Extension talk:CirrusSearch/2015

omit results from search results page

Hi,

Is it possible to omit some of the results from the results page?

I have pages with acl, that are not allowed for anyone. If someone searches in the wiki for a phrase that's in one of those pages, he will get the page in the results (although it's without the description, so that's good).

So the user can now know that the secret page contains some specific string.

I would like to omit those pages from the results page, so the users won't even know that the string exist in that page.

I tried looking at some hooks, but didn't found something helpful.

Any idea how can I accomplish that? Maybe there's a hook I've missed?

Thanks,

Ariel. Arieljannai (talk) 01:58, 14 January 2015 (UTC)

What ACL extension are you using? See also Security issues with authorization extensions.

Extension:Lockdown disables snippets too. Nemo 11:06, 14 January 2015 (UTC)

Thanks for your answer. I'm using a plugin that was developed inside my organization a few years ago, and is similar to Lockdown extension you mentioned, although ours is a bit more comprehensive (also checking rss and api).

I looked again at Lockdown, and saw that also there they're not omitting search results.

I've tried using SpecialSearchResultsPrepend, but the results are not part of the html over there.

Do you have an idea of how can I accomplish that? Arieljannai (talk) 06:34, 20 January 2015 (UTC)

By "looked", do you mean tested? According to Extension:Lockdown/hiding_pages#Special:Search, snippets are removed; but to also delist results you'll need to apply that hack. It would be great if you published the code of your custom solution, or if you reported what you are missing in Lockdown that prevents you from adopting it. Nemo 15:10, 22 January 2015 (UTC)

Probably missed that part :O

I'll look at this again, thanks! Arieljannai (talk) 05:13, 28 January 2015 (UTC)

Login not working

Has anyone experienced an issue where a user can not login when CirrusSearch is enabled on 1.23.8? It is the only extension turned on and everytime a user logs in, it will never log them through. The cookies and session data is there, but they are not logged in. ProgramCeltic (talk) 22:10, 29 January 2015 (UTC)

Does not work with https?

Warning: Search backend error during full_text search for 'test' after 4. SearchPhaseExecutionException[Failed to execute phase [dfs], all shards failed; shardFailures {[QGlYV-7aT6GhYjKRf74CZg][zhmoegirl_content][0]: SearchParseException[[zhmoegirl_content][0]: query[((all.plain:test | all:test^0.5) all_near_match:test)~1],from[-1],size[-1]: Parse Failure [Failed to parse source [{"_source":["id","title","namespace","redirect.*","timestamp","text_bytes"],"fields":"text.word_count","query":{"bool":{"minimum_number_should_match":1,"should":[{"query_string":{"query":"test","fields":["all.plain^1","all^0.5"],"auto_generate_phrase_queries":true,"phrase_slop":0,"default_operator":"AND","allow_leading_wildcard":false,"fuzzy_prefix_length":2,"rewrite":"top_terms_128"}},{"multi_match":{"fields":["all_near_match^2"],"query":"test"}}]}},"highlight":{"pre_tags":["<span class=\"searchmatch\">"],"post_tags":[""],"fields":{"title":{"number_of_fragments":0,"type":"fvh","order":"score","matched_fields":["title","title.plain"]} in /wiki/includes/debug/MWDebug.php on line 302 Zoglun (talk) 00:36, 6 February 2015 (UTC)

Works for me. Where did you perform your search? Can you share the URL of your search? Nemo 13:52, 6 February 2015 (UTC)

Solved! The reason is nginx did not pass request to right address. It should used localhost/127.0.0.1 instead of outside IP.

Nemo, Thank you for your help. Zoglun (talk) 00:59, 7 February 2015 (UTC)

not picking up pages imported with importDump.php

I installed CirrusSearch on a wiki (with MW 1.24.1) which gets its edits and its new pages from a database tool, which generates the page text (in xml) and imports it into the wiki with maintenance/importDump.php

It seems that CirrusSearch is not indexing those pages.

What can I do to force CirrusSearch to quickly index these new or edited pages?

