Log message:
Remove the stability entity, it has no meaning outside of an official context.

Log message:
Change the service_bundle name to "export" to reduce diffs between the
original manifest.xml file and the output from "svccfg export".

Log message:
Update databases/elasticsearch to 2.3.3.

== 2.3.3 Release Notes

=== New features

Geo::
- Add CONTAINS relation to geo_shape query

=== Enhancements

Core::
- Kill thread local leak

REST::
- CORS handling triggered whether User-Agent is a browser or not

=== Bug fixes

Core::
- Don't try to compute completion stats on a reader after we
  already closed it

Highlighting::
- Exclude all but string fields from highlighting if wildcards are
  used...

Query DSL::
- Fix NullPointerException in SimpleQueryParser when analyzing
  text produces a null query

REST::
- CORS should permit same origin requests
- Allow CORS requests to work with HTTP compression enabled

=== Upgrades

Internal::
- Update Jackson 2.6.2 -> 2.6.6 (latest and final 2.6 patch)

Log message:
Update databases/elasticsearch to 2.3.2.

== 2.3.2 Release Notes

=== Deprecations

REST::
- Deprecate camelCase usages

Search::
- Deprecate sort option reverse.

Settings::
- Deprecate camelCase settings magic

=== Enhancements

Analysis::
- Analyze API : Fix/deprecated filters in analyze in 2x

=== Bug fixes

Aggregations::
- Aggregations: Adds serialisation of sigma to
  extended_stats_bucket pipeline aggregation
- ExtendedStatsAggregator should also pass sigma to emtpy aggs.

Allocation::
- Rebalancing policy shouldn't prevent hard allocation decisions
- When considering the size of shadow replica shards, set size to 0

Logging::
- Add missing index name to search slow log.

Mapping::
- Fail if an object is added after a field with the same name.

More Like This::
- Alias items are not ignored anymore

REST::
- Fixes reading of CORS pre-flight headers and methods
- index is a required url part for update by query

Recovery::
- Backports Notify GatewayRecoveryListener on failure

Snapshot/Restore::
- On restore, selecting concrete indices can select wrong index

== 2.3.1 Release Notes

=== Bug fixes

Aggregations::
- Prevents exception being raised when ordering by an aggregation
  which wasn't collected

Cluster::
- Cluster stats: fix memory available that is always set to 0

== 2.3.0 Release Notes

=== Breaking changes

Allocation::
- Speed up shard balancer by reusing shard model while moving
  shards that can no longer be allocated to a node

Mapping::
- Add per-index setting to limit number of nested fields

=== Deprecations

Geo::
- Deprecate lat_lon and precision_step
- Deprecate geo_point mapping params

Plugin Discovery Multicast::
- Deprecate Discovery Multicast Plugin

Query DSL::
- Add deprecation logging for deprecated queries.

=== New features

Reindex API::
- Backport reindex to 2.x

Task Manager::
- Backport of task management api to 2x

=== Enhancements

Allocation::
- Skip capturing least/most FS info for an FS with no total
- Speed improvements for BalancedShardsAllocator

Cache::
- Make TermsQuery considered costly.

Cluster::
- Shard state action request logging
- Safe cluster state task notifications

Core::
- Remove log4j exception hiding
- Avoid cloning MessageDigest instances
- Add a hard check to ensure we are running with the expected
  lucene version

Exceptions::
- Added file name to exceptions when failing to read index state
- Add Exception class name to message in
  `NotSerializableExceptionWrapper`

Geo::
- Upgrade GeoPointField to use Lucene 5.5 PrefixEncoding

Internal::
- Cleanup search sub transport actions and collapse
  o.e.action.search.type package into o.e.action.search

Java API::
- TransportClient: Add exception when using plugin.types, to help
  migration to addPlugin

Logging::
- Log suppressed stack traces under DEBUG

Mapping::
- Add deprecation logging for mapping transform
- Expose the reason why a mapping merge is issued.
- Add sub-fields support to `bool` fields.
- Improve cross-type dynamic mapping updates.
- Make mapping updates more robust.

Network::
- Add additional fallback to http.publish_port and restrict
  fallback to transport.publish_port

Packaging::
- Windows service: Use JAVA_HOME environment variable in registry
- Default standard output to the journal in systemd
- Use egrep instead of grep -E for Solaris

Plugin Cloud Azure::
- Add timeout settings (default to 5 minutes)

Plugin Discovery EC2::
- Add ap-northeast-2 (seoul) endpoints for EC2 discovery and S3
  snapshots

Plugin Mapper Attachment::
- Fix attachments plugins with docx

REST::
- More robust handling of CORS HTTP Access Control

Reindex API::
- Implement helpful interfaces in reindex requests
- Reindex should timeout if sub-requests timeout
- Teach reindex to retry on rejection

Scripting::
- Logs old script params use to the Deprecation Logger
- Remove suppressAccessChecks permission for Groovy script plugin
- Class permission for Groovy references
- Scripting: Allow to get size of array in mustache
- Added plumbing for compile time script parameters
- Enhancements to the mustache script engine

Search::
- Add deprecation logging for ignore_unmapped parameter
- Add deprecation logging when users use the SCAN search type.

Settings::
- Log warning if max file descriptors too low

=== Bug fixes

Aggregations::
- Setting 'other' bucket on empty aggregation
- Build empty extended stats aggregation if no docs collected for
  bucket
- Set meta data for pipeline aggregations
- Filter(s) aggregation should create weights only once.
- Make `missing` on terms aggs work with all execution modes.
- Fix NPE in Derivative Pipeline when current bucket value is null

Aliases::
- Fix _aliases filter and null parameters
- Check lenient_expand_open after aliases have been resolved

Allocation::
- IndicesStore checks for `allocated elsewhere` for every shard
  not alocated on the local node
- Prevent peer recovery from node with older version

Analysis::
- Analysis : Allow string explain param in JSON
- Analysis : Fix no response from Analyze API without specified
  index

Bulk::
- Bulk api: fail deletes when routing is required but not
  specified

CAT API::
- Fix column aliases in _cat/indices, _cat/nodes and _cat/shards
  APIs

CRUD::
- fixed propagation of autoGeneratedId on document creation
- Throw exception if content type could not be determined in
  Update API

Cache::
- Handle closed readers in ShardCoreKeyMap

Cluster::
- Index deletes not applied when cluster UUID has changed
- Only fail the relocation target when a replication request on it
  fails

Core::
- Migrate template settings (add missing byte & time units)

Fielddata::
- Make disabled fielddata loading fail earlier.

