Migrating from Elasticsearch to Vespa

This document will show how to move data from Elasticsearch to Vespa. By the end of this guide you will have generated a deployable Vespa application package from your existing Elasticsearch cluster. To consider whether Vespa is a better choice for your use case, take a look at this comparison

1. Get all documents from Elasticsearch with ElasticDump

It is possible to use ElasticDump to get all documents from Elasticsearch in a JSON-file. Assuming starting in a empty folder.

$ git clone --depth 1 https://github.com/taskrabbit/elasticsearch-dump.git

Then get all documents and mapping from your cluster(s) with:

$ `pwd`/elasticsearch-dump/bin/elasticdump \
  --input=http://localhost:9200/my_index \
  --output=/path/to/empty/folder/my_index.json \

 $ `pwd`/elasticsearch-dump/bin/elasticdump \
  --input=http://localhost:9200/my_index \
  --output=/path/to/empty/folder/my_index_mapping.json \
  • --input should be the url to your Elasticsearch index
  • --output should be the path to your intially empty folder

2. Parse the ES-documents to Vespa-documents and generate an Application Package

Download ES_Vespa_parser.py here, and place it in your intitially empty directory.


$ ES_Vespa_parser.py [-h] [--application_name APPLICATION_NAME] documents_path mappings_path

Run this command in your folder to parse the documents, so that it can be feeded to Vespa:

$ python ES_Vespa_parser.py my_index.json my_index_mapping.json
  • --application_name defaults to “application_name” - just change if you want
    • The document ids will become id:application_name:doc_name::elasticsearch_id

The directory has now a folder application:

      ├── documents.json
      ├── hosts.xml
      ├── services.xml
      └── /searchdefinitions
            ├── sd1.sd
            └── ... 

Which contains your converted documents, their search definitions, a hosts.xml and a services.xml - a whole application package.

3. Deploying Vespa:

Go into your initially empty folder. This tutorial have been tested with a Docker container with 10GB RAM. We will map the this directory into the /app directory inside the Docker container. Now, to start the Vespa container:

 $ docker run -m 10G --detach --name vespa --hostname vespa-es-tutorial \
    --privileged --volume `pwd`:/app \
    --publish 8080:8080 --publish 19112:19112 vespaengine/vespa

Make sure that the configuration server is running:

 $ docker exec vespa bash -c 'curl -s --head http://localhost:19071/ApplicationStatus'

Deploy the application package:

$ docker exec vespa bash -c '/opt/vespa/bin/vespa-deploy prepare /app/application && \
    /opt/vespa/bin/vespa-deploy activate'

(or alternatively, run the equivalent commands inside the docker container). After a short while, pointing a browser to http://localhost:8080/ApplicationStatus returns JSON-formatted information about the active application. The Vespa node is now configured and ready for use.

For more detailed explanation of deploying application packages read this.

4. Feeding the parsed documents to Vespa:

Send this to Vespa using one of the tools Vespa provides for feeding. In this part of the tutorial, the Java feeding API is used:

$ docker exec vespa bash -c 'java -jar /opt/vespa/lib/jars/vespa-http-client-jar-with-dependencies.jar \
    --verbose --file /app/application/documents.json --host localhost --port 8080'

You can also inspect the search node state by: $ docker exec vespa bash -c '/opt/vespa/bin/vespa-proton-cmd --local getState'

5. Fetching the documents:

Fetch documents by document id using the Document API:

$ curl -s http://localhost:8080/document/v1/application_name/doc_name/docid/elasticsearch_id

6. The first query:

Feel free to use our GUI for building queries at http://localhost:8080/querybuilder/ (with Vespa-container running) which can help you building queries with e.g. autocompletion of YQL. Also take a look at Vespa’s Search API.

Click for more information about developing applications and application packages like application.

Please take a look at how to secure your Vespa installation


Vespa can be feeded with either Vespa Http Feeding Client or using Hadoop, Pig, Oozie.

The Vespa Http Feeding Client is a Java API and command line tool to feed document operations to Vespa. The Vespa feedig client allows you to combine high throughput with feedig over HTTP.

Enabling in your application

Add the <document-api> to a container cluster to enable it to receive documents:

<?xml version="1.0" encoding="utf-8" ?>
<services version="1.0">

  <container version="1.0" id="default">


Using from the command line

Use the API by running a binary. This binary supports feeding document operations and is installed with Vespa - found at $VESPA_HOME/lib/jars/vespa-http-client-jar-with-dependencies.jar.

Example of usage:

$ java -jar $VESPA_HOME/lib/jars/vespa-http-client-jar-with-dependencies.jar --file file.json --host gatewayhost --port 8080

You can also feed Using the Java Vespa Feeding Client API, which you can read more about here, where you also can find sample code.


For more experience with the Vespa Platform, feel free to do one of Vespas tutorials:


If you find errors, spelling mistakes, faulty pieces of code or want to improve the migration tutorial, please submit a pull request or create an issue