Do I have to use forceSearchIndex.php --from or is there another way, maybe a modification to importDump.php ? Aloist (talk) 06:52, 13 February 2015 (UTC)

Yes, you have to reindex. Nemo 23:34, 14 February 2015 (UTC)

Yes, that is what I am doing now. The wiki is part of a multi-wiki installation and called 'adb'.

In my Ruby app it looks about like this

dt = Time.now.utc

dts = dt.strftime("%Y-%m-%dT%H:%M:%SZ")

... (create xml file with the new or changed wiki pages)

do_cmd "export REQUEST_URI=/wiki/adb && php /wwd/wiki/maintenance/importDump.php #{wdir}/#{fn}", log

do_cmd "export REQUEST_URI=/wiki/adb && php /wwd/wiki/extensions/CirrusSearch/maintenance/forceSearchIndex.php --from '#{dts}'", log

do_cmd "run_adb_recentchanges", log Aloist (talk) 09:38, 15 February 2015 (UTC)

wgCirrusSearchServers with or without port?

$wgCirrusSearchServers = array( '333.333.333.333:9200' );

or

$wgCirrusSearchServers = array( '333.333.333.333' );

? Zoglun (talk) 02:48, 14 February 2015 (UTC)

Without: WMF production $wgCirrusSearchServers setting BDavis (WMF) (talk) 23:07, 19 March 2015 (UTC)

Is it possible to forbid search of certain namespace?

For example. I don't want namespace "private" to be searched by user. Zoglun (talk) 23:09, 14 February 2015 (UTC)

What do you mean by "private" namespace? See Extension:Lockdown and Extension:Lockdown/hiding_pages#Special:Search for a way to achieve that. Nemo 23:31, 14 February 2015 (UTC)

Install CirrusSearch for local wiki - no change in search results

Hi, i installed CirrusSearch on my win7 64bit machine like the discription on https://www.mediawiki.org/wiki/Extension:CirrusSearch an even the readme-file on https://git.wikimedia.org/blob/mediawiki%2Fextensions%2FCirrusSearch.git/HEAD/README.

the one an only stone on the way was the message
when i call
php $MW_INSTALL_PATH/extensions/CirrusSearch/maintenance/forceSearchIndex.php --skipLinks --indexOnSkip

now i am in hope of successfull search results - there is a page named "ALKIS" and when i put the word "alkis" into the textfield only the side "ALKIS" will automatically open. but the word "alkis" will written in many other pages.

first i search a way to test if all my installation work fine! is there a possibility to look to the elasticsearch-index words ??

in my opinion there must even the possiblity to search vor "auto" and words like "autocad" or "autolisp" will be listed in the result !

regards Jan JanTappenbeck (talk) 12:16, 26 February 2015 (UTC)

Update of index

hi !

one questions again - did i understand Help:CirrusSearch#Updates correct:

there is no cronjob or windows task require to update the CirrusSearch-Index?

regards Jan :-) JanTappenbeck (talk) 14:14, 26 February 2015 (UTC)

Generally, no cronjob or Windows Task is required to update the index -- it just happens. You can just test this yourself by adding a new page and then searching for something in the body.

The index updates will happen as part of the normal operation of the wiki engine. If you've left $wgJobRunRate alone, then some maintenance tasks get executed as part of periodic user page requests. If you set that variable to zero, you're responsible for setting up a cron job to call maintenance/runJobs.php, which will perform the required maintenance tasks. Lsilverman (talk) 19:22, 26 February 2015 (UTC)

hi !

thanks for answer. there is no search result of my new page. i looked into the data-folder of elasticsearch and no file will be change since the first index call. is there a special way to set allow for writing to the elasticsearch-files ?? in documentation i did not found them !

for manuall cron-jobs you tell about the file "wp-cron.php" - are you shure this is right? when i search to google only word in context of wordpress will listed.

regards Jan JanTappenbeck (talk) 07:37, 27 February 2015 (UTC)

I'm sorry, I wrote the wrong command. Instead of wp-cron.php, I should have written maintenance/runJobs.php Lsilverman (talk) 14:21, 27 February 2015 (UTC)

Installationproblem again - look to error-file (win7 64bit)