Geo::
- Remove .geohash suffix from GeoDistanceQuery and
  GeoDistanceRangeQuery

Highlighting::
- Don't override originalQuery with request filters

Inner Hits::
- Query and top level inner hit definitions shouldn't overwrite
  each other

Internal::
- Log uncaught exceptions from scheduled once tasks
- FunctionScoreQuery should implement two-phase iteration.

Java API::
- Fix boundary chars in java api

Logging::
- Add missing index name to indexing slow log
- ParseFieldMatcher should log when using deprecated settings.

Mapping::
- Put mapping operations must update metadata of all types.
- Fix serialization of `search_analyzer`.
- Reuse metadata mappers for dynamic updates.
- Fix MapperService#searchFilter(...)
- MapperService: check index.mapper.dynamic during index creation

Network::
- Only accept transport requests after node is fully initialized

Packaging::
- Fix waiting for pidfile
- Enable es_include at init

Parent/Child::
- Check that parent_type in Has Parent Query has child types
- Has child query forces default similarity

Percolator::
- Don't replace found fields if map unmapped fields as string is
  enabled
- mpercolate api should serialise start time

Plugin Cloud Azure::
- Fix calling ensureOpen() on the wrong directory

Plugin Discovery GCE::
- Add setFactory permission to GceDiscoveryPlugin

Query DSL::
- Fix FunctionScore equals/hashCode to include minScore and
  friends

REST::
- Remove detect_noop from REST spec
- Make text parsing less lenient.

Recovery::
- Try to renew sync ID if `flush=true` on forceMerge
- sync translog to disk after recovery from primary

Reindex API::
- Properly register reindex status
- Be much much much careful about context
- Make search failure cause rest failure

Scripting::
- Check that _value is used in aggregations script before setting
  value to specialValue
- Add permission to access sun.reflect.MethodAccessorImpl from
  Groovy scripts
- Fixes json generation for scriptsort w/ deprecated params
- Security permissions for Groovy closures

Search::
- Do not apply minimum_should_match on auto generated boolean
  query if the coordination factor is disabled.
- Do not apply minimum-should-match on a boolean query if the
  coords are disabled
- Fix blended terms take 2

Settings::
- TransportClient should use updated setting for initialization of
  modules and service

Snapshot/Restore::
- Prevent closing index during snapshot restore
- Add node version check to shard allocation during restore

Stats::
- Fix recovery translog stats totals when recovering from store
- Fix calculation of age of pending tasks

Task Manager::
- Fix TaskId#isSet to return true when id is set and not other way
  around
- Don't wait for completion of list tasks tasks when
  wait_for_completion flag is set

Translog::
- Call ensureOpen on Translog#newView() to prevent
  IllegalStateException
- Make sure IndexShard is active during recovery so it gets its
  fair share of the indexing buffer
- Avoid circular reference in exception
- Initialize translog before scheduling the sync to disk
- Catch tragic even inside the checkpoint method rather than on
  the caller side
- Never delete translog-N.tlog file when creation fails
- Close recovered translog readers if createWriter fails

Tribe Node::
- Passthrough environment and network settings to tribe client
  nodes
- Tribe node: pass path.conf to inner tribe clients

=== Regressions

Analysis::
- Add PathHierarchy type back to path_hierarchy tokenizer for
  backward compatibility with 1.x

Plugin Cloud Azure::
- Fix Unknown [repository] type [azure] error with 2.2.0

REST::
- Don't return all indices immediately if count of expressions >1
  and first expression is *

=== Upgrades

Core::
- Upgrade to Lucene 5.5.0 official release
- Upgrade to lucene 5.5.0-snapshot-850c6c2
- Upgrade to lucene 5.5.0-snapshot-4de5f1d
- Update lucene to r1725675
- Upgrade to lucene-5.5.0-snapshot-1721183.

Plugin Cloud Azure::
- Upgrade Azure Storage client to 4.0.0

Plugin Discovery Azure::
- Upgrade azure SDK to 0.9.3

Scripting::
- Upgrade groovy dependency in lang-groovy module to version 2.4.6

== 2.2.1 Release Notes

=== Enhancements

Cluster::
- Shard state action request logging

Core::
- Avoid cloning MessageDigest instances

Exceptions::
- Add Exception class name to message in
  `NotSerializableExceptionWrapper`

Network::
- Add additional fallback to http.publish_port and restrict
  fallback to transport.publish_port

Packaging::
- Windows service: Use JAVA_HOME environment variable in registry

Scripting::
- Class permission for Groovy references

Settings::
- Log warning if max file descriptors too low

=== Bug fixes

Aggregations::
- Set meta data for pipeline aggregations

Aliases::
- Fix _aliases filter and null parameters

Bulk::
- Bulk api: fail deletes when routing is required but not
  specified

Inner Hits::
- Query and top level inner hit definitions shouldn't overwrite
  each other

Logging::
- Add missing index name to indexing slow log

Parent/Child::
- Check that parent_type in Has Parent Query has child types
- Has child query forces default similarity

Percolator::
- Don't replace found fields if map unmapped fields as string is
  enabled

Plugin Cloud Azure::
- Fix calling ensureOpen() on the wrong directory

Plugin Discovery GCE::
- Add setFactory permission to GceDiscoveryPlugin

Query DSL::
- Use MappedFieldType.termQuery to generate simple_query_string
  queries

REST::
- Remove detect_noop from REST spec

Scripting::
- Add permission to access sun.reflect.MethodAccessorImpl from
  Groovy scripts

Snapshot/Restore::
- Add node version check to shard allocation during restore

Stats::
- Fix recovery translog stats totals when recovering from store

Tribe Node::
- Passthrough environment and network settings to tribe client
  nodes

=== Regressions

Plugin Cloud Azure::
- Fix Unknown [repository] type [azure] error with 2.2.0

=== Upgrades

Plugin Repository Azure::
- Upgrade Azure Storage client to 4.0.0

== 2.2.0 Release Notes

=== Breaking changes

Index APIs::
- Change Field stats API response to include both number and
  string based min and max values

Scripting::
- Filter classes loaded by scripts
- Lock down javascript and python script engines better

=== Deprecations

Java API::
- Deprecate the count api in favour of search with size 0
- Deprecates defaultRescoreWindowSize

Plugin Discovery Multicast::
- Deprecate the discovery-multicast plugin

Query DSL::
- Deprecate the `missing` query
- Deprecate NotQueryBuilder

Search::
- Deprecate _search/exists in favour of regular _search with size
  0 and terminate_after 1

=== New features

Geo::
- Add CONTAINS relation to geo_shape query
- Add support for Lucene 5.4 GeoPoint queries
- Add GeoPointV2 Field Mapping

Network::
- Allow binding to multiple addresses.

