Introduction
Since Wedia 11.3.0, the full text search is based on ElasticSearch. The objective is to be able to respond more easily and quickly to different problems such as multilingual management, load sharing, content denormalization for the following purposes optimize the search. In addition, SQL criteria, fulltext and security rules are now directly managed by ElasticSearch, which solves performance and consistency problems when using ElasticSearch.
Prerequisites
Support for major versions of ElasticSearch depends on the Wedia version :
-
WEDIA 11.3: ElasticSearch 1.7
-
WEDIA 11.6: ElasticSearch 1.7, 2.x and 5.x
-
WEDIA 11.9: ElasticSearch 5.x recommended.
The ElasticSearch support for versions prior to 5 will be dropped in 2018.
Installation
The basic installation consists of downloading the latest version of ElasticSearch and starting it on a server. The ElasticSearch instance can be deployed locally on the Wedia server or on another server accessible by the latter. It is also possible and advisable to deploy an entire ElasticSearch cluster in production.
Once the ElasticSearch server is started, just go to the Wedia system settings (/admin/ebnAdminstration.ebn) and enter the URL of one of the ElasticSearch servers, the name of the index to use and the default indexing language. The default language is the parsing language for text fields that do not have language information such as non-internationalized fields or the default field for internalized fields. This section will be described later in this documentation.
|
Note
|
It is not necessary to create the ElasticSearch index in advance. The Wedia engine automatically takes care of this. However, care should be taken not to provide the name of an existing index for another application as it will be cleaned of any unnecessary mapping and configuration when using WXM. |
Below: an example of configuration.
Check ElasticSearch indexation
ElasticSearch indexing has its own administration screen:
/administration/ElasticSearch.
This screen allows:
-
to check if the index is configured correctly,
-
to check if the index is available,
-
to restart a complete, incremental indexing or indexing of a particular object.
-
to check the indexing state object by object and follow an indexation in progress.
|
Note
|
When an object appears as "unindexed" or "requiring an update", it does not prevent the application from being used. The object will be available again once its index is restored. |
|
Note
|
It is normally not necessary to start indexing manually in this Admin screen. The index is regularly checked and updated in the background without human intervention. |