Ryan Bigg

Truncating indexes in Elasticsearch 2

22 Feb 2016

In pre-2.0 versions of elasticsearch, you may have been truncating indexes using delete_by_query. This has been moved in Elasticsearch 2.0 out to a plugin, which can be installed with:

bin/plugin install delete-by-query

Where bin/plugin is located wherever you installed Elasticsearch. On my Mac, that path is /usr/local/Cellar/elasticsearch/2.2.0_1/libexec because I installed Elasticsearch with Homebrew. If you're on Ubuntu and you installed it from Elasticsearch's own package repo, it will be at /usr/share/elasticsearch.

I'm using the Elasticsearch Ruby gem to interact with my Elasticsearch instance, and this is the code I'm using to truncate the indexes:

module Index
  class Indexers
    def truncate_indexes
      client = Elasticsearch::Client.new
      client.delete_by_query(index: index_names, body: { query: { match_all: {} } })
      client.indices.flush(index: index_names)

This method is called after each spec has finished running. The index_names is just a list of indexes that the Index::Indexers class knows about. The delete_by_query functionality is available here from the delete-by-query plugin that was installed. The indices.flush call is necessary because the next query may return data that the delete_by_query request was supposed to delete. The flush call clears the transaction log and memory and writes data to disk, ensuring that the next query won't return data that was deleted.

blog comments powered by Disqus