Plugin Analysis Phonetic::
- Add support for `daitch_mokotoff`

Plugin Repository S3::
- Add support for S3 storage class

Plugins::
- Decentralize plugin security

Search::
- Add query profiler

=== Enhancements

Aliases::
- Add support to _aliases endpoint to specify multiple indices and
  aliases in one action

Allocation::
- Simplify delayed shard allocation
- Add cluster-wide setting for total shard limit
- Early terminate high disk watermark checks on single data node
  cluster

Analysis::
- Add detail response support for _analyze API

CAT API::
- Add sync_id to cat shards API
- Add duration field to /_cat/snapshots
- Add cat API for repositories and snapshots
- Adds disk used by indices to _cat/allocation

Cluster::
- Safe cluster state task notifications
- Reroute once per batch of shard failures
- Add callback for publication of new cluster state
- Split cluster state update tasks into roles
- Use general cluster state batching mechanism for shard started
- Use general cluster state batching mechanism for shard failures
- Set an newly created IndexShard's ShardRouting before exposing
  it to operations
- Uniform exceptions for TransportMasterNodeAction

Core::
- If we can't get a MAC address for the node, use a dummy one
- Simplify shard inactive logging
- Simplify IndexingMemoryController#checkIdle
- IndexingMemoryController should not track shard index states
- Make PerThreadIDAndVersionLookup per-segment
- Verify Checksum once it has been fully written to fail as soon
  as possible

Discovery::
- Don't allow nodes with missing custom meta data to join cluster

Exceptions::
- Deduplicate cause if already contained in shard failures
- Give a better exception when running from freebsd jail without
  enforce_statfs=1
- Make root_cause of field conflicts more obvious

Fielddata::
- Update GeoPoint FieldData for GeoPointV2

Geo::
- Geo: Fix toString() in GeoDistanceRangeQuery and GeoPolygonQuery
- Enable GeoPointV2 with backward compatibility testing
- Refactor Geo utilities to Lucene 5.4

Index APIs::
- Add option to disable closing indices

Index Templates::
- Disallow index template pattern to be the same as an alias name

Internal::
- Simplify the Text API.
- Simpler using compressed oops flag representation
- Info on compressed ordinary object pointers
- Explicitly log cluster state update failures
- Use transport service to handle RetryOnReplicaException to
  execute replica action on the current node
- Make IndexShard operation be more explicit about whether they
  are expected to run on a primary or replica
- Avoid trace logging allocations in
  TransportBroadcastByNodeAction
- Only trace log shard not available exceptions
- Transport options should be immutable
- Fix dangling comma in ClusterBlock#toString
- Improve some logging around master election and cluster state
- Add System#exit(), Runtime#exit() and Runtime#halt() to
  forbidden APIs
- Simplify XContent detection.
- Add threadgroup isolation.
- Cleanup plugin security
- Add workaround for JDK-8014008
- Refactor retry logic for TransportMasterNodeAction
- Remove MetaDataSerivce and it's semaphores
- Cleanup IndexMetaData
- TransportNodesAction shouldn't hold on to cluster state
- Add SpecialPermission to guard exceptions to security policy.
- Clean up scripting permissions.
- Factor groovy out of core into lang-groovy
- Factor expressions scripts out to lang-expression plugin

Java API::
- TransportClient: Add exception when using plugin.types, to help
  migration to addPlugin
- Align handling of interrupts in BulkProcessor
- BulkProcessor backs off exponentially by default
- Add java API for synced flush
- Allow to get and set ttl as a time value/string
- Reject refresh usage in bulk items when using and fix NPE when
  no source
- BulkProcessor retries after request handling has been rejected
  due to a full thread pool

Logging::
- Add circuit breaker name to logging package
- Log cluster health status changes
- Map log-level 'trace' to JDK-Level 'FINEST'

Mapping::
- Make mapping serialization more robust.
- Make mappings immutable.
- Make MappedFieldType.checkTypeName part of
  MappedFieldType.checkCompatibility.
- throw exception if a copy_to is within a multi field
- Register field mappers at the node level.

Network::
- only allow code to bind to the user's configured port
  numbers/ranges
- Port of publishAddress should match port of corresponding
  boundAddress

Packaging::
- Default standard output to the journal in systemd
- Use egrep instead of grep -E for Solaris
- punch thru symlinks when loading plugins/modules
- set ActiveProcessLimit=1 on windows
- set RLIMIT_NPROC = 0 on bsd/os X systems.
- Drop ability to execute on Solaris
- Packaging: change permissions/ownership of config dir
- Release: Fix package repo path to only consist of major version
- Nuke ES_CLASSPATH appending, JarHell fail on empty classpath
  elements
- Get lang-javascript, lang-python, securemock ready for script
  refactoring
- Remove some bogus permissions only needed for tests.
- Remove java.lang.reflect.ReflectPermission
  "suppressAccessChecks"

Plugin Cloud AWS::
- Add support for proxy authentication for s3 and ec2

Plugin Cloud Azure::
- Add timeout settings (default to 5 minutes)
- Add support for secondary azure storage account

Plugin Cloud GCE::
- Adding backoff from retries on GCE errors
- Add `_gce_` network host setting

Plugin Discovery EC2::
- Add ap-northeast-2 (seoul) endpoints for EC2 discovery and S3
  snapshots
- Adding US-Gov-West
- Improved building of disco nodes

Plugin Mapper Attachment::
- Backport mapper-attachments plugin to 2.x

Plugin Repository S3::
- Add aws canned acl

Plugins::
- Expose http.type setting, and collapse al(most all) modules
  relating to transport/http
- Ban RuntimePermission("getClassLoader")
- Add nicer error message when a plugin descriptor is missing
- Don't be lenient in PluginService#processModule(Module)
- Check "plugin already installed" before jar hell check.
- Plugin script to set proper plugin bin dir attributes
- Plugin script to set proper plugin config dir attributes

Query DSL::
- Allow CIDR notation in query string query

REST::
- Make XContentGenerator.writeRaw* safer.
- Filter path refactoring

Recovery::
- Handle cancel exceptions on recovery target if the cancel comes
  from the source
- Decouple routing and primary operation logic in
  TransportReplicationAction

Scripting::
- Add property permissions so groovy scripts can serialize json
- Remove ScriptEngineService.unwrap.
- Remove ScriptEngineService.execute.

