documentation/content/en/admin/optional/elasticsearch.md

title	description	menu
Full-text search	Setting up ElasticSearch to search for statuses authored, favourited, or mentioned in.	docs weight parent 10 admin-optional

Mastodon supports full-text search when it ElasticSearch is available. Mastodon’s full-text search allows logged in users to find results from their own toots, their favourites, and their mentions. It deliberately does not allow searching for arbitrary strings in the entire database.

Installing ElasticSearch

ElasticSearch requires a Java runtime. If you don’t have Java already installed, do it now. Assuming you are logged in as root:

apt install openjdk-8-jre-headless

Add the official ElasticSearch repository to apt:

wget -qO - https://artifacts.elastic.co/GPG-KEY-elasticsearch | apt-key add -
echo "deb https://artifacts.elastic.co/packages/6.x/apt stable main" | tee -a /etc/apt/sources.list.d/elastic-6.x.list
apt update

Now you can install ElasticSearch:

apt install elasticsearch

{{< hint style="warning" >}} Security warning: By default, ElasticSearch is supposed to bind to localhost only, i.e. be inaccessible from the outside network. You can check which address ElasticSearch binds to by looking at network.host within /etc/elasticsearch/elasticsearch.yml. Consider that anyone who can access ElasticSearch can access and modify any data within it, as there is no authentication layer. So it’s really important that the access is secured. Having a firewall that only exposes the 22, 80 and 443 ports is advisable, as outlined in the main installation instructions. If you have a multi-host setup, you must know how to secure internal traffic. {{< /hint >}}

To start ElasticSearch:

systemctl enable elasticsearch
systemctl start elasticsearch

Configuring Mastodon

Edit .env.production to add the following variables:

ES_ENABLED=true
ES_HOST=localhost
ES_PORT=9200

If you have multiple Mastodon servers on the same machine, and you are planning to use the same ElasticSearch installation for all of them, make sure that all of them have unique REDIS_NAMESPACE in their configurations, to differentiate the indices. If you need to override the prefix of the ElasticSearch index, you can set ES_PREFIX directly.

After saving the new configuration, create the index in ElasticSearch with:

RAILS_ENV=production bundle exec rake chewy:upgrade

Then restart Mastodon processes for the new configuration to take effect:

systemctl restart mastodon-sidekiq
systemctl reload mastodon-web

Now new statuses will be written to the ElasticSearch index. The last step is importing all of the old data as well. This might take a long while:

RAILS_ENV=production bundle exec rake chewy:sync

{{< hint style="warning" >}} Compatibility note: There is a known bug in Ruby 2.6.0 that prevents the above task from working. Other versions of Ruby, such as 2.6.1, are fine. {{< /hint >}}

Search optimization for other languages

Chinese search optimization

The default analyzer of the ElasticSearch is the standard analyzer, which may not be the best especially for Chinese. To improve search experience, you can install language specific analyzer. Before create the index in ElasticSearch:

Installing elasticsearch-analysis-ik, elasticsearch-analysis-stconvert to ElasticSearch.

And do modify as follows:

diff --git a/app/chewy/accounts_index.rb b/app/chewy/accounts_index.rb
--- a/app/chewy/accounts_index.rb
+++ b/app/chewy/accounts_index.rb
@@ -4,7 +4,7 @@ class AccountsIndex < Chewy::Index
   settings index: { refresh_interval: '5m' }, analysis: {
     analyzer: {
       content: {
-        tokenizer: 'whitespace',
+        tokenizer: 'ik_max_word',
         filter: %w(lowercase asciifolding cjk_width),
       },
 
diff --git a/app/chewy/statuses_index.rb b/app/chewy/statuses_index.rb
--- a/app/chewy/statuses_index.rb
+++ b/app/chewy/statuses_index.rb
@@ -16,9 +16,17 @@ class StatusesIndex < Chewy::Index
         language: 'possessive_english',
       },
     },
+    char_filter: {
+      tsconvert: {
+        type: 'stconvert',
+        keep_both: false,
+        delimiter: '#',
+        convert_type: 't2s',
+      },
+    },
     analyzer: {
       content: {
-        tokenizer: 'uax_url_email',
+        tokenizer: 'ik_max_word',
         filter: %w(
           english_possessive_stemmer
           lowercase
@@ -27,6 +35,7 @@ class StatusesIndex < Chewy::Index
           english_stop
           english_stemmer
         ),
+        char_filter: %w(tsconvert),
       },
     },
   }
diff --git a/app/chewy/tags_index.rb b/app/chewy/tags_index.rb
--- a/app/chewy/tags_index.rb
+++ b/app/chewy/tags_index.rb
@@ -2,10 +2,19 @@
 
 class TagsIndex < Chewy::Index
   settings index: { refresh_interval: '15m' }, analysis: {
+    char_filter: {
+      tsconvert: {
+        type: 'stconvert',
+        keep_both: false,
+        delimiter: '#',
+        convert_type: 't2s',
+      },
+    },
     analyzer: {
       content: {
-        tokenizer: 'keyword',
+        tokenizer: 'ik_max_word',
         filter: %w(lowercase asciifolding cjk_width),
+        char_filter: %w(tsconvert),
       },
 
       edge_ngram: {