Can't get "docsearch" to work

tommes

Hi,

I'm new to dokuwiki, but had no problems installing it and all of the plugins are working, except docsearch. I've uploaded a PDF and the filelist-plugin show's it. Running cron.php (under the same user as the web server) build's a text file under /data/docsearch/pages out of the PDF without errors. But the /data/docsearch/index directory stay's empty (or does it have to?) and I get no search results.

Any ideas?

dominik

looks like the index isn't build properly. the first tip: check the permissions on the data/docsearch/index folder. Your webserver user needs write access to it.

tommes

Thanks for your answer. That was the first I've checked. I run the cron.php with the same user as the web server, so all files/dirs created are accessible. But I'll check it again. Just for understanding: Who creates the index? Does the cron.php build the index using the dokuwiki indexer?

dominik

docsearch uses the dokuwiki indexer.

tommes

Spending a few hours I dint' come any further. This is what I found out:

-permissions are ok, if I set the "docsearch/index" dir to root access I get an error
-cron.php deletes the "docsearch/index" and "docsearch/pages" directorys an creates them again, including the txt-files converted by pdftotxt
-the indexer is called by cron.php
-after cron.php is finished the "docsearch/index" directory is still empty

I'm not familiar with PHP, but I found the verbose-option in the "idx_addPage" function and set it to true.

Here is the output:

Indexer: locked
<div class="error">Creating directory /b failed</div>
<div class="error">Writing /b/b38b56844836212dcb466487609c6f62.i failed</div>
<div class="error">Unable to save cache file. Hint: disk full; file permissions; safe_mode setting.</div>
<div class="error">Creating directory /dokumente failed</div>
<div class="error">Writing /dokumente/klemmkasten.pdf.meta failed</div>
<div class="error">Unable to save metadata file. Hint: disk full; file permissions; safe_mode setting.</div>
<div class="error">Creating directory /e failed</div><div class="error">Creating directory /e failed</div>
<div class="error">Creating directory /e failed</div>
<div class="error">Writing /e/e509f2b11b6fed8cb125381050ef44e3.i failed</div>
<div class="error">Unable to save cache file. Hint: disk full; file permissions; safe_mode setting.</div>
<div class="error">Creating directory /dokumente failed</div>
<div class="error">Writing /dokumente/k_manual_de.pdf.meta failed</div>
<div class="error">Unable to save metadata file. Hint: disk full; file permissions; safe_mode setting.</div>
Indexer: locked

But I cant't explain this. There are 92GB of free disk space, permissions are ok (all files and dirs under /srv/www2/htdocs belong to www-data), safe_mode in /etc/php5/cli/php.ini is "Off". Does the script try do write in /var or /tmp? My full path to docsearch-data is /srv/www2/htdocs/data/docsearch. With wiki-pages the indexer works fine.

dominik

looks like permission issues or php safe_mode ;-)

<div class="error">Creating directory /b failed</div>
[...]
<div class="error">Creating directory /e failed</div><div class="error">Creating directory /e failed</div>
[...]
<div class="error">Creating directory /e failed</div>
[...]
<div class="error">Creating directory /dokumente failed</div>
[...]

dokuwiki cannot write the files as long as dokuwiki cannot create the appropriate folders ;-)

another tip, check if php safe_mode is enabled. When its enabled it may result in error on creating folders/writing to files.

if youre on linux and safe_mode is off try:

chown -R <your-webserver-user> dokuwikis/data/folder
chmod -R 775 dokuwikis/data/folder

tommes

Yes, I'm on Linux. Setting user and permissions recursive was on of the first things I've done. ALL files and dirs under /srv/www2/htdocs belong to www-data, folder permissions are 755. Safe_mode in /etc/php5/cli/php.ini is "Off".

If the permissions were wrong the script could'nt create the /data/docsearch/index and /data/docsearch/pages folders, but it does. It also creates the /data/docsearch/pages/dokumente folder , including the converted txt files. So why not the index? It's the same script running, the same privileges, the same user. But I have an idea, I'll check it on monday at work. Thanks so far.

tommes

So, I've come a little further. One have to read the ouput. The problem are the cache files. I ran the script as root and it creates the folders "/b","/e" and "/dokumente" in the root directory ("/")! Running the script as www-data it has no permissions to write there of course . After the script can save the cache files the index is correctly built in "/srv/www2/htdocs/data/docsearch/index".

I've had dokuwiki installed in the www-server-root and left the "basedir" option empty. Now I've created a "doku" subdir, moved all files there and set the basedir to "/doku/". But no changes. What I dont get: Most parts (converting pdf's and creating the index) work and use the correct folders, but creating the cache files does not and writes in "/".

tommes

Ok, I've found the problem. If the Indexer is called by cron.php it seems like the config variables for the directories "meta", "cache" and "locks" are not set. Therefore the script tries to write in the root-dir "/" and runs only if called by root. I dont know why and I wonder if I'm the only one encountering this issue, but I've found a workaround. I've added a few lines similar to the ones for the "pages" and "index" directories.

$cache  = $base . 'docsearch/cache';
$meta   = $base . 'docsearch/meta';
$locks  = $base . 'docsearch/locks';

io_mkdir_p($cache);
io_mkdir_p($meta);
io_mkdir_p($locks);

$conf['cachedir'] = $cache;
$conf['metadir']  = $meta;
$conf['lockdir']  = $locks;

I don't know if it's better to use the standard directories "/data/cache|meta|locks/" instead of "/data/docsearch/cache|meta|locks", but it works for me.

joemouth

tommes:1315784228 wrote Ok, I've found the problem.
.....

Hi,

sounds good. I have similar problems, but did not have any time to analyse.

Could you please give us more detailled infos regarding your patch?

Which file (cron.php?) ?
Inserted after/before which line?

tommes

Have a look at cron.php. Beginning at line 167 I added the paths for cache, meta and locks.

original:

// build the important pathes
$input = $base . ((isset($conf['mediadir'])) ? $conf['mediadir'] : 'media' );
$output = $base . 'docsearch/pages';
$index = $base . 'docsearch/index';

// create output dir
io_mkdir_p($output);
io_mkdir_p($index);

// change the datadir and the indexdir
$conf['datadir'] = $output;
$conf['indexdir'] = $index;

new:

// build the important pathes
$input  = $base . ((isset($conf['mediadir'])) ? $conf['mediadir'] : 'media' );
$output = $base . 'docsearch/pages';
$index  = $base . 'docsearch/index';
$cache  = $base . 'docsearch/cache';
$meta   = $base . 'docsearch/meta';
$locks  = $base . 'docsearch/locks';

// create directories
io_mkdir_p($output);
io_mkdir_p($index);
io_mkdir_p($cache);
io_mkdir_p($meta);
io_mkdir_p($locks);

// change the config
$conf['datadir']  = $output;
$conf['indexdir'] = $index;
$conf['cachedir'] = $cache;
$conf['metadir']  = $meta;
$conf['lockdir']  = $locks;

joemouth

Thanks for your detailled answer.

audiomobster

Thanks, worked for me too.

I posted a bug report about this.

And closed it short after because it was a duplicate.

There is an advanced cron.php with some error checking and messages in the git hub.

alexanderb

This solved the problem described here https://forum.dokuwiki.org/post/33401 (files and folders created outside of basedir).

Global DokuWiki Links