Modernizing WordPress Search with Elasticsearch

Modernizing WordPress
Search with Elasticsearch

Who Am I?
• My name is Taylor Lovett!
• Director of Web Engineering at 10up
• Open source community member
• WordPress core contributor
• ElasticPress team member
@tlovett12

10up is hiring!
@tlovett12
taylor.lovett@10up.com

Doesn’t WordPress have
search built-in?

WordPress Search is Rudimentary
• Only searches post title, content, and excerpt.
• Relies on MySQL and thus is slow.
• Relevancy calculations are poor and overly
simplistic.
• Not able to handle any advanced ﬁltering.

What is Elasticsearch?
http://www.elasticsearch.org/

Elasticsearch
• Open-source search server written in Java
based on a technology called Lucene (open-
source search software by Apache).
• A standalone database server that provides a
RESTful interface to accept and store data in a
way that is optimized for search.
• Extremely scalable, performant, and reliable

Elasticsearch
• Relevant results
• Autosuggest
• Fuzzy matching
• Geographic searches
• Filterable searches
• Data weighting
• Much more

Get an Elasticsearch Server
• Very ﬂexible and customizable. There is not
really a “one size ﬁts all” setup. Generally, you
have two options:
• Option 1: Pay someone else to manage/host
your Elasticsearch cluster (SaaS)
• Option 2: Host your own cluster

Elasticsearch SaaS
• found.no
• qbox.io
• heroku.com
• etc…

Let’s setup our own
Elasticsearch cluster.

Elasticsearch Self-Hosted Cluster
• Make sure you have Java installed.
• Download Elasticsearch: 
http://www.elasticsearch.org/downloads/1-4-2/
• Assuming you are installing on a Linux
distribution, you can easily install using DEB or
RPM packages.

Conﬁguring Your Cluster
Install Elasticsearch on it’s own box or
VM!

Configuring Your Cluster (cont.)
• How much memory is available to the heap? 
Configure your Elasticsearch heap size (~half of available RAM,
defaults to 256M min and 1G max): 
 
ES_HEAP_SIZE=512m in /etc/default/elasticsearch
• Can the current process be swapped? 
Disable process swapping for Elasticsearch: 
 
bootstrap.mlockall: true in config/elasticsearch.yml 
 
“Linux kernel tries to use as much memory as possible for file system caches and
swaps out unused application memory, possibly resulting in the Elasticsearch
process being swapped. Swapping is very bad for performance and for node
stability.” See: http://www.elasticsearch.org/guide/en/elasticsearch/reference/current/
setup-configuration.html.

• How much memory can the current system user acquire?  
Raise resources that system user can acquire: 
 
In /etc/security/limits.conf: 
elasticsearch - memlock unlimited 
 
In /etc/default/elasticsearch 
MAX_LOCKED_MEMORY=unlimited
• How many files can the current system user have open?  
Raise *nix file limit (65535 is ES default): 
 
In /etc/security/limits.conf: 
elasticsearch - nofile 65535

• Can Elasticsearch execute dynamic scripts in
queries? 
Disable dynamic scripting (potentially
dangerous and usually unnecessary) 
 
In /etc/elasticsearch/elasticsearch.yml 
script.disable_dynamic: true

Start Your Server
• Assuming you installed Elasticsearch as a
service on a Linux/Unix distribution: 
 
sudo service elasticsearch start

Hello World!
• You should be able to view http://
yourdomainorIP:9200 and see some JSON.
• If this doesn’t work, try SSH’ing into your ES
server and doing a cURL to localhost: 
curl http://localhost:9200

Cluster Health
• View http://yourdomainorIP:9200/_cluster/health 
 
You will see something like this: 
{ 
"cluster_name":"elasticsearch", 
"status":"yellow", 
"timed_out":false, 
“number_of_nodes":1, 
"number_of_data_nodes":1, 
“active_primary_shards":5, 
"active_shards":5, 
"relocating_shards":0, 
"initializing_shards":0, 
"unassigned_shards":5 
}

Cluster Health (cont.)
• “active_primary_shards”: Number of primary
(non-replica) shards that are active and available.
• shard: Every document (post) lives on a shard.
Indexes are divided into shards (5 by default).
There are primary and replica shards. Replica
shards help with failover. Shards are distributed
across nodes evenly and help with parallel
processing. A shard replica must live on a
different node than the primary.

• “cluster_name”: Elasticsearch is meant to be a
distributed search database. A cluster is one or
more nodes.
• Node: A single server on your cluster. Nodes
can be set up to store data or as master nodes.
• “number_of_nodes”: This is the number of
nodes in your cluster.

• “status”: There are three statuses possible: 
 
Red: Not all primary shards are available 
 
Yellow: Primary shards are available but
replicas are not. 
 
Green: Cluster is fully operational with primary
and replica shards

• Cluster status is yellow because there is only
one node. Replica shards need to be assigned
to a different node than their primary; this is not
possible in our current setup.
• It’s good practice to create at least 2 nodes in
our production clusters.

Security
• Elasticsearch has no concept of security,
authentication, or authorization built-in.
• By default, your instance is open to the world.
Good for development, but bad for production.

