The versionMap holds all versions (keyed by _uid) for recently indexed
documents. Previously we only cleared it during flush, which can be
infrequent if the translog flush thresholds are high, and can cause
excessive heap usage especially for small documents.
Now we clear it during refresh which is usually more frequent
(e.g. once per second by default).
Closes#6379
The GeoBounds Aggregation is a new single bucket aggregation which outputs the coordinates of a bounding box containing all the points from all the documents passed to the aggregation as well as the doc count. Geobound Aggregation also use a wrap_logitude parameter which specifies whether the resulting bounding box is permitted to overlap the international date line. This option defaults to true.
This aggregation introduces the idea of MetricsAggregation which do not return double values and cannot be used for sorting. The existing MetricsAggregation has been renamed to NumericMetricsAggregation and is a subclass of MetricsAggregation. MetricsAggregations do not store doc counts and do not support child aggregations.
Closes#5634
This commit reverts the commit for issue #5900 introduced
in `1.2.0`. The unlimited translog size can cause memory pressure
on ES instances with low memory and high indexing load.
Closes#6377
Routing has been inadvertly changed in #5562 resulting in documents going to
different shards in 1.2. This is a terrible bug because an indexing request
would not necessarily go to the same shard anymore, potentially leading to
duplicates.
Close#6391
Previously, More Like This would create a new mlt query for each value of a
multi-value field. This could result in all the values of the field to be
selected, which defeats the purpose of More Like This. Instead, the correct
behavior is to generate only one mlt query for all the values of the field.
This commit provides the correct behavior for More Like This DSL. The fix for
More Like This API will be coming in another commit.
Closes#6310
The BigArrays limit is currently shared by the translog, netty, http and some
queries/aggregations. If any of these consumers starts taking a lot of memory,
then other ones might fail to allocate memory, which could have bad
consequences, eg. if ping requests can't be sent. The plan is to come up with
a better solution in 1.3.
Close#6332
If a polygon is constructed which overlaps the date line but has a hole which lies entirely one to one side of the date line, JTS error saying that the hole is not within the bounds of the polygon because the code which splits the polygon either side of the date line does not add the hole to the correct component of the final set of polygons. The fix ensures this selection happens correctly.
Closes#6179
We perform some management operations that require the cluster to be
consistent with respect to the number of nodes in the cluster state
/ visible to the master in order to rely on the ack mechanism. This
only applies to the test infrastructure when nodes are not explicitly
started / stopped as well as while tearing down the cluster and wiping
indices after the tests.
Reusing Lucene's TermsEnum for _uid/version lookups gives a small
indexing (updates) speedup and brings us a closer to not having
to spend RAM on bloom filters.
Closes#6212
Randomize rewrite methods instead of trying them all when highlighting multi term queries with postings highlighter
Rely on search type randomization and remove all the explicit setSearchType calls as they are not needed anymore
Remove explicit `.from`, `.size` and `.explain`, not needed and might slow tests down (especially explain)
Added support for min_children and max_children parameters to
the has_child query and filter. A parent document will only
be considered if a match if the number of matching children
fall between the min/max bounds.
Closes#6019
Using ping.timeout, which defaults to 3s, to use as a timeout value on the join request a node makes to the master once its discovered can be too small, specifically when there is a large cluster state involved (and by definition, all the buffers and such on the nio layer will be "cold"). Introduce a dedicated join.timeout setting, that by default is 10x the ping.timeout (so 30s by default).
closes#6342
The top_hits aggregation returned an empty InternalTopHits instance with no fields set when there were no result, causing reduce and serialization errors down the road. This is fixed by setting all required fields when a there are no results.
Closes#6346
This default type has been inherited from its ancestor, the (non-paged) recycler whose memory
usage was unbounded and required soft references to make sure it could release memory eventually.
On the contrary, the page cache recycler memory usage is bounded so we could remove soft
references in order to remove load on the garbage collector.
Note: the cache type is already randomized in integration tests.
Close#6320
Mustache extracts the key/value pairs for parameter substitution from
objects and maps but it's decided on the first execution. We need to
make sure if the params are null we pass an empty map to ensure we
bind the map based extractor
Closes#6318
At the moment plain highligher only uses an analyzer defined for on the type
level. However, during the indexing stage it is possible to define analyzer on
per document level, for example mapping '_analyzer' to another field, containing
required name. This commit attempts to make sure that highlighting works
correctly in this scenario.
Closes#5497