Search::
- Support * wildcard to retrieve stored fields in the 'fields'
  option
- Caching Weight wrappers should propagate the BulkScorer.
- fix numerical issue in function score query

Snapshot/Restore::
- Support wildcards for getting repositories and snapshots
- Add ignore_unavailable parameter to skip unavailable snapshot

Stats::
- Add system CPU percent to OS stats
- Pull Fields instance once from LeafReader in completion stats
- Add os.allocated_processors stats

Top Hits::
- Put method addField on TopHitsBuilder

Translog::
- Check for tragic event on all kinds of exceptions not only ACE
  and IOException

Tribe Node::
- Tribe nodes should apply cluster state updates in batches

=== Bug fixes

Aggregations::
- Filter(s) aggregation should create weights only once.
- Make `missing` on terms aggs work with all execution modes.
- Run pipeline aggregations for empty buckets added in the Range
  Aggregation
- [Children agg] fix bug that prevented all child docs from being
  evaluated
- Correct typo in class name of StatsAggregator
- Fix significant terms reduce for long terms
- Pass extended bounds into HistogramAggregator when creating an
  unmapped aggregator
- Added correct generic type parameter on ScriptedMetricBuilder

Aliases::
- Check lenient_expand_open after aliases have been resolved

Allocation::
- Prevent peer recovery from node with older version
- Fix calculation of next delay for delayed shard allocation
- Take ignored unallocated shards into account when making
  allocation decision
- Only allow rebalance operations to run if all shard store data
  is available
- Delayed allocation can miss a reroute
- Check rebalancing constraints when shards are moved from a node
  they can no longer remain on

Bulk::
- Do not release unacquired semaphore

CAT API::
- Properly set indices and indicesOptions on subrequest made by
  /_cat/indices

CRUD::
- Throw exception if content type could not be determined in
  Update API

Cluster::
- Only fail the relocation target when a replication request on it
  fails
- Handle shards assigned to nodes that are not in the cluster
  state
- Bulk cluster state updates on index deletion

Core::
- BitSetFilterCache duplicates its content.
- Limit the max size of bulk and index thread pools to bounded
  number of processors
- AllTermQuery's scorer should skip segments that never saw the
  requested term
- Include root-cause exception when we fail to change shard's
  index buffer
- Restore thread interrupt flag after an InterruptedException
- Use fresh index settings instead of relying on @IndexSettings
- Record all bytes of the checksum in VerifyingIndexOutput
- When shard becomes active again, immediately increase its
  indexing buffer
- Close TokenStream in finally clause

Engine::
- Never wrap searcher for internal engine operations

Exceptions::
- Fix ensureNodesAreAvailable's error message

Fielddata::
- Don't cache top level field data for fields that don't exist

Geo::
- Geo: Allow numeric parameters enclosed in quotes for
  'geohash_grid' aggregation

Highlighting::
- Fix spans extraction to not also include individual terms.

Index APIs::
- Field stats: Index constraints should remove indices in the
  response if the field to evaluate is empty
- Field stats: Fix NPE for index constraint on empty index
- Field stats: Added `format` option for index constraints
- Restore previous optimize transport action name for bw comp
- Forbid index name `.` and `..`

Internal::
- Log uncaught exceptions from scheduled once tasks
- Make sure the remaining delay of unassigned shard is updated
  with every reroute
- Throw a meaningful error when loading metadata and an alias and
  index have the same name
- fixup issues with 32-bit jvm
- Failure to update the cluster state with the recovered state
  should make sure it will be recovered later
- Properly bind ClassSet extensions as singletons
- Streamline top level reader close listeners and forbid general
  usage
- Gateway: a race condition can prevent the initial cluster state
  from being recovered
- Verify actually written checksum in VerifyingIndexOutput
- Remove all setAccessible in tests and forbid
- Remove easy uses of setAccessible in tests.
- Ban setAccessible from core code, restore monitoring stats under
  java 9

Logging::
- Settings in log config file should not overwrite custom
  parameters
- Plugin cli tool should not create empty log files

Mapping::
- Fix MapperService#searchFilter(...)
- Fix initial sizing of BytesStreamOutput.
- MetaDataMappingService should call MapperService.merge with the
  original mapping update.
- Fix copy_to when the target is a dynamic object field.
- Only text fields should accept analyzer and term vector
  settings.
- Mapper parsers should not check for a `tokenized` property.
- Validate that fields are defined only once.
- Check mapping compatibility up-front.
- Don't treat _default_ as a regular type.
- Don't ignore mapping merge failures.
- Treat mappings at an index-level feature.
- Multi field names may not contain dots
- Mapping: Allows upgrade of indexes with only search_analyzer
  specified

Packaging::
- Fix Windows service installation failure
- fix command line options for windows bat file
- Handle system policy correctly
- Startup script exit status should catch daemonized startup
  failures

Plugin Cloud AWS::
- Discovery-ec2 plugin should check `discovery.type`

Plugin Delete By Query::
- Fix Delete-by-Query with Shield
- Delete by query to not wrap the inner query into an additional
  query element

Plugin Mapper Attachment::
- Fix toXContent() for mapper attachments field

Plugin Repository S3::
- Hack around aws security hole of accessing sun.security.ssl, s3
  repository works on java 9 again

Plugins::
- Fix plugin list command error message
- Parse Java system properties in plugin.bat
- PluginManager: Dont leave leftover files on unsuccessful
  installs
- Fix HTML response during redirection

Query DSL::
- Fix FunctionScore equals/hashCode to include minScore and
  friends
- Min should match greater than the number of optional clauses
  should return no result
- Return a better exception message when `regexp` query is used on
  a numeric field

REST::
- Throw exception when trying to write map with null keys
- XContentFactory.xContentType: allow for possible UTF-8 BOM for
  JSON XContentType
- Restore support for escaped '/' as part of document id
- Add missing REST spec for `detect_noop`
- Expose nodes operation timeout in REST API
- Ensure XContent is consistent across platforms

