pkgsrc changes
==============
- Install sigar on supported platforms
- Add SMF support
- Cleanup
Changes in 1.3.0
================
breaking changes
- Allocation: Enable disk-space allocation decider by default
- Analysis: Improvements to StemmerTokenFilter
- Internal:
- Make transport action name available in TransportAction base class
- Thread pool rejection status code should be 429
- Remove unnecessary intermediate interfaces
- Remove unsafe unaligned memory access - illegal on SPARC
- Mappings: Update mapping on master in async manner
- REST API: Replace error code 503 with 429 when appropriate to tell client
to back off
- Security: Disable JSONP by default
new features
- Aggregations:
- Created infrastructure for changing easily the significance terms
heuristic
- Added the percentiles_rank aggregation
- Deferred aggregations prevent combinatorial explosion
- Added top_hits aggregation
- Support bounding box aggregation on geo_shape/geo_point data types
- Scripting:
- Add script engine for Lucene expressions
- Add Groovy as a scripting language, add groovy sandboxing
- Allow search templates stored in an index to be retrieved and used
at search time
- Allow to run scripts/templates stored in .scripts index
- Mappings: Add transform to document before index
- Suggester: Phrase suggest option to limit suggestions to existing phrases
enhancements
- Resiliency:
- Resend failed shard messages when receiving a cluster state still
referring to the failed shards
- Send shard exists requests if shard exists locally but is not allocated
to the node
- Don't attempt to start or fail shard if no master node can be found
- Improve handling of failed primary replica handling
- During discovery, verify connect when sending a rejoin cluster request
- During relocation, process pending mapping update in phase 2
- Set a default of 5m to recover_after_time when any of the expected*Nodes
is set
- Start master/node fault detection pinging immediately during discovery
- Before deleting shard verify that another node holds an active shard
instance
- Ensure index.version.created is consistent
- Cancel recovery if shard on the target node closes during recovery
operation
- Raise proper failure if not fully reading translog entry
- Wait till node is part of cluster state for join process
- Have a dedicated join timeout that is higher than ping.timeout for join
- Before deleting a local unused shard copy, verify we're connected
to the node it's supposed to be on
- Use Lucene built-in checksumming
- Indexing:
- Set default translog flush_threshold_ops to unlimited, to flush by byte
size by default and not penalize tiny documents
- Don't acquire dirtyLock on autoid for create
- Clear versionMap on refresh not flush
- If versionMap is too large we should trigger refresh
- Reuse Lucene's TermsEnum for faster _uid/version lookup during indexing
- Don't use AllTokenStream if no fields were boosted
- Analysis:
- Improve Hunspell error messages
- Share numeric date analyzer instances between mappings
- Added missing pre built analysis components
- PatternAnalyzer should use PatternTokenFilter instead
- More resource efficient analysis wrapping usage
- Added additional Analyzers, Tokenizers, and TokenFilters from Lucene
- Administration & Monitoring:
- Improve pending API to include current executing class
- Stats: Expose IndexWriter and versionMap RAM usage
- Improve indices stats options
- Improve cluster update settings api
- Remove field names in stats url
- Index template API: Unified PUT/POST behaviour in relation to create
parameter
- Snapshot/Restore:
- Restore an index without restoring its aliases
- Add ability to restore partial snapshots
- Aggregations:
- Extend allowed characters in aggregation name
- Add from support to top_hits aggregator
- Moved BucketsAggregator#docCounts field to IntArray
- GlobalOrdinalsStringTermsAggregator is inefficient for high-cardinality
fields
- Remove ordinals execution hint.
- Delegation of nextReader calls
- Fielddata:
- Make BytesValues.WithOrdinals more similar to Lucene's SortedSetDocValues
- Don't expose hashes in Fielddata anymore.
- Allocation:
- Allow primaries that have never been allocated to be allocated if under
the low watermark
- Change high/low watermark defaults for disk based allocation decider
- Disk-aware allocation decider should allow initial primary allocation
if under the high watermark
- Mapping:
- Improve performance for many new fields introduction in mapping
- Better logic on sending mapping update new type introduction
- Wait for mapping updates during local recovery
- Scripting:
- Add a transformer to translate constant BigDecimal to double
- Add Groovy sandboxing for GString-based method invocation
- Fix optional default script loading
- Exposed _uid, _id and _type fields as stored fields (_fields notation)
- Search:
- Wrap filter only once in ApplyAcceptedDocsFilter
- Remove Queries#optimizeQuery - already handled in BooleanQuery
- Speed up exists and missing filters on high-cardinality fields
- Parent/Child: Support min_children and max_children on has_child
query/filter
- Geo: Added caching support to geohash filter
- Highlighting:
- Make the HighlightQuery class public
- Highlight fields in request order
- Plain highlighter does not honor _analyzer
- More Like This Query:
- Ensure selection of best terms is indeed O(n)
- creates only one MLT query per field for all queried items
- Values of a multi-value fields are compared at the same level
- replaced exclude with include to avoid double ...
- allow for both like_text and docs/ids to be specified.
- Plugins:
- Enables plugins to define default logging configuration for their needs.
- bin/plugin tests for missing plugin name when passing --url
- Build:
- Check for tabs and nocommits in the code on validate
- Generate source jars for tests
- Packaging: Remove java-6 directories from debian init script
- Startup: Reset locale to C in bin/elasticsearch
- Windows: Modify command window title (windows)
- REST API:
- Copy the headers from REST requests to the corresponding
TransportRequest(s)
- Ensure 503 responses lead to retry on another node
- Security: Make JSONP responses optional.
