https://www.testcontainers.org/modules/elasticsearch/
This project is no longer maintained. We recommend using Testcontainers instead:embedded-elasticsearch
Small utility for creating integration tests that use Elasticsearch. Instead of using Node
it downloads Elasticsearch in specified version and starts it in a separate process. It also allows you to install required plugins which is not possible when using NodeBuilder
. Utility was tested with 1.x, 2.x, 5.x and 6.x versions of Elasticsearch.
Introduction
All you need to do to use this tool is create EmbeddedElastic
instance. To do so, use provided builder:
final embeddedElastic = EmbeddedElastic.builder()
.withElasticVersion("5.0.0")
.withSetting(PopularProperties.TRANSPORT_TCP_PORT, 9350)
.withSetting(PopularProperties.CLUSTER_NAME, "my_cluster")
.withPlugin("analysis-stempel")
.withIndex("cars", IndexSettings.builder()
.withType("car", getSystemResourceAsStream("car-mapping.json"))
.build())
.withIndex("books", IndexSettings.builder()
.withType(PAPER_BOOK_INDEX_TYPE, getSystemResourceAsStream("paper-book-mapping.json"))
.withType("audio_book", getSystemResourceAsStream("audio-book-mapping.json"))
.withSettings(getSystemResourceAsStream("elastic-settings.json"))
.build())
.build()
.start()
When you are done with creating it, you can start it real simple:
embeddedElastic.start()
And that's all, you can connect to your embedded-elastic instance on specified port and use it in your tests.
Available builder options
Method | Description |
---|---|
withElasticVersion(String version) |
version of Elasticsearch; based on that version download url to official Elasticsearch repository will be created |
withDownloadUrl(URL downloadUrl) |
if you prefer to download Elasticsearch from a different location than official repositories you can do that using this method |
withInResourceLocation(String inResourcePath) |
you can also have your Elasticsearch package inside resource directory, specify it's location with this option |
withSetting(String key, Object value) |
setting name and value as in elasticsearch.yml file |
withPlugin(String expression) |
plugin that should be installed into Elasticsearch; treat expression as argument to ./elasticsearch-plugin install <expression> command; use multiple times for multiple plugins |
withIndex(String indexName, IndexSettings indexSettings) |
specify index that should be created and managed by EmbeddedElastic |
withTemplate(String templateName, String templateBody) |
specify a template that should be created and managed by EmbeddedElastic |
withStartTimeout(long value, TimeUnit unit) |
specify timeout you give Elasticsearch to start |
withInstallationDirectory(File installationDirectory) |
specify custom installation directory |
withDownloadDirectory(File downloadDirectory) |
specify custom download directory where downloaded distribution packages will be saved |
withCleanInstallationDirectoryOnStop(boolean cleanInstallationDirectoryOnStop) |
specify whether clean the installation directory after Elasticsearch stop |
withEsJavaOpts(String javaOpts) |
value of ES_JAVA_OPTS variable to be set for Elasticsearch process |
withJavaHome(JavaHomeOption javaHomeOption) |
select java environment to run in. For available options see below |
getTransportTcpPort() |
get transport tcp port number used by Elasticsearch instance |
getHttpPort() |
get http port number used by Elasticsearch instance |
withDownloadProxy(Proxy proxy) |
proxy that should be used for downloading Elasticsearch package |
withDownloaderConnectionTimeout(long value, TimeUnit unit) |
connection timeout that should be used by downloader |
withDownloaderReadTimeout(long value, TimeUnit unit) |
socket timeout that should be used by downloader |
Available IndexSettings.Builder
options
Method | Description |
---|---|
withType(String type, String mapping) |
specify type and it's mappings |
withSettings(String settings) |
specify index settings |
Availabe JavaHomeOption
options
Method | Description |
---|---|
useSystem() |
default behavior, lets elasticsearch startup script determine the JRE to run in |
inheritTestSuite() |
use the same JRE as the process starting it |
path(String path) |
manually set the path of the JRE to execute the embedded elastic |
Available operations
EmbeddedElastic
provides following operations:
Method | Description |
---|---|
start() |
downloads Elasticsearch and specified plugins, setups everything and finally starts your Elasticsearch instance |
stop() |
stops your Elasticsearch instance and removes all data |
index |
index your document, comes with variants that take only document, or document and it's id |
deleteIndex(String indexName) , deleteIndices() |
deletes index with name specified during EmbeddedElastic creation |
createIndex(String indexName) , createIndices() |
creates index with name specified during EmbeddedElastic creation; note that this index is created during EmbeddedElastic startup, you will need this method only if you deleted your index using deleteIndex method |
recreateIndex(String indexName) , recreateIndices() |
combination of deleteIndex and createIndex |
refreshIndices() |
refresh index; useful when you make changes in different thread, and want to check results instantly in tests |
deleteTemplate(String templateName) , deleteTemplates() |
deletes a template of the specified name/all templates during EmbeddedElastic creation |
createTemplate(String templateName) , createTemplates() |
creates a template with the specified name/all templates during EmbeddedElastic creation; note that this template is created during EmbeddedElastic startup, you will need this method only if you deleted your template using deleteTemplate method |
recreateTemplate(String templateName) , recreateTemplates() |
combination of deleteTemplate and createTemplate |
Example
If you want to see example, look at this spec: pl.allegro.tech.search.embeddedelasticsearch.EmbeddedElasticSpec
Dependency
To start using embedded-elasticsearch in your project add it as a test dependency:
Gradle:
testCompile 'pl.allegro.tech:embedded-elasticsearch:2.7.0'
Maven:
<dependency>
<groupId>pl.allegro.tech</groupId>
<artifactId>embedded-elasticsearch</artifactId>
<version>2.7.0</version>
<scope>test</scope>
</dependency>
SBT:
libraryDependencies ++= Seq("pl.allegro.tech" % "embedded-elasticsearch" % "2.7.0" % "test")
Known problems
If you build your project on Travis, you may have problems with OOM errors when using default settings. You can change Elasticsearch memory settings using withEsJavaOpts
method. Example (from spec pl.allegro.tech.embeddedelasticsearch.EmbeddedElasticSpec
):
static EmbeddedElastic embeddedElastic = EmbeddedElastic.builder()
.withElasticVersion(ELASTIC_VERSION)
.withSetting(TRANSPORT_TCP_PORT, TRANSPORT_TCP_PORT_VALUE)
.withSetting(CLUSTER_NAME, CLUSTER_NAME_VALUE)
.withEsJavaOpts("-Xms128m -Xmx512m")
.withIndex(CARS_INDEX_NAME, CARS_INDEX)
.withIndex(BOOKS_INDEX_NAME, BOOKS_INDEX)
.withStartTimeout(1, MINUTES)
.build()
.start()
Running more then one Elasticsearch instance
There are cases where you might want to run more than one Elasticsearch instance e.g.:
- running tests of one project in parallel (e.g. using gradle --parallel or mvn -T1C)
- running tests of different projects on the same physical mashine (e.g. Jenkins jobs running on the same server)
- running integration tests which require more than one Elasticsearch instance
In such situations you should use distinct values for following settings for each instance:
withSetting(PopularProperties.TRANSPORT_TCP_PORT, ...)
withSetting(PopularProperties.HTTP_PORT, ...)
withInstallationDirectory(...)
With such configuration embedded-elasticsearch will redownload elasticsearch installation package for every distinct
installation directory. To avoid this behavior and reuse downloaded installation package you should
set common location of downloaded files with withDownloadDirectory(...)
for every embedded-elasticsearch configuration.
License
embedded-elasticsearch is published under Apache License 2.0.