Recovery::
- sync translog to disk after recovery from primary
- CancellableThreads should also treat ThreadInterruptedException
  as InterruptedException

Scripting::
- Security permissions for Groovy closures

Search::
- Fix NPE when a segment with an empty cache gets closed.
- Fix the quotes in the explain message for a script score
  function without parameters

Settings::
- TransportClient should use updated setting for initialization of
  modules and service

Shadow Replicas::
- Don't pull translog from shadow engine

Snapshot/Restore::
- Snapshot restore and index creates should keep index settings
  and cluster blocks in sync
- Snapshot restore operations throttle more than specified
- Fix blob size in writeBlob() method

Stats::
- Fix calculation of age of pending tasks
- serialize os name, arch and version too
- Add extra validation into `cluster/stats`
- Omit current* stats for OldShardStats

Translog::
- Make sure IndexShard is active during recovery so it gets its
  fair share of the indexing buffer
- Avoid circular reference in exception
- Initialize translog before scheduling the sync to disk
- Translog base flushes can be disabled after replication
  relocation or slow recovery
- Catch tragic even inside the checkpoint method rather than on
  the caller side
- Never delete translog-N.tlog file when creation fails
- Close recovered translog readers if createWriter fails
- Fail and close translog hard if writing to disk fails
- Prevent writing to closed channel if translog is already closed
- Don't delete temp recovered checkpoint file if it was renamed
- Translog recovery can repeatedly fail if we run out of disk

Tribe Node::
- Tribe node: pass path.conf to inner tribe clients
- Fix tribe node to load config file for internal client nodes

=== Regressions

Analysis::
- Add PathHierarchy type back to path_hierarchy tokenizer for
  backward compatibility with 1.x

Internal::
- Deduplicate concrete indices after indices resolution

Plugin Cloud Azure::
- Filter cloud azure credentials

Query DSL::
- RangeQueryParser should accept `_name` in inner field

=== Upgrades

Core::
- Upgrade to lucene-5.4.1.
- Upgrade to lucene-5.4.0.
- Upgrade Lucene to 5.4.0-snapshot-1715952
- Upgrade Lucene to 5.4.0-snapshot-1714615
- Upgrade to lucene-5.4.0-snapshot-1712973.
- update to lucene-5.4.x-snapshot-1711508
- Upgrade to lucene-5.4-snapshot-1710880.
- Upgrade to lucene-5.4-snapshot-1708254.
- upgrade lucene to r1702265
- Upgrade master to lucene 5.4-snapshot r1701068

Geo::
- Update to spatial4j 0.5 for correct Multi-Geometry

Network::
- Upgrade Netty to 3.10.5.final

Plugin Discovery Azure::
- Update Azure Service Management API to 0.9.0

Plugin Discovery EC2::
- Upgrade to aws 1.10.33

Plugin Lang JS::
- upgrade rhino for plugins/lang-javascript

== 2.1.2 Release Notes

=== Enhancements

Internal::
- Only trace log shard not available exceptions

Plugin Cloud Azure::
- Add timeout settings (default to 5 minutes)

Translog::
- Check for tragic event on all kinds of exceptions not only ACE
  and IOException

=== Bug fixes

Aggregations::
- Make `missing` on terms aggs work with all execution modes.

Aliases::
- Check lenient_expand_open after aliases have been resolved

Allocation::
- Prevent peer recovery from node with older version

Cluster::
- Only fail the relocation target when a replication request on it
  fails

Core::
- BitSetFilterCache duplicates its content.
- Limit the max size of bulk and index thread pools to bounded
  number of processors
- AllTermQuery's scorer should skip segments that never saw the
  requested term

Highlighting::
- Fix spans extraction to not also include individual terms.

Internal::
- Log uncaught exceptions from scheduled once tasks
- Fix ConcurrentModificationException from nodes info and nodes
  stats

Mapping::
- Fix MapperService#searchFilter(...)
- Fix initial sizing of BytesStreamOutput.

Packaging::
- Fix Windows service installation failure
- fix command line options for windows bat file

Query DSL::
- Fix FunctionScore equals/hashCode to include minScore and
  friends

Recovery::
- sync translog to disk after recovery from primary

Stats::
- Fix calculation of age of pending tasks

Translog::
- Make sure IndexShard is active during recovery so it gets its
  fair share of the indexing buffer
- Avoid circular reference in exception
- Initialize translog before scheduling the sync to disk
- Translog base flushes can be disabled after replication
  relocation or slow recovery
- Catch tragic even inside the checkpoint method rather than on
  the caller side
- Never delete translog-N.tlog file when creation fails
- Close recovered translog readers if createWriter fails

Tribe Node::
- Tribe node: pass path.conf to inner tribe clients

=== Regressions

Analysis::
- Add PathHierarchy type back to path_hierarchy tokenizer for
  backward compatibility with 1.x

Log message:
Update databases/elastisearch to 2.1.1.

elasticsearch 2.1.1
===================

Enhancements

Aggregations
- [Children agg] fix bug that prevented all child docs from being evaluated
Core
- If we can't get a MAC address for the node, use a dummy one
- Simplify shard inactive logging
- Simplify IndexingMemoryController#checkIdle
- IndexingMemoryController should not track shard index states
Index Templates
- Disallow index template pattern to be the same as an alias name
Mapping
- throw exception if a copy_to is within a multi field
- Register field mappers at the node level.

Bug fixes

Index APIs
- Field stats: Index constraints should remove indices in the response
  if the field to evaluate is empty
Internal
- Throw a meaningful error when loading metadata and an alias and index
  have the same name
Mapping
- Multi field names may not contain dots
Search
- Fix NPE when a segment with an empty cache gets closed.
Translog
- Fail and close translog hard if writing to disk fails
- Prevent writing to closed channel if translog is already closed
- Don't delete temp recovered checkpoint file it was renamed
Tribe Node
- Fix tribe node to load config file for internal client nodes

Regressions

Query DSL
- RangeQueryParser should accept _name in inner field

elasticsearch 2.1.0
===================

Breaking changesedit

CRUD
- Default detect_noop to true
Fielddata
- Remove the experimental indices.fielddata.cache.expire
Index APIs
- Add Force Merge API, deprecate Optimize API
Internal
- Forbid changing thread pool types
Java API
- Deprecates defaultRescoreWindowSize
Nested Docs
- If sorting by nested field then the nested_path should always be
  specified
Search
- Limit the size of the result window to a dynamic property
Stats
- The queue_size value should be shown as an integer.

