Breaking changes in 6.5

This section discusses the changes that you need to be aware of when migrating your application to Elasticsearch 6.5.

See also Release Highlights and Release Notes.

To make the node name appear consistently in Elasticsearch’s logs and to clean up much of the internal logging code we’ve removed the node name from the "marker" in the logging configuration. If you do not customize your logging configuration file then there is nothing for you to do. If you do customize your logging configuration properties then you should replace %marker with `[%node_name]%marker `, including the trailing space. If you do not do so Elasticsearch will log a warning on startup and log with the new pattern. It will not change the logging configuration files though. You should make this change before 7.0 because in 7.0 Elasticsearch will no longer automatically add the node name to the logging configuration if it isn’t already present.

Using rescore with a scroll query now raises a deprecation warning and ignores the parameter. In earlier 6.x releases, rescore on scroll queries was silently ignored. In 7.0 and later, we will return a 400 - Bad Request with a validation error.

An additional group will be present in the result of requests containing a GROUP BY for a column that has missing values in the returned documents. The records with missing values in the grouped by column will be collectively considered a single bucket.

Settings is no longer required by Retry so the withBackoff methods that take Settings are now deprecated and there are new versions of those methods that do not require Settings.

Custom role providers previously accepted an ActionListener with a response type of Set<RoleDescriptor>, but in 6.5.0 the role provider requires an ActionListener with a response type of RoleRetrievalResult. The RoleRetrievalResult object accepts a Set<RoleDescriptor> if the provider was successful; if the provider was not successful the RoleRetrievalResult should be populated with the error.

In 6.5.0 a new format for the logfile auditing has been introduced. The new file format prints audit entries as JSON documents. Entries in the previous format contained positional fields. These fields are no longer positional in the new format, as they are part of the structured log entry (the JSON document). Consequently, the following settings, which toggled some of the prefix positional fields, have been renamed ("prefix" was spliced out):

In 7.0.0 the prefix variant will be removed.