Hi! in addition to my other postings i look current into the php_error.log-file. there are two error parts i found 11:2x is the time of generate the first index by

php $MW_INSTALL_PATH/extensions/CirrusSearch/maintenance/forceSearchIndex.php --skipLinks --indexOnSkip
php $MW_INSTALL_PATH/extensions/CirrusSearch/maintenance/forceSearchIndex.php --skipParse 

here is the report:
[... 11:47:42 ...] PHP Warning: is_file(): open_basedir restriction in effect. File(/usr/bin/git) is not within the allowed path(s): (C:\....) in C:\mediawiki\includes\GitInfo.php on line 132
...
[... 11:49:41 ...] PHP Fatal error: Call to undefined function Elastica\Transport\curl_init() in C:\mediawiki\extensions\Elastica\Elastica\lib\Elastica\Transport\Http.php on line 183
...
[... 10:41:24 ...] PHP Warning: file_exists(): open_basedir restriction in effect. File(C:\mediawiki/../wmf-config/wikimedia-mode) is not within the allowed path(s): (C:\....) in C:\mediawiki\maintenance\doMaintenance.php on line 80
...
[... 11:28:44 ...] PHP Warning: Search backend error during sending 10 documents to the content index after 31. NullPointerException[null] [Called from CirrusSearch\ElasticsearchIntermediary::failure in C:\mediawiki\extensions\CirrusSearch\includes\ElasticsearchIntermediary.php at line 97] in C:\mediawiki\includes\debug\Debug.php on line 303
[... 11:33:55 ...] PHP Warning: file_exists(): open_basedir restriction in effect. File(C:\mediawiki/../wmf-config/wikimedia-mode) is not within the allowed path(s): (C:\....) in C:\mediawiki\maintenance\doMaintenance.php on line 80

php_curl.php is in the extension-folder and listed into phpinfo.php.

can someone help to me fix the problems?

• (/usr/bin/git) - which folder i had to put into list of base_dir on windows ?
• C:\mediawiki/../wmf-config/wikimedia-mode i did not found this folder to add into base_dir - what is the complete name?

with steps had to be repeated after this for the index-process ?

reagards Jan JanTappenbeck (talk) 11:32, 27 February 2015 (UTC)

hi !

a additional question again:

what is the job of the curl funktion in depend of

[... 11:49:41 ...] PHP Fatal error: Call to undefined function Elastica\Transport\curl_init() in C:\mediawiki\extensions\Elastica\Elastica\lib\Elastica\Transport\Http.php on line 183

is it a "high risk" for the use of CirrusSearch when this message come...?

regards Jan JanTappenbeck (talk) 09:09, 3 March 2015 (UTC)

That error looks like your PHP install does not include the curl PHP extension. The curl_init function is provided by that extension. Your PHP is not finding that function in the global scope or in the local namespace of the Elastica\Transport\Http class and reporting the function as missing in the class namespace. See a tutorial like http://www.tonyspencer.com/2003/10/22/curl-with-php-and-apache-on-windows/ for help installing and configuring curl for PHP on Windows. You will know you have it installed when php -m includes curl.

The CirrusSearch extension requires curl support to communicate with the backend Elasticsearch server. BDavis (WMF) (talk) 23:03, 19 March 2015 (UTC)

[Regex] interval syntax error

What's an "interval syntax error"? Example: [1] Nemo 11:53, 28 February 2015 (UTC)

In the Lucene regular expression language, <n-m> is the representation of a "numerical interval". The regex parser is balking at insource:/<!-- categories -->.*\\n/ containing an unescaped < that doesn't delimit a numerical interval. This syntax seems to get past the regex parser: insource:/\<!-- categories -->.*\\n/. BDavis (WMF) (talk) 22:21, 19 March 2015 (UTC)

Thanks, I'll add a link from the help page. Nemo 20:25, 25 March 2015 (UTC)

Fatal Error: Class Not Found (new installation)

I'm encountering an error with CirrusSearch both from CLI and MW1.24.