Deprecations

Java API
- Deprecate the count api in favour of search with size 0
Parent/Child
- Deprecate score_type option in favour of the score_mode option
Query DSL
- Deprecate NotQueryBuilder
Search
- Deprecate _search/exists in favour of regular _search with size 0 and
  terminate_after 1
- Deprecate the scan search type.

New features

Aggregations
- Add percentiles_bucket pipeline aggregation
- Add stats_bucket / extended_stats_bucket pipeline aggs
Analysis
- Lithuanian analysis
Geo
- Adds geo_centroid metric aggregator

Enhancements

Allocation
- Add cluster-wide setting for total shard limit
- Early terminate high disk watermark checks on single data node cluster
- Also use PriorityComparator in shard balancer
- Add support for filtering by publish IP address
CAT API
- Add duration field to /_cat/snapshots
- Add cat API for repositories and snapshots
- Adds disk used by indices to _cat/allocation
Core
- Verify Checksum once it has been fully written to fail as soon as
  possible
Exceptions
- Deduplicate cause if already contained in shard failures
- Give a better exception when running from freebsd jail without
  enforce_statfs=1
- Make root_cause of field conflicts more obvious
- Use a dedicated id to serialize EsExceptions instead of it's class name.
- Validate class before cast.
- Improve error message of ClassCastExceptions
Geo
- Refactor geo_point validate* and normalize* for 2.x
Index APIs
- Limit type name length
Index Templates
- Accumulate validation errors when validating index templates
Internal
- Fix dangling comma in ClusterBlock#toString
- Improve some logging around master election and cluster state
- Add workaround for JDK-8014008
- Cleanup IndexMetaData
- More helpful error message on parameter order
- Cleanup InternalClusterInfoService
- Remove and forbid use of com.google.common.base.Throwables
- Remove cyclic dependencies between IndexService and FieldData/BitSet
  caches
- Remove and forbid use of com.google.common.base.Objects
- Remove and forbid use of com.google.common.collect.ImmutableList
- Remove and forbid use of com.google.common.collect.Lists
- Remove unused code from query_string parser and settings
- Consolidate duplicate logic in RoutingTable all*ShardsGrouped
- Turn DestructiveOperations.java into a Guice module.
- Remove CachedDfSource
- Enable indy (invokedynamic) compile flag for Groovy scripts by
  default
Java API
- Prevents users from building a BulkProcessor with a null client
Logging
- Move logging for the amount of free disk to TRACE
Packaging
- Drop ability to execute on Solaris
- Nuke ES_CLASSPATH appending, JarHell fail on empty classpath elements
- improve seccomp syscall filtering
- Block process execution with seccomp on linux/amd64
- Remove JAVA_HOME detection from the debian init script
Plugin Cloud AWS
- Enable S3SignerType
- Remove cloud.account and cloud.key settings
Plugin Cloud GCE
- cloud-gce plugin should check discovery.type
Plugin Discovery EC2
- Adding US-Gov-West
- Improved building of disco nodes
Plugin Repository S3
- Add aws canned acl
Plugins
- Don't be lenient in PluginService#processModule(Module)
- Adds a validation for plugins script to check if java is set
- Plugins: Removed plugin.types
- Improve java version comparison and explicitly enforce a version format
- Output plugin info only in verbose mode
Query DSL
- Internal: simplify filtered query conversion to lucene query
- Remove unsupported rewrite from multi_match query builder
- Remove unsupported rewrite option from match query builder
- Make FunctionScore work on unmapped field with missing parameter
Scripting
- Add property permissions so groovy scripts can serialize json
Scroll
- Optimize sorted scroll when sorting by _doc.
Search
- fix numerical issue in function score query
- Optimize scrolls for constant-score queries.
- Optimize counts on simple queries.
Search Templates
- Adds template support to _msearch resource
Snapshot/Restore
- Simplify the BlobContainer blob writing interface
- Add readonly option for repositories
Stats
- Add os.allocated_processors stats
- Adds stats counter for failed indexing requests

Bug fixes

Aggregations
- Pass extended bounds into HistogramAggregator when creating an
  unmapped aggregator
- Added correct generic type parameter on ScriptedMetricBuilder
- Pipeline Aggregations at the root of the agg tree are now validated
- Estimate HyperLogLog bias via k-NN regression
Allocation
- Fix calculation of next delay for delayed shard allocation
- Take ignored unallocated shards into account when making allocation
  decision
- Only allow rebalance operations to run if all shard store data is
  available
- Delayed allocation can miss a reroute
- Check rebalancing constraints when shards are moved from a node they
  can no longer remain on
CAT API
- Properly set indices and indicesOptions on subrequest made by
  /_cat/indices
CRUD
- Index name expressions should not be broken up
Cluster
- Handle shards assigned to nodes that are not in the cluster state
Core
- Use fresh index settings instead of relying on @IndexSettings
- Fork Lucene PatternTokenizer to apply LUCENE-6814 (closes
- Record all bytes of the checksum in VerifyingIndexOutput
- When shard becomes active again, immediately increase its indexing buffer
- Close TokenStream in finally clause
- LoggingRunnable.run should catch and log all errors, not just Exception?
Exceptions
- Fix ensureNodesAreAvailable's error message
Fielddata
- Don't cache top level field data for fields that don't exist
Geo
- Geo: Allow numeric parameters enclosed in quotes for geohash_grid
  aggregation
- Resync Geopoint hashCode/equals method
- Fix GeoPointFieldMapper to index geohash at correct precision.
Index APIs
- Field stats: Fix NPE for index constraint on empty index
- Field stats: Added format option for index constraints
- Restore previous optimize transport action name for bw comp
- Forbid index name . and ..
Index Templates
- Validate settings specified in index templates at template creation time
Internal
- fix mvn verify on jigsaw with 2.1
- fixup issues with 32-bit jvm
- Failure to update the cluster state with the recovered state should
  make sure it will be recovered later
- Gateway: a race condition can prevent the initial cluster state from
  being recovered
- Verify actually written checksum in VerifyingIndexOutput
- An inactive shard is activated by triggered synced flush
Logging
- Don't log multi-megabyte guice exceptions.
- Moving system property setting to before it can be used
Mapping
- Make _type use doc values
- Mapping: Allows upgrade of indexes with only search_analyzer specified
Packaging
- Handle system policy correctly
- Startup script exit status should catch daemonized startup failures
- Don't let ubuntu try to install its crazy jayatana agent.
Parent/Child
- Remove unnecessary usage of extra index searchers
- Plugin Delete By Query
- Fix Delete-by-Query with Shield
- Delete by query to not wrap the inner query into an additional query
  element
Plugins
- Fix plugin list command error message
- Fix HTML response during redirection
REST
- XContentFactory.xContentType: allow for possible UTF-8 BOM for JSON
  XContentType
- RestUtils.decodeQueryString ignores the URI fragment when parsing
  a query string
Search
- Fix the quotes in the explain message for a script score function
  without parameters
Settings
- ByteSizeValue.equals should normalize units
- Snapshot/Restore
- Snapshot restore and index creates should keep index settings and
  cluster blocks in sync
- Fix blob size in writeBlob() method
Stats
- Add extra validation into cluster/stats
- Omit current* stats for OldShardStats
Translog
- Translog recovery can repeatedly fail if we run out of disk
- Pending operations in the translog prevent shard from being marked
  as inactive

