /var/lock/rsyncrestart");
        die "Already running: $num - restart scheduled\n";
}
$pidfile->write;

### build commandline
$command="rsync $rsyncopts -e \"$myssh\" ";
foreach (@excludelist) {
  $command.=" --exclude '$_'";
}
$command.=" $src $dst";
### and execute it..
system("date");
system($command);
### fix group permissions
system("$myssh $dsthost \"chgrp -R www-data $dstpath 2>/dev/null\"");

# check if restart was attempted -> rm restart file and rsync again..
while ( -e "/var/lock/rsyncrestart" ) {
        unlink "/var/lock/rsyncrestart";
        system($command);
}

$pidfile->remove;
</code>

{{tag>dokuplugin FLOSS WWW development}}