- Internal:
- Make XContentBuilder implement Releasable
- Don't close/reopen IndexWriter when changing RAM buffer size
- Added more utility methods to Settings
- Cleanup of the transport request/response messages
- Add local node to cluster state
- Control whether MapperService docMapper iterator should contain
DEFAULT_MAPPING
- Upgrade to Jackson 2.4.1.1
- Don't replace indices within ActionRequest and check blocks against
concrete indices
- Improve Settings#get lookup for camel case support
- Improve large bytes request handling by detecting content composite
buffer
- Call callback on actual mapping processed
- Remove intern calls on FieldMapper#Names for better performance
- Clean shard bulk mapping update to only use type name
- Disable explicit GC by default
- Make a hybrid directory default using mmapfs/niofs
- Make sure we don't reuse arrays when sending an error back
- Wrap RateLimiter rather than copy RateLimitedIndexOutput
- Upgrade to Lucene 4.9
- Re-shade MVEL as a dependency
- Refactored AckedClusterStateUpdateTask & co. to remove code repetitions
in subclasses
- Better default size for global index to alias map
- Suppress Unsafe is an internal proprietary API... compilation warnings
- use ConcurrentHashMapV8 for lower memory overhead
- Base64 decode parsing detects more errors
- Upgrade to netty 3.9.1
- Added plugins to .gitignore
- Change the default type of the page recycler to CONCURRENT instead
of SOFT_CONCURRENT
- Entirely cut over to TopDocs#merge for merging shard docs in the reduce
phase
- Add support for Byte and BytesRef to the XContentBuilder
- Make Node and Client interfaces Closeable
- Settings: Throw error when incorrect setting applied
to auto_expand_replicas
- TransportClient: Improve logging, fix minor issue
- Java API: Client intermediate interfaces removal follow-up
bug fixes
- Resiliency:
- Increment Store refcount on RecoveryTarget
- Improve handling of failed primary replica handling
- Do not ignore ConnectTransportException for shard replication operations
- Recovering replicas might get stuck in initializing state
- During recovery, only send mapping updates to master if needed
- If the node initialisation fails, make sure the node environment
is closed correctly
- Handle ConnectionTransportException during a Master/Node fault detection
ping during discovery
- Recovery from local gateway should re-introduce new mappings
- Also send Refresh and Flush actions to relocation targets
- Honor time delay when retrying recoveries
- Aggregations:
- Histogram Aggregation key bug
- Fix JSON response for significant terms
- Fix reducing of range aggregations.
- Fix cardinality aggregation when doc values field is empty
- ReverseNestedAggregator does not compute parent documents correctly
- ClassCastException when sibling aggregations have the same name
- date_histogram aggregation breaks on date fields with multiple formats
- DateHistogramBuilder uses wrong data type for pre_offset and post_offset
- Administration & Monitoring:
- Admin: Recovery API should also report ongoing relocation recoveries
- Cat API: Fix NullPointerException in recovery API
- Index Templates API: GET templates doesn't honor
the flat_settings parameter.
- Snapshot/Restore: Allow deleting of interrupted snapshot
- Indexing:
- Translog: Better support for partial buffer reads/writes in translog
infrastructure
- Force refresh when versionMap is using too much RAM
- Lower the translog flush triggers to workaround
- Clear versionMap on refresh not flush
- DocumentMissingException is uncaught if thrown during retry
of update request
- Geo: Valid polygon crossing dateline fails to parse
- Search:
- The query_string cache should returned cloned Query instances.
- Replace empty bool queries with match_all to prevent NullPointerException
- Match query with operator and, cutoff_frequency and stacked tokens
- Search template not replacing parameter after initial failure
in parameter substitution
- Nested: queries/filters/aggregations expect FixedBitSets, yet it isn't
the case with NoneFilterCache
- Query DSL:
- Fix MatchQueryParser not parsing fuzzy_transpositions
- Function score without function throws NPE
- Improved explanation for match_phrase_prefix
- Mapping:
- MapperParsingException when create default mapping with 'include_in_all'
nested
- Fix possibility of losing meta configuration on field mapping update
- Allow _version to use disk as a doc values format.
- Analysis:
- stem_exclusion as array not working in language analyzers
- Default analyzer includes stopwords
- Bulk API:
- Fix return of wrong request type on failed updates
- Bulk request which try and fail to create multiple indices may
never return
- BulkProcessor does not call afterBulk when bulk throws
eg NoNodeAvailableException
- Percolator:
- Fix memory leak when percolating with nested documents
- Fix handling of nested documents
- Allow significant terms and geo hash grid aggregations in the percolator
- Plugins:
- bin/plugin removes itself
- Removing plugin does not fail when plugin dir is read only
- Fix github download link when using specific version
- Java API:
- Fix source excludes setting if no includes were provided
- BulkRequest#add(Iterable) to support UpdateRequests
- Internal:
- Fix possible NPE during shutdown for requests using timeouts
- Prevent NPE if engine is closed while version map is checked
- Fixes Logger class for BackgroundIndexer
- Check for index blocks against concrete indices on master operations
- IndexingMemoryController should only update buffer settings of fully
recovered shards
- Fix possible race condition in checksum name generator
- Lucene: Use XNativeFSLockFactory instead of the buggy
Lucene 4.8.1 version
- Disable circuit breaking for BