24 11월 2015 출시

Elasticsearch 2.1.0 and 2.0.1 released

By Clinton Gormley

Today we are pleased to announce the release of Elasticsearch 2.1.0 based on Lucene 5.3.1 and a bug fix release of Elasticsearch 2.0.1.

Latest stable release:

Bug fixes in 2.0:

If you can't wait to get your hands on the new, shiny goodness…we understand. Get on with downloading the latest and greatest, or spin up a cluster on Found — the only hosted Elasticsearch service built and supported by the people building the software.

Elasticsearch 2.1.0 adds some great new features and has a number of important enhancements:

Geo-centroid aggregation

Kibana makes use of the geohash_grid aggregation to overlay aggregation results in a grid on a map, based on geolocation. Up until now, the results in each cell in the grid could only be positioned in the center of each cell, resulting in a row-and-column style display. Occasionally, the centre of a cell would fall squarely into the Atlantic Ocean instead of the neighbouring island of Manhattan.

The new geo_centroid metric aggregation calculates the weighted center — the point closest to the majority of the geo-points in the cell. This will result in a much more natural, realistic layout once it is supported by Kibana.

New pipeline aggregations

This release adds three new pipeline aggregations, used to post-process the results of other aggregations: the stats_bucket and extended_stats_bucket aggs calculate a number of statistics such as the min, max, avg, sum, sum_of_squares, variance, and std_deviation, while the percentiles_bucket agg calculates the specified percentiles.

Allocation and recovery

From Elasticsearch 1.6.0 onwards, we have made significant improvements to the time it takes to restart a node by adding synced flushing, delayed shard allocation, and the ability to cancel ongoing shard relocations if a node rejoins the cluster with a good copy of the shard.

These changes revealed some flaws in the shard rebalancing process which have been fixed in this release, such as rebalancing shards to a newly joined node before knowing what shards the node already holds (#14591), making rebalancing decisions before the cluster reaches green status (#14678), or ignoring constraints such as the cluster_concurrent_rebalance setting when changing shard awareness settings (#14259).

Incremental enhancements

We have made a number of enhancements which, though small, will improve your experience with Elasticsearch:

  • We have a placed a soft limit on the maximum number of results (default 10,000) that can be requested in any single search request. (#13188) This will protect the sysadmin from users who request the top billion results… just in case.
  • An update request will, by default, only write a new version of the document if the _source has actually changed. (#11306)
  • The count and scan search types have been replaced by searching with size set to 0 (#14165) and sorting by _doc (#12983) respectively.

Conclusion

Please download Elasticsearch 2.1.0, try it out, and let us know what you think on Twitter (@elastic) or in our forum. You can report any problems on the GitHub issues page.