CLI ouput:

$> php updateSearchIndexConfig.php PHP Fatal error: Class 'CirrusSearch\Connection' not found in /opt/www/sites/mediawiki/extensions/CirrusSearch/maintenance/updateSearchIndexConfig.php on line 47

Output of error.log:

2015/04/08 13:27:16 [error] 2266#0: *146 FastCGI sent in stderr: "PHP message: PHP Fatal error: Cannot instantiate abstract class ElasticaConnection in /opt/www/sites/mediawiki/extensions/Elastica/ElasticaConnection.php on line 37" while reading response header from upstream, client: [ipaddress], server: [servername], request: "GET /index.php?search=audit&title=Special%3ASearch&go=Go HTTP/1.1", upstream: "fastcgi://unix:/var/run/php-fpm.sock:", host: "[servername]", referrer: "[servername]/Main_Page"

LocalSettings.php

require_once "$IP/extensions/Elastica/Elastica.php";

require_once "$IP/extensions/CirrusSearch/CirrusSearch.php";

$wgDisableSearchUpdate = true;

$wgCirrusSearchServers = array( 'localhost' );

$wgSearchType = 'CirrusSearch';

File permissions appear to be properly set. All other extensions are functioning (including Elastica).

Can anyone provide some insight into how this may be resolved? 206.116.111.163 18:20, 8 April 2015 (UTC)

Managed to resolve the CLI issue with some additional settings due to a conflict with ElasticSearch 1.5. Error is persisting with web access. 206.116.111.163 19:34, 8 April 2015 (UTC)

PHP Fatal error: Cannot instantiate abstract class ElasticaConnection in /opt/www/sites/mediawiki/extensions/Elastica/ElasticaConnection.php on line 37

This looks to be a late static binding failure.

	public static function getSingleton() {
		static $instance;
		if ( !$instance ) {
			$instance = new static;
		}
		return $instance;
	}

Line 37 is the $instance = new static; line and static should presumably be bound to the \CirrusSearch\Connection class rather than the ElasticaConnection parent class where the method is declared.

What PHP version are you using? I can't find any obvious PHP interpreter bugs related to this but I may be missing something. BDavis (WMF) (talk) 00:22, 15 April 2015 (UTC)

$> php -v returns 5.4.37. Nothing particularly exotic. 198.84.189.163 10:30, 20 April 2015 (UTC)

Search backend error during full_text search

I've got an error after installation of Elastica and CirrusSearch

Search backend error during full_text search for 'Morrowind' after 17.  SearchPhaseExecutionException[Failed to execute phase [dfs], all shards failed; shardFailures {[yWCaARxnRqCQyeqtEyzy0g][almanach-wiki__content_first][0]: SearchParseException[[almanach-wiki__content_first][0]: query[filtered(((all.plain:morrowind | all:morrowind^0.5) all_near_match:morrowind)~1)->cache(namespace: � in <b>/var/www/wiki/includes/debug/MWDebug.php</b> on line <b>300</b><br /> Scharesoft (talk) 14:09, 1 July 2015 (UTC)

Have you solved? i have the same issue Davidemaglio (talk) 14:58, 10 November 2015 (UTC)

I have solved because i tried to use git clone in MW 1.25.3 Davidemaglio (talk) 09:13, 11 November 2015 (UTC)

I Also have this problem on MW 1.25.5 , How did you solve it daviddemaglio 91.192.31.192 (talk) 07:32, 27 January 2016 (UTC)

Could you post the full debug output? Perhaps create a paste at https://phabricator.wikimedia.org/paste/create/ or another site to include the entire thing (it's likely very verbose). EBernhardson (WMF) (talk) 18:33, 8 February 2016 (UTC)

I also had this issue (because I've upgraded elasticsearch/libelasticsearch by accident), and solved it by verifying (and reinstalling if necessary) the following packages:

elasticsearch            all    1.7.1

libelasticsearch1.7-java    all    1.7.3+dfsg-3

libjna-java            all    4.1.0-1

libjna-jni            armhf    4.1.0-1

(It's a Debian-for-ARM-system)

I couldn't find the 1.7.1-version of libelasticsearch1.7-java so I wasn't able to downgrade it, but it works.

As stated on CirrusSearch's extension page, CirrusSearch needs elasticsearch 1.7.1, version 1.7.3 (as distributed by Debian stretch) doesn't work with CirrusSearch.

