Oak supports Elasticsearch (Elastic for short) based indexes for both property constraint and full text constraints. Elastic indexes support similar features as Lucene indexes, however there are differences:
type
is elasticsearch
./oak:index
. Other locations are not supported.async
property needs to be set to elastic-async
. Synchronous, nrt
or other lanes are not supported. Indexes are updated asynchronously.refresh
is ignored. Changes take effect immediately after changing them. Existing documents in Elasticsearch are not changed.reindex
property to true
or by using the oak-run
tool. We recommend to build them using the oak-run
tool.evaluatePathRestrictions
cannot be disabled. The parent paths are always indexed. Queries with path restrictions are evaluated at index level when possible, otherwise they are evaluated at repository level.codec
is ignored.compatVersion
is ignored.useIfExists
is ignored.blobSize
is ignored.name
is ignored.indexPath
is ignored.analyzers
support the Lucene configuration plus Elasticsearch specific options. Since Elasticsearch uses a more recent version of Lucene compared to the one in oak-lucene
module, there might be differences in configuration options that could require changes when migrating from Lucene to Elasticsearch.useInExcerpt
does not support regexp relative properties.sync
and unique
are ignored. Synchronous indexing, and enforcing uniqueness constraints is not currently supported in elastic indexes.dynamicBoost
is slightly different: For Lucene indexes, boosting is done in indexing, while for Elastic it is done at query time.suggest
is slightly different: For Lucene indexes, the suggestor is updated every 10 minutes by default and the frequency can be changed by suggestUpdateFrequencyMinutes
property in suggestion node under the index definition node. In Elastic indexes, there is no such delay and thus no need for the above config property. This is an improvement in ES over lucene.