Private Instance
• We can easily lock down your ES instance 
 
In /etc/elasticsearch/elasticsearch.yml 
network.host: localhost!
• Note: this will make it difﬁcult to navigate your ES
instance in your web browser which is useful for
debugging. We can also do something like
setup a reverse proxy (using Nginx) with a
password protected endpoint to get around this.

Now that we have gone through
Elasticsearch basics, let’s talk
about integrating Elasticsearch
with WordPress.

What is ElasticPress?
https://github.com/10up/elasticpress

ElasticPress
A free 10up WordPress plugin that powers
WordPress search with Elasticsearch.

ElasticPress
• Build ﬁlterable performant queries (ﬁlter by taxonomy
term, post meta key, etc.)
• Fuzzy search post title, content, excerpt, taxonomy
terms, post meta, and authors
• Search across multiple blogs in a multi-site instance
• Results returned by relevancy. Relevancy calculations
are highly customizable.
• Very extensible and performant

ElasticPress Requirements
!
• WordPress 3.7+
• A host (not WordPress.com VIP) that either gives
SSH access or the ability to run WP-CLI
commands.
• An instance of Elasticsearch.
• WP-CLI

Installation
• Github: http://github.com/10up/elasticpress
• WordPress.org: http://wordpress.org/plugins/
elasticpress

Point to Your ES Instance
• In your wp-config.php file, add the following
where the URL is the domain or IP of your
Elasticsearch instance: 
 
define( 'EP_HOST', 'http://192.168.50.4:9200' );

Index Your Posts
• Just activating the plugin will do nothing. We need
to run a WP-CLI command: 
 
wp elasticpress index --setup [--network-wide]!
• --network-wide will force indexing across all the
blogs on a network of sites in multisite. This is
required if you plan to do cross-site search. It’s
basically a shortcut so you don’t have to run the
command once with the --url parameter for each
of your blogs.

What Happens Next?
• Once ElasticPress is activated and posts are
indexed, it will integrate with WP_Query to run
queries against Elasticsearch instead of MySQL.
• WP_Query integration will only happen on
search queries (“s” parameter) or when the
ep_integrate parameter is passed.

Query Integration
new WP_Query( array( 
’s’ => ‘search terms’, 
‘author_name’ => ‘taylor’, 
… 
) );!
’ep_integrate’ => ‘true’, 
… 
) );!
… 
) );

Note on Search
• We can use ElasticPress to power queries
OTHER than search! This is powerful.

Advanced Queries
• ElasticPress uses Elasticsearch to do many cool types of queries in a
performant manner. By passing special params to a WP_Query object we can:
• Search taxonomy terms
• Filter by taxonomy terms (unlimited dimensions)
• Search post meta
• Filter by post meta (unlimited dimensions)
• Search authors
• Filter by authors
• Search across blogs in multisite
• more!

Example Queries
’s’ => ‘paris france’, 
‘sites’ => ‘all’, 
) );

Example Queries
‘search_ﬁelds’ => array( 
‘post_title’, 
‘post_content’, 
‘taxonomies’ => array( ‘category’ ), 
), 
) );

Example Queries
‘post_type’ => ‘page’, 
‘meta’ => array( ‘city_name’ ), 
), 
) );

Example Queries
‘tax_query’ => array( 
array( 
‘taxonomy’ => ‘category’, 
‘terms’ => array( ‘term1’, ‘term2’ ), 
), 
), 
‘meta’ => array( ‘city_name’ ), 
‘author_name’, 
), 
‘site’ => 3, 
) );

WP_Query Integration
• The goal for ElasticPress is to make WP_Query
work behind the scenes through Elasticsearch
and provide extra functionality.
• This means we have to support every query
parameter which isn’t the case yet. Github
contains a full list of parameters WP_Query
supports with ElasticPress and usage for each
parameter.

Elasticsearch in Your Language
• Elasticsearch is designed to be
internationalized.
• Out of the box, it will work ﬁne with most
languages.

Analysis and Analyzers
• When a document is indexed in Elasticsearch,
text is analyzed, broken into terms (tokenized),
and normalized with token ﬁlters.
• In normalization, strings might be lowercased
and plurals stripped.

Custom Analyzers
• ElasticPress by default uses a pretty standard set of
analyzers intended for the English language.
• We can easily customize our analyzers for use with
other languages by filtering ep_config_mapping
(see EP source code).!
• You can read about language specific analyzers here: 
 
http://www.elasticsearch.org/guide/en/elasticsearch/
reference/current/analysis-lang-analyzer.html

Documentation
Full documentation with installation instructions: 
 
https://github.com/10up/ElasticPress

Feedback and Continuing Development
• If you are using ElasticPress on a project, please
let us know and give us feedback!
• Pull requests are welcome!

Questions?
We need to send a PUT request to this endpoint with
our post data. Of course we must authenticate before
doing this.
@tlovett12!
taylor.lovett@10up.com!
taylorlovett.com

Modernizing WordPress Search with Elasticsearch

More Related Content

What's hot

Viewers also liked

Similar to Modernizing WordPress Search with Elasticsearch

Recently uploaded

In this document

Modernizing WordPress Search with Elasticsearch