Groonga CloudSearch blog
Today, we’ve shipped Groonga CloudSearch version 1.5.0 and Groonga CloudSearch Console 1.2.0. You can download the packages and read the tutorial on the home page of this project site.
What’s new? : Groonga CloudSearch
Better compatibility to the Amazon CloudSearch
- Id part of “DocService/Arn”, “SearchService/Arn”, and “DomainId” of “DomainStatus” are now returned with the value different from the domain id string in the endpoint host name correctly. You always need to refer “DocService/Endpoint” or “SearchService/Endpoint” to know endpoint host names.
- “DefineIndexField” never merges newly specified options with previously defined options. (In other words, now “DefineIndexField” always cleared old options at old versions.)
- XML SDF batches are now acceptable. (Note: all responses from documents/batch API are still returned in JSON.)
CLI tool changes
gcs-configure-fields command becomes more compatible to
cs-configure-fields. Existing field options are always cleared, so you have to set multiple field options at once. For this change, now the command accepts multiple ”–option” options, like:
$ gcs-configure-fields -d domain –name field –type literal –option search result
$ gcs-configure-fields -d domain –name field –type literal –option search –option result
gcs-configure-from-sdf is also now available. You can define multiple index fields from existing SDF file.
Changes as a daemon
Because you always have to call APIs via HTTP, now the ”–base-host” command line option for bin/gcs and GCS_BASE_HOST environment variable are obsolete. It is automatically detected from the Host header. You don’t need to give a base host name of the server to Groonga CloudSearch processes anymore.
Now Groonga CloudSearch creates the directory to store data with the permission 0700 (full access by author, otherwise no access). If you use old versions of Groonga CloudSearch, you should set the permission of the data directory (~gcs/.gcs, ~/.gcs, and so on) to 0700 to the data directory by the
What’s new? : Groonga CloudSearch Console
Groonga CloudSearch Console now activates BASIC Auth always. So it asks you to register account name and password for the administration user on the initial access. The web console can’t be accessed without authentication anymore.
Account information is stored in the directory ~/.gcs-console by default. If you need to reset the password, remove the directory.
Flexible ”–endpoint” option
Simple host name is now acceptable for the ”–endpoint” option (and GCS_CONSOLE_ENDPOINT environment variable) like ”–endpoint gcs-console.example.com:7575”. In old versions, you always had to specify full URL like “http://gcs-console.example.com:7575”.
Now we are planning to suspend this project and concentrate human resources to the fluent-plugin-groonga project for scalable backend based on groonga. After that, we’ll be back and work to improve Groonga CloudSearch with scalable backends.
Currently Groonga CloudSearch doesn’t support default value, source attribute, and more features of Amazon CloudSearch. We have to re-construct the database layer of Groonga CloudSearch to support these features.
On the other hand, now groonga project has started a search system based on scalable backends, with fluentd. It is the most important feature of Amazon CloudSearch but not supported by Groonga CloudSearch.
Because both changes require many many codes around the database layer. We’ll need to rewrite codes for the scalable backends in the feature, even if we now implement default value and other features. So we’ve decided to suspend this Groonga CloudSearch project for a while and concentrate development resources to the scalable backend.