Release notes

The release notes provide information about changes to the Retrieve and Rank service since the previous release.

Known issues

  • The cluster sizing enhancements introduced on 30 June 2016 are not fully enabled. Cluster sizes of 8 to 14 units are not currently available. You can still resize a cluster of 1 to 7 units provided that your solrconfig.xml file contains the following new line, as listed in Configuring the Retrieve and Rank service.

    <queryParser name="serializedQueryParser" class="" />

    The example solrconfig.xml file provided with the configuration has been updated with the required new entry.

  • The Try it out! button on the Upload Solr configuration (/v1/solr_clusters/{solr_cluster_id}/config/{config_name}) command in the Retrieve and Rank API Explorer does not work. This is due to a limitation in the Swagger code used to create the API Explorer. The limitation prevents zip files from being uploaded.

    To upload a Solr configuration, see the Upload Solr configuration example in the API Reference.

Updates as of 21 September 2016

The cURL command for the search and search-and-rank methods has been changed from GET to POST. The POST command does not encounter errors that the GET command can sometimes encounter with Solr searches. The change is documented in the Retrieve and Rank API Reference and API Explorer.

Previous updates

This section lists issues that were addressed in previous versions of this document.

Updates as of 3 August 2016

The Retrieve and Rank Web Interface is now compatible with a ranker trained by using the curl and commands described on the Tutorial page. To enable compatibility, use the refreshed versions of the and, if needed, the files to configure your cluster.

30 June 2016

  • New APIs to show Solr cluster statistics and to resize Solr clusters. See Sizing your Retrieve and Rank cluster for details. Be sure to note the limitations of the resizing API before attempting to resize a cluster. [22 June 2016]

  • The ability to display the ranker-confidence score for the results of a query to a trained ranker. See Working with ranker confidence scores for information.

  • The introduction of Retrieve and Rank tooling to simplify working with the service. See the linked documentation for details.

    • Enhanced Information Retrieval provides command-line tools for setting up a Retrieve and Rank instance and uploading documents to a search index.
    • The Retrieve and Rank Web Interface provides a web interface for setting up a Retrieve and Rank instance, uploading documents, and answering questions to train a ranker.
  • Enhanced service-specific support (tokenization) for German in addition to previously supported languages. See Configuring the Retrieve and Rank service for details.

29 April 2016

  • There is a new sample application that uses the Retrieve and Rank service. See the Overview for details.

28 March 2016

  • The Retrieve and Rank service now provides enhanced service-specific support (tokenization and lemmatization) for Japanese as well as for the languages listed in the updates for 7 January 2016.

  • The Retrieve and Rank documentation incorrectly listed the maximum size of a ranker's training-data set as 300 GB (gigabytes). The correct maximum size is 300 MB (megabytes). The documentation has been corrected.

  • The output listed in the API Explorer application for the "List Solr configurations" method (GET /v1/solr_clusters/{solr_cluster_id}/config) does not match the current output of the method. Although the syntax listed in the API Explorer application is correct according to the specification, a known software defect displays the output incorrectly.

    The API Explorer application lists the following output format for the method:

      "solr_configs": [
          "config_name": "string"

    The current version of the method lists the following output format: {"solr_configs":["string"]} where string is the name or names of the Solr configuration or configurations.

    The output of the method is planned to be corrected in an upcoming release.

10 February 2016

The Retrieve and Rank service now enforces Apache Solr's recommendation that newly created collection names use only alphanumeric characters (a-z, A-Z, and 0-9), periods (.), and underscores (_). If you use a collection name that includes other characters, the service returns an error.

Note: The character constraints apply only to newly created collections. If you already have collections that include other characters such as dashes (-), you can continue to administer them by using the existing APIs without changing their names or deleting and recreating them.

The Retrieve and Rank API reference has been updated with this information.

4 February 2016

The 4 February 2016 release included information about the Java API for the Retrieve and Rank service in the service documentation's API reference.

14 January 2016 (documentation erratum)

The Retrieve and Rank service documentation incorrectly stated that the maximum number of Solr servers in a cluster was 14. The correct maximum number is seven (7) servers in a cluster. The documentation was updated with the correct information.

7 January 2016

The 7 January 2016 release included the following new features:

  • Enhanced service-specific support (tokenization and lemmatization) for the following languages in addition to English:

    • Arabic (ar)
    • Portuguese (ptbr)
    • Spanish (es)

    For information on using these languages with the Retrieve and Rank service, see Configuring Retrieve and Rank.

    Note: The Retrieve and Rank service supports all of the same languages that standard Solr supports. However, the service provides enhanced support only for the languages listed above.

  • The Retrieve and Rank service no longer blocks the RELOAD call from the Solr Collections API.