Regressions

Internal
- Deduplicate concrete indices after indices resolution

Upgrades

Core
- Upgrade Lucene to 5.3.1
- Upgrade to lucene-5.3.0.
Geo
- Update to spatial4j 0.5 for correct Multi-Geometry
Internal
- Update to Jackson 2.6.2
Plugin Cloud AWS
- Update AWS SDK version to 1.10.19
Plugin Discovery EC2
- Upgrade to aws 1.10.33

Log message:
Update databases/elasticsearch to 2.0.0.

Breaking changes
- Remove option to configure custom config file via CONF_FILE
  or -Des.default.conf

Deprecations
- Deprecate index.analysis.analyzer.default_index in favor
  of index.analysis.analyzer.default.

Enhancements
- Make PerThreadIDAndVersionLookup per-segment
- TransportNodesAction shouldn't hold on to cluster state
- Packaging: change permissions/ownership of config dir
- Release: Fix package repo path to only consist of major version
- Plugin Cloud GCE: Add _gce_ network host setting
- Check "plugin already installed" before jar hell check.
- Plugin script to set proper plugin bin dir attributes
- Plugin script to set proper plugin config dir attributes

Bug fixes
- CRUD: Send response for update request when it timed out
- Engine: Never wrap searcher for internal engine operations
- Fielddata: Add GeoPoint in StreamInput/StreamOutput
- Properly bind ClassSet extensions as singletons
- Streamline top level reader close listeners and forbid
  general usage
- Settings in log config file should not overwrite custom parameters
- Plugin cli tool should not create empty log files
- Mappings: Enforce metadata fields are not passed in documents
- Plugin Cloud AWS: Discovery-ec2 plugin should check discovery.type
- Parse Java system properties in plugin.bat
- PluginManager: Dont leave leftover files on unsuccessful installs
- REST: Restore support for escaped '/' as part of document id
- REST: Add missing REST spec for detect_noop
- REST: Expose nodes operation timeout in REST API
- REST: Ensure XContent is consistent across platforms
- Shadow Replicas: Don't pull translog from shadow engine
- Snapshot restore operations throttle more than specified

Upgrades
- Upgrade Netty to 3.10.5.final

NOT CLASSIFIED
- [DOCS] terminate_after is not experimental anymore
- Remove unused clear(IndexReader) method from IndexFieldData
- changed ben to been
- move render search template methods to cluster admin client
- Docs: Document ES_CLASSPATH removal in breaking changes
- Rewrite native script documentation
- Docs: Update repository docs with new major version structure
- Add a note about shard failure in the api

Log message:
Remove duplicate SHA512 digests that crept in.

Log message:
Add SHA512 digests for distfiles for databases category

Problems found with existing distfiles:
	distfiles/D6.data.ros.gz
	distfiles/cstore0.2.tar.gz
	distfiles/data4.tar.gz
	distfiles/sphinx-2.2.7-release.tar.gz
No changes made to the cstore or mariadb55-client distinfo files.

Otherwise, existing SHA1 digests verified and found to be the same on
the machine holding the existing distfiles (morden).  All existing
SHA1 digests retained for now as an audit trail.

Log message:
Update databases/elasticsearch to 1.7.3.
Fix the plugin script.

elasticsearch 1.7.3
===================

Enhancements
  Tribe Node:
    TransportNodesAction shouldn't hold on to cluster state

Bug fixes
  Internal:
    - An inactive shard is activated by triggered synced flush
    - Pending operations in the translog prevent shard from being
      marked as inactive
  Mapping:
    - Ensure more specific analyzer is used independent of the mapping order
  REST:
    - Expose nodes operation timeout in REST API
  Snapshot/Restore:
    - Snapshot restore operations throttle more than specified
  Tribe Node:
    - Increment tribe node version on updates

Upgrades
  Network:
    - Upgrade Netty to 3.10.5.final

elasticsearch 1.7.2
===================

Enhancements
  Allocation:
    - Avoid extra reroutes of delayed shards in RoutingService
  Packaging:
    - Shade joda-convert
  Query DSL:
    - Add a check if FieldMappers exists

Bug fixes
  Allocation:
    - Take relocating shard into consideration during awareness allocation
    - Take initializing shards into consideration during awareness allocation
    - Fix messaging about delayed allocation
  Internal:
    - Fix concurrency issue in PrioritizedEsThreadPoolExecutor.
  Nested Docs:
    - Nested query should only use bitset cache for parent filter
  Packaging:
    - Fix upgrade RPM script
  Plugins:
    - Plugin script: Fix ES_HOME with spaces
  Query DSL:
    - Do not track named queries that are null
    - multi_match query applies boosts too many times.
  Recovery:
    - Rethrow exception during recovery finalization even if source is
      not broken
  Snapshot/Restore:
    - Snapshot restore request should accept indices options
  Suggesters:
    - Prevent DirectCandidateGenerator to reuse an unclosed analyzer

Log message:
Update databases/elasticsearch to 1.7.1.

