0.18.0 has just been released. You can download it here. This is a major release, and includes the following major features:
0.18.2 was quickly released to fix a bug with custom
path.data location configuration and fetching inner fields from a source doc.
Shard Allocation Awareness and Filtering
Shard allocation within a cluster can now be “aware” of which node replicas are allocated on (similar to rack awareness). It allows to make sure that a shard and a replica will be allocated across logical grouping of nodes. The awareness can also be forced, by not over allocating replicas within the same logical node group. Search and Get operations will automatically prefer shards that are allocated within the same logical node group the search is executing on.
Filtering allows to explicitly control which nodes indices are allowed, or not allowed, to be allocated on, again, based on custom logical grouping of nodes (called node attributes).
Both settings can be updated in realtime using either the cluster wide or index level update settings API. More info can be found here.
Cluster Update Settings API
The cluster update settings API allow to update node level settings (and not index level settings) in realtime on a live cluster. The settings allowed and more docs on the API are provided here.
Timestamp and TTL
A document can now automatically be indexed with a timestamp associated with it, and have a TTL (time to live), which, when expired, will cause the document to be deleted. More info can be found here and here.
Improved Geo Execution
Geo distance filter and facet good a considerable performance boost by doing bounding box optimization. Also, an option to use indexed lat lon for the checks (must be enabled in the
geo_type mapping) is provided which can provide even faster executing under certain conditions (compared to in memory checks).
More statistics are now provided out of the box, including a new index stats API (the status API should not be used for stats anymore). Statistics for APIs such as index, get, and search are also being aggregated, with search allowing to specific specific stats grouping aggregation (to further distinguish search “types”).
Multi Data locations
A node can now work with multiple data locations, spreading the index files across those locations in a RAID 0 like behavior.
Smaller improvements and bug fixes
Many more bug fixes and smaller improvement have went into this release. Full release notes can be found here.