Use Relevance without Elasticsearch

Introduction

Goal

Use Relevance without Elasticsearch at the expensive of giving up Trends and Experiments.

Background

In order to use the full functionality of the Relevance Module, Elasticsearch is required. However, as of Bloomreach Experience Manager 14.1.0 (also backported to 13.4.2 and 14.0.1), it is possible to use Relevance without Elasticsearch, at the expense of the Trends and Experiments functionalities.

In new implementation projects created using the Bloomreach Experience Manager 14.1.0 Maven archetype, this can be achieved through Essentials by installing only the Relevance feature and not the Relevance Experiments and Trends feature.

In existing projects that were created using an earlier Maven archetype, some manual steps are required to remove the Elasticsearch requirement. This page described those steps.

Prerequisites

The instructions below assume your implementation project was created using the Maven archetype, had Relevance added through Essentials, and has a platform properties configuration file (see HST Container Configuration for more details on platform properties).

Also, complete all running experiments and publish channel changes before you remove Elasticsearch from a production environment,

Remove ElasticSearch Requirement

In cms/src/main/webapp/WEB-INF/hst-config.properties), add the property:

targeting.elastic.disabled = true

The above makes Relevance run without ElasticSearch. The trade-off is that the Trends and Goals tabs is not available any more in the Content audiences application and that you cannot create or start an experiment any more in the Experience manager: the experiments feature is not available any more. 

Further Cleanup

Without the ElasticSearch requirement, more configuration can be removed from the project.

In conf/context.xml, the elasticsearch environment variable can be removed:

<Environment name="elasticsearch/targetingDS"
                          value="{'indexName':'visits', 'locations':['http://localhost:9200/']}"
                          type="java.lang.String"/>

For local development, the elasticsearch plugin can be removed from the root pom.xml so that it does not start any more with cargo.run. To achieve this, remove from the root pom.xml:

<plugin>
  <groupId>com.github.alexcojocaru</groupId>
  <artifactId>elasticsearch-maven-plugin</artifactId>
  <version>${maven.plugin.elasticsearch.version}</version>
  <executions>
    <execution>
      <id>start-elasticsearch</id>
      <phase>validate</phase>
      <goals>
        <goal>runforked</goal>
      </goals>
    </execution>
    <execution>
      <id>stop-elasticsearch</id>
      <phase>post-integration-test</phase>
      <goals>
        <goal>stop</goal>
      </goals>
    </execution>
  </executions>
  <configuration>
    <clusterName>esDevCluster</clusterName>
    <httpPort>${es.httpPort}</httpPort>
    <transportPort>${es.tcpPort}</transportPort>
    <version>${maven.plugin.elasticsearch.configuration.version}</version>
    <keepExistingData>true</keepExistingData>
    <timeout>60</timeout>
    <skip>${es.skip.start}</skip>
  </configuration>
</plugin>

The Elasticsearch client dependencies can be excluded from hippo-addon-targeting-cms-dependencies (or hippo-addon-targeting-platform-dependencies if you use that one).

Replace

<dependency>
  <groupId>com.onehippo.cms7</groupId>
  <artifactId>hippo-addon-targeting-cms-dependencies</artifactId>
  <type>pom</type>
</dependency>

with

<dependency>
  <groupId>com.onehippo.cms7</groupId>
  <artifactId>hippo-addon-targeting-cms-dependencies</artifactId>
  <type>pom</type>
  <exclusions>
    <exclusion>
      <groupId>com.onehippo.cms7</groupId>
      <artifactId>hippo-addon-targeting-state-elastic5</artifactId>
    </exclusion>
    <exclusion>
      <groupId>com.onehippo.cms7</groupId>
      <artifactId>hippo-addon-targeting-state-elastic6</artifactId>
    </exclusion>
  </exclusions>
</dependency>

Remove the following configuration from the repository:

/targeting:targeting/targeting:datastores/targeting:visits

(Optional) Docker Cleanup

In case of using Docker, in addition to the actions above, execute the following steps:

1. Remove the following element from context-mysql.xml & context-postgres.xml:

<Environment name="elasticsearch/targetingDS" type="java.lang.String"
 value="{'indexName':'visits', 'locations':['http://@es.host@:@es.port@/']}"/>

2. Remove the following instructions from mysql/setup-db.sh or postgres/setup-db.sh:

sed --in-place 's/@es.host@/'"$ELASTICSEARCH_HOST"'/' /usr/local/tomcat/conf/context-$profile.xml
sed --in-place 's/@es.port@/'"$ELASTICSEARCH_PORT"'/' /usr/local/tomcat/conf/context-$profile.xml

3. Remove previously added Maven 'docker.es' profile.