elasticsearch 1.7.1
===================

Deprecations
  Geo:
    Deprecate validate_* and normalize_*

Enhancements
  Logging:
    Add -XX:+PrintGCDateStamps when using GC Logs

Bug fixes
  Aggregations:
    Fix cidr mask conversion issue for 0.0.0.0/0 and add tests
  Core:
    ThreadPools: schedule a timeout check after adding command to queue
  Internal:
    IndicesStore shouldn't try to delete index after deleting a shard
  Plugins:
    Plugin script: Fix ES_HOME with spaces
  Query DSL:
    Fix malformed query generation
    QueryString ignores maxDeterminizedStates when creating a WildcardQuery
    Fix RegexpQueryBuilder#maxDeterminizedStates
  Scripting:
    Consistently name Groovy scripts with the same content
  Search:
    _only_nodes preference parsed incorrectly
    Copy headers from the MLT request before calling the multi-termvectors API
  Settings:
    Add explicit check that we have reached the end of the settings stream
      when parsing settings
    Copy the classloader from the original settings when checking for prompts

elasticsearch 1.7.0
===================

Breaking changes
  Allocation:
    Default delayed allocation timeout to 1m from 0

New features
  Allocation:
    Optional Delayed Allocation on Node leave
  Recovery:
    Add basic recovery prioritization to GatewayAllocator

Enhancements
  Allocation:
    Simplify ShardRouting and centralize move to unassigned
  Cluster:
    Remove scheduled routing
    Reset registeredNextDelaySetting on reroute
    Add Unassigned meta data
  Exceptions:
    Reduce the size of the XContent parsing exception
  Internal:
    Remove reroute with no reassign
    Mark store as corrupted instead of deleting state file on engine failure
  REST:
    Create Snapshot: remove _create from POST path to match PUT
    Add rewrite query parameter to the indices.validate_query API spec
  Search:
    Search preference based on node specification
  Snapshot/Restore:
    Backport to 1.7 - Snapshot info should contain version of elasticsearch
      that created the snapshot
    Add validation of snapshot FileInfo during parsing
  Term Vectors:
    Only load term statistics if required
  Upgrade:
    Upgrade groovy from 2.4.0 to 2.4.4

Bug fixes
  Allocation:
    Shard Started messages should be matched using an exact match
    Reroute after node join is processed
  Core:
    Throw LockObtainFailedException exception when we can't lock
      index directory
    Only clear open search ctx if the index is delete or closed via API
    Workaround deadlock on Codec initialisation
  Discovery:
    ZenDiscovery: #11960 failed to remove eager reroute from node join
  Highlighting:
    Fix exception for plain highlighter and huge terms for Lucene 4.x
  Index APIs:
    Use IndexWriter.hasPendingChanges() to detect if a flush is needed.
  Internal:
    Fix FieldDataTermsFilter.equals.
    Add a null-check for XContentBuilder#field for BigDecimals
  More Like This:
    Fix potentially unpositioned enum
  Packaging:
    Fix endless looping if starting fails
    Create PID_DIR in init.d script
  Percolator:
    Support filtering percolator queries by date using now
    Fail nicely if nested query with inner_hits is used in a percolator query
  Query DSL:
    CommonTermsQuery fix for ignored coordination factor
  Scroll:
    Append the shard top docs in such a way to prevent AOOBE
  Search:
    Free all pending search contexts if index is closed or removed
  Settings:
    Do not prompt for node name twice
  Shadow Replicas:
    Fail engine without marking it as corrupt when recovering on SharedFS
  Snapshot/Restore:
    Add url repository whitelist - backport of #11687 to 1.6 and 1.7
    Improve repository verification failure message
    Aborting snapshot might not abort snapshot of shards in very early
      stages in the snapshot process
    Improve logging of repository verification exceptions.
  Stats:
    Fix wrong reused file bytes in Recovery API reports
    Fix RecoveryState timestamps

Regression
  More Like This:
    Support for deprecated percent_terms_to_match REST parameter

elasticsearch 1.6.1
===================

Breaking changes
  Snapshot/Restore:
    Url repository should respect repo.path for file urls

Enhancements
  Exceptions:
    Reduce the size of the XContent parsing exception
  REST:
    Create Snapshot: remove _create from POST path to match PUT
    Add rewrite query parameter to the indices.validate_query API spec
  Snapshot/Restore:
    Add validation of snapshot FileInfo during parsing
    Add snapshot name validation logic to all snapshot operations
  Term Vectors:
    Only load term statistics if required
  Upgrade:
    Upgrade groovy from 2.4.0 to 2.4.4

Bug fixes
  Core:
    Throw LockObtainFailedException exception when we can't lock
      index directory
    Only clear open search ctx if the index is delete or closed via API
    Workaround deadlock on Codec initialisation
    Consistently add one more maxMerge in ConcurrentMergeSchedulerProvider
  Highlighting:
    Fix exception for plain highlighter and huge terms for Lucene 4.x
  Index APIs:
    Use IndexWriter.hasPendingChanges() to detect if a flush is needed.
  Internal:
    Fix FieldDataTermsFilter.equals.
    Add a null-check for XContentBuilder#field for BigDecimals
    AsyncShardFetch can hang if there are new nodes in cluster state
  Logging:
    Use task's class name if not a TimedPrioritizeRunnable
  More Like This:
    Fix potentially unpositioned enum
  Packaging:
    Fix endless looping if starting fails
    Postrm script should not fail
    Create PID_DIR in init.d script
  Percolator:
    Support filtering percolator queries by date using now
    Fail nicely if nested query with inner_hits is used in a percolator query
  Query DSL:
    CommonTermsQuery fix for ignored coordination factor
    Fix support for _name in some queries
  Scroll:
    Append the shard top docs in such a way to prevent AOOBE
  Search:
    Free all pending search contexts if index is closed or removed
  Settings:
    Do not prompt for node name twice
  Shadow Replicas:
    Return empty CommitID from ShadowEngine#flush
  Snapshot/Restore:
    Add url repository whitelist - backport of #11687 to 1.6 and 1.7
    Improve repository verification failure message
    Aborting snapshot might not abort snapshot of shards in very early
      stages in the snapshot process
    Improve logging of repository verification exceptions.
  Stats:
    Fix wrong reused file bytes in Recovery API reports
    Fix RecoveryState timestamps

Regression
  More Like This:
    Support for deprecated percent_terms_to_match REST parameter
    Add back support for deprectated percent_terms_to_match REST parameter