31 juillet 2013

This Week in Elasticsearch - July 31, 2013

Par Alexander Reelsen

Welcome to the fifth issue of This week in elasticsearch. In this format we try to inform you about the latest and greatest changes in elasticsearch. We try to cover what happened in the github repositories, as well as all the events happening about elasticsearch and give you a small peek in the future.

Elasticsearch core

  • Switching to JUnit and Randomized testing (commit, master and 0.90)
  • Git build information has been added, when a distribution is build (#3370, master and 0.90)
  • Mapping: The dynamic flag can now be set at runtime (#3384, master and 0.90)
  • The master node creates IndexService instances in a more efficient way when executing a batch of master node operations. (Commits 1, 2, master and 0.90)
  • Rescoring query now supports score_mode (#3258, master and 0.90)
  • The new percolator engine now has support for percolating indexed documents (#3380, master only)
  • Code cleanup, removal of unused code (1, 2, 3, 4, 5, 6, 7, 8, done in master and 0.90)
  • Closing a not yet completely allocated index now throws an exception to make sure that a close operation was successful and leaves the index in a usable state (#3313, master and 0.90)
  • Get and Mget APIs got support for versions (#3404, master only)
  • Open/Close index API now can optionally wait for an acknowledgement that all cluster states have been updated (#3400, master and 0.90)
  • Using BigDoubleArrayList instead of arrays in geo point field data (commit, master and 0.90)
  • Transport layer: Better handling of failures to catch Throwable as well (1, 2, master and 0.90)
  • Version can now be injected, which makes it easier to test. In addition the smallest available version will be used as a common denominator between two nodes (commit, master and 0.90)
  • Do not use guava ordering to prevent copying data twice in SearchPhaseController (commit, master and 0.90)
  • Dynamic update of index.gc_deletes works (#3396, 0.90, was already fixed in master)
  • Getting back only parts from the source using the fields parameter may result in getting back a different document structure. Therefore a new mechanism has been introduced (#3301, master only)
  • Alias handling in the cluster state has been improved (less objects being created). The adding of an alias to a cluster state with 100k aliases improved  from 800ms to 230ms. (#3410, master and 0.90)
  • If you have auto_expand_replicas configured, the amount of cluster state changes has been significantly reduced (#3399, master and 0.90)
  • Lucene 4.4 allows to reuse string readers internally in the tokenstream, elasticsearch has been adapted (#3409, master and 0.90)
  • A index delete call with an acknowledgement set now waits for the actual deletion of shards on disk (#3413, master and 0.90)

Elasticsearch ecosystem

We try to give you some more information about what is happening in the ecosystem we are maintaining around elasticsearch, be it plugin or driver releases or news about kibana.

Elasticsearch community

Got an interesting open source project, plugin, driver or anything else for elasticsearch? Here is your time to shine! Just drop us a note and we will list it here (and on the .org website, of course!).

Meetups

Also, if you are interested in a core elasticsearch training, the next locations are San Francisco (limited offer: get three seats for the price of two!) and Boston at the beginning of August. For more locations, check the training page

If you are interested in all this, we are hiring. We are interested in your skills, not in your location. Just drop us a note.