ArangoDB River Plugin for ElasticSearch
Clone or download
Latest commit afdd6db Oct 20, 2018

This project is dead. ElasticSearch removed the River API in Version 2.0 on which this plugin relied. To read the ArangoDb Replication Log and load the data into your own ElasticSearch (or any other datastore), have a look at

ArangoDB River Plugin for ElasticSearch

This is an ElasticSearch plugin that will connect to your ArangoDB server, read the Write-Ahead-Log of a collection and update an ElasticSearch index with the data.

Build Status

Branch Status
Master master
Devel devel

Version Compatibility

Release ArangoDB Version ElasticSearch built ElasticSearch integration-tested
1.0.1 2.[6-8].x 1.4.2 1.4 through 1.7
  • ElasticSearch built is the ES version against which the release was built
  • ElasticSearch integration-tested are the ES versions against which the integration tests work

The ArangoDB river artifact is named elasticsearch-river-arangodb-<version>.jar.


The zip file contains all necessary dependencies and nothing else is needed to install it in ElasticSearch.

Install from URL

/usr/share/elasticsearch/bin/plugin \
    --install arangodb \

install from file (after manual download)

/usr/share/elasticsearch/bin/plugin \
    --install arangodb \
    --url file:///${HOME}/Downloads/

Manual install

To install manually, unzip the archive into the ES plugins directory see: 'paths.plugin' config value of ES. The file structure should look like this:

$ cd <ES plugins directory>
$ find . -type f
./arangodb/... more jars (dependencies of the plugin) ...

Remove Plugin

Removing the plugin is required for upgrades.

/usr/share/elasticsearch/bin/plugin --remove arangodb

Start & Configuration

To create the river, run a curl statement from a shell:

curl -XPUT 'http://localhost:9200/_river/arangodb/_meta' -d '{
    "type": "arangodb",
    "arangodb": {
        "db": "<DATABASE_NAME>",
        "collection": "<COLLECTION>",
    "index": {
        "name": "<ES_INDEX_NAME>",
        "type": "<ES_TYPE_NAME>"


curl -XPUT 'http://localhost:9200/_river/arangodb_test_car/_meta' -d '{
    "type": "arangodb",
    "arangodb": {
        "db": "test",
        "collection": "car"
    "index": {
        "name": "cars",
        "type": "car"

Here is a more complex configuration with user credentials and a simple javascript example:

curl -XPUT 'http://localhost:9200/_river/arangodb_test_car2/_meta' -d '{
    "type": "arangodb",
    "arangodb": {
        "host": "carhost",
        "port": carport,
        "db": "test",
        "collection": "car",
        "credentials": {
            "username": "riveruser",
            "password": "rivauser"
        "reader_min_sleep": "100ms",
        "reader_max_sleep": "10s",
        "script" : "ctx.doc.title = ctx.doc.manufacturer + \" \" + ctx.doc.model;"
    "index": {
        "name": "cars",
        "type": "car",
        "bulk_size": 5000,
        "bulk_timeout": "500ms"


To use scripting to filter or manipulate documents before they are indexed, you must install the ElasticSearch scripting plugin. Please refer to the ElasticSearch Plugins documentation. Specifically, the scripting plugins.

Tips & Tricks

ArangoDB Authentication

To add a user for basic authentication, open an ArangoDB shell and run the following command:

require("org/arangodb/users").save("<username>", "<password>");

This will create a user with the given user name and the given password. Currently, ArangoDB just provides user authentication, but no authorization on a collection or operation level.

When starting the database, you can set the ArangoDB daemon parameter

--server.disable-authentication true

if you like to run the database without any authentication mechanism.


Use scripting if you want to filter your data. Let's say your documents have a boolean field named "available". The following script will filter your data due to their availability flags:

"script" : "if ( ctx.doc.available == false ) { ctx.operation = 'SKIP' };"

This script checks the "available" flag, and it changes the operation to "SKIP" for the given document context. This document will not be indexed. Please note that, if the document is already indexed, this will mean it won't be updated or deleted.


  • extend integration test:
    • stop river, make a change in arangodb, must not be reflected in ES anymore
    • start river again, should pick up where we stopped or start 'now' which means we've lost the last update

Development & Contributing

If you want to change and test something, the basic development process is simple:

  • fork and clone repo, checkout devel branch
  • create your own branch from devel
  • eclipse users run gradle cleanEclipse eclipse
  • make your changes
  • run the integration tests (requires ArangoDB and ElasticSearch to be installed) with gradle clean test intTest
  • build a distribution with gradle clean dist (see directory build/distributions)
  • test your custom build in your environment
  • if happy commit and push, then create a pull request against the devel branch


This software is licensed under the Apache 2 license, see the supplied LICENSE file.



  • uses new version 1.1.8 or arangodb-wal-client, which supports new properties of the write ahead log that are introduced in arangodb 2.6


  • Works with ArangoDB versions of 2.2 and higher
  • Uses the Write-Ahead-Log and is not compatible with ArangoDB versions below 2.2
  • This is a complete rewrite of the existing code
  • Tested with ArangoDB 2.4 and ElasticSearch 1.4
  • The configuration and featureset is slightly different. Existing users please carefully re-read the documentation


  • Works with ArangoDB versions below 2.2
  • Works with ArangoDB versions of 2.2 and higher if the legacy Replication Logger is enabled
  • Update all dependencies to latest versions (ElasticSearch 1.4.x)
  • Complete Refactoring, Cleanups, Simplifications


  • Minor changes for ElasticSearch v1.0.0.


  • Initial alpha version.
  • Supports script filters.
  • Using Apache HTTP standard components.
  • No support yet for deleting or renaming collections.
  • Based on former river layouts like CouchDB and MongoDB.