After that, I've rebuild the index: http://pastebin.com/zpkLr6QN

This caused an exception, but waiting ~20 seconds and retrying the command helped.

MediaWiki version information:

MediaWiki         1.26.0

CirrusSearch (Plugin)    0.2

Elastica (Plugin)    1.3.0.0 Ezivert (talk) 15:56, 14 February 2016 (UTC)

Thanks for the detail! CirrusSearch should now work with elasticsearch 1.7.5 (on the master branch). We've just updated our production servers to 1.7.5 and things are looking smooth so far.

The exception looks to be a temporary one, and the posted elasticsearch exception is directly related to the exception thrown by CirrusSearch. In this case elasticsearch had not yet recovered a maintenance index we use after starting up (or possibly being created). I'm a bit surprised it was in the recovery state, the index is typically completely empty and has a very simple mapping. It should create/restore crazy quick. That index is only used at write time, I will think about potentially adding a wait and retry call there as it won't effect user perceived search latency. Regardless that particular problem would have gone away as soon as elastic finished initializing the index, so you did the right thin by retrying. EBernhardson (WMF) (talk) 22:01, 3 March 2016 (UTC)

Upercase wiki names do not work

We have several mediawiki installation and the databases are both upper and lowercase. When a MW is created with a lower case database (wikisandboxdb, for example) works with no trouble but when a DB with some uppercase is used, the index processes die.

Unexpected Elasticsearch failure. Elasticsearch failed in an unexpected way.  This is always a bug in CirrusSearch. Error type: Elastica\Exception\ResponseException Message: InvalidIndexNameException[[wikiTUdb_content_first] Invalid index name [wikiTUdb_content_first], must be lowercase]

Trace: #0 /usr/local/mediawiki-extensions/Elastica/vendor/ruflin/elastica/lib/Elastica/Request.php(168): Elastica\Transport\Http->exec(Object(Elastica\Request), Array) #1 /usr/local/mediawiki-extensions/Elastica/vendor/ruflin/elastica/lib/Elastica/Client.php(592): Elastica\Request->send() #2 /usr/local/mediawiki-extensions/Elastica/vendor/ruflin/elastica/lib/Elastica/Index.php(444): Elastica\Client->request('wikiTUdb_con...', 'PUT', Array, Array) #3 /usr/local/mediawiki-extensions/Elastica/vendor/ruflin/elastica/lib/Elastica/Index.php(250): Elastica\Index->request('', 'PUT', Array, Array) #4 /usr/local/mediawiki-extensions/CirrusSearch/includes/Maintenance/Validators/IndexValidator.php(124): Elastica\Index->create(Array, true) #5 /usr/local/mediawiki-extensions/CirrusSearch/includes/Maintenance/Validators/IndexValidator.php(88): CirrusSearch\Maintenance\Validators\IndexValidator->createIndex(true) #6 /usr/local/mediawiki-extensions/CirrusSearch/maintenance/updateOneSearchIndexConfig.php(283): CirrusSearch\Maintenance\Validators\IndexValidator->validate() #7 /usr/local/mediawiki-extensions/CirrusSearch/maintenance/updateOneSearchIndexConfig.php(224): CirrusSearch\Maintenance\UpdateOneSearchIndexConfig->validateIndex() #8 /usr/local/mediawiki-extensions/CirrusSearch/maintenance/updateSearchIndexConfig.php(52): CirrusSearch\Maintenance\UpdateOneSearchIndexConfig->execute() #9 /usr/local/mediawiki-1.25.1/maintenance/doMaintenance.php(103): CirrusSearch\Maintenance\UpdateSearchIndexConfig->execute() #10 /usr/local/mediawiki-extensions/CirrusSearch/maintenance/updateSearchIndexConfig.php(59): require_once('/usr/local/medi...') #11 {main}

I've notice the key is wfWikiId, that might be uppercase, so I've replaced in the code all apperance of wfWikiId with strtolower(wfWikiId()):

/usr/local/mediawiki/extensions/CirrusSearch# grep -ir wfWikiId * includes/Api/SettingsDump.php:                Connection::getIndex( strtolower(wfWikiId()), $index )->getSettings()->get() ); includes/Api/MappingDump.php:                Connection::getPageType( strtolower(wfWikiId()), $index )->getMapping() ); includes/Updater.php:            $pageType = Connection::getPageType( strtolower(wfWikiId()), $indexType ); includes/Updater.php:                        Connection::getPageType( strtolower(wfWikiId()), $indexType )->deleteIds( $ids ); includes/Updater.php:                    Connection::getPageType( strtolower(wfWikiId()), $indexType )->deleteIds( $ids ); includes/Searcher.php:     * @param string $index Base name for index to search from, defaults to wfWikiId() includes/Searcher.php:        $this->indexBaseName = $index ?: strtolower(wfWikiId()); includes/Searcher.php:                array( 'local_sites_with_dupe' => strtolower(wfWikiId()) ) ); includes/Job/OtherIndex.php:        $otherIdx = new OtherIndexes( strtolower(wfWikiId()) ); includes/BuildDocument/RedirectsAndIncomingLinks.php:        $type = Connection::getPageType( strtolower(wfWikiId()) ); includes/InterwikiSearcher.php:     * @param string $index Base name for index to search from, defaults to wfWikiId() maintenance/saneitize.php:                $this->output( sprintf( "[%20s]%10d/%d\n", strtolower(wfWikiId()), $pageId, maintenance/forceSearchIndex.php:        $wiki = sprintf( "[%20s]", strtolower(wfWikiId()) ); maintenance/forceSearchIndex.php:        $wiki = strtolower(wfWikiId()); maintenance/backupIndexes.php:        $baseName = $this->getOption( 'baseName', strtolower(wfWikiId()) ); maintenance/cirrusNeedsToBeBuilt.php:                $count = Connection::getPageType( strtolower(wfWikiId()), $indexType )->count(); maintenance/indexNamespaces.php:        $type = Connection::getNamespaceType( strtolower(wfWikiId()) ); maintenance/updateOneSearchIndexConfig.php:        $this->indexBaseName = $this->getOption( 'baseName', strtolower(wfWikiId()) ); maintenance/updateVersionIndex.php:        $baseName = $this->getOption( 'baseName', strtolower(wfWikiId()) );

I assume you would like to fix it properly on the code. 94.16.81.106 (talk) 15:52, 16 July 2015 (UTC)

This issue is still present ... any better work around? Uwe a (talk) 00:20, 4 September 2018 (UTC)

yes it's tracked in T135021. DCausse (WMF) (talk) 08:28, 11 September 2018 (UTC)

For a database named "DBMyCoolDatabase" I use the configuration parameter $wgCirrusSearchIndexBaseName = 'dbmycooldatabase'; and CirrusSearch works fine. [[kgh]] (talk) 22:11, 8 December 2019 (UTC)

Need help upgrading CirrusSearch

I have been unable to successfully upgrade Cirrus (my "version" file says CirrusSearch: master 2015-02-03T22:04:18 9f1f0f1) to the latest version. Here is what I was doing, moving the old extensions/CirrusSearch folder away, and getting the new one. Following the instructions in the README file about updating I ran

>> php update_search.php --reindexAndRemoveOk --indexIdentifier now
this returned:

>> Can't use multiple processes without $wgCirrusSearchWikimediaExtraPlugin[ 'id_hash_mod_filter' ] = true

So I set $wgCirrusSearchWikimediaExtraPlugin[ 'id_hash_mod_filter' ] = true (after installing said plugin) in "CirrusSearch.php and tried again. This time I got back:

Reindexing...

                        [0] Starting child process reindex

PHP Fatal error:  Call to undefined method MediaWiki\Logger\LoggerFactory::getLogger() in /home/public_html/live/public/extensions/CirrusSearch/includes/Maintenance/Reindexer.php on line 318

Which is where I've been stuck. Any help is greatly appreciated.

T0lk (talk) 05:26, 15 August 2015 (UTC)

Got same error. Help needed! Deletedaccount4567435 (talk) 22:06, 15 May 2017 (UTC)

If the error is about MediaWiki\Logger\LoggerFactory not existing, that means the versions of CirrusSearch and MediaWiki are not in sync. What versions of cirrus and mediawiki are installed? EBernhardson (WMF) (talk) 12:29, 19 May 2017 (UTC)

Connecting to ElasticSearch server

I have installed elastic search on a synology NAS and installed the elastica and cirrussearch extensions for mediawiki. I have followed the procedure outlined here:

https://git.wikimedia.org/blob/mediawiki%2Fextensions%2FCirrusSearch.git/HEAD/README

but when I run

> php extensions/CirrusSearch/maintenance/updateSearchIndexConfig.php

I get the following output:

content index...

        Fetching Elasticsearch version...

Unexpected Elasticsearch failure.

Http error communicating with Elasticsearch:  Couldn't connect to host, ElasticSearch down?.

I have checked that the elastic server is running:

> curl 192.168.1.3:32770

{

  "status" : 200,

  "name" : "Ent",

  "cluster_name" : "elasticsearch",

  "version" : {

    "number" : "1.7.1",

    "build_hash" : "b88f43fc40b0bcd7f173a1f9ee2e97816de80b19",

    "build_timestamp" : "2015-07-29T09:54:16Z",

    "build_snapshot" : false,

    "lucene_version" : "4.10.4"

  },

  "tagline" : "You Know, for Search"

}

and that cURL is installed:

> php -i | grep -i curl

curl

cURL support => enabled

cURL Information => 7.36.0

any ideas what to do next or where to post this question? Any help is much appreciated

Sohansenit (talk) 12:10, 20 August 2015 (UTC)

We have the same problem.

After putting some debugging output into the CirrusSearch and Elastica extension, it looks like there is a problem with POST requests to elasticsearch server.

But currently it is unclear how to slove the problem or what is the root cause of it. 192.35.17.13 (talk) 13:05, 4 September 2015 (UTC)

Here the problem is due to non availability of the instance of ElasticSearch service. You need to start the ElasticSearch in the console and after that it may be used in CirrusSearch for Media WIKI.

Thanks,

Sanjay 202.60.62.100 (talk) 13:52, 15 September 2015 (UTC)

How can I index MS Office (doc, docx, xls, etc) documents?

Hello,

I'm testing MediaWiki and CirrusSearch for a new project. I've setup a minimal installation, and it's all working: contents of the Wiki and Pdf documents are indexed in the ElasticSearch server, and I can do searches.

But I also need to index MS Office documents (Word and Excel, mainly). I've installed the elasticsearch-mapping-attachments in the ElasticSearch server, but Office documents still aren't indexed (they appear empty, without content).

What should I do to tell CirrusSearch to index them? And another question: When are documents indexed? Should I force the indexing using the forceSearchIndex.php script, or is there any automatized way? 83.55.9.46 (talk) 10:06, 3 September 2015 (UTC)

Installing Elastica requires 'composer update'

This should be mentioned along with the 'require_once' in installation steps. Without the dependency, PHP Fatals with the message '[fatal] [3f648c60] PHP Fatal Error: Class 'Elastica\Transport\Http' not found'

01tonythomas (talk) 19:35, 29 September 2015 (UTC)

The most relevant pages aren't the top hits?

We are having problems with CirrusSearch returning results that are not the most relevant pages on the site. Full description from mediawiki-l:

https://lists.wikimedia.org/pipermail/mediawiki-l/2015-October/044841.html

Any help appreciated, especially how to start debugging this. Thank you. Maiden taiwan (talk) 17:07, 12 October 2015 (UTC)

Community Liaison job opening at WMF

Hi. There's a new job posting for a Community Liaison to work with the Discovery department. Please pass it along, if you know someone who might be interested or a good fit. Thanks. Quiddity (WMF) (talk) 20:11, 21 October 2015 (UTC)

Deleted pages showing up in search

As per topic. Is there some troubleshooting I should do? ☠MarkAHershberger☢(talk)☣ 21:31, 5 November 2015 (UTC)

Cirrus Search

When I try to search I'm having to use wildcards. Our end users will never understand what thats doing or why its needed. Is there any way to not require that? 192.174.72.128 (talk) 20:01, 7 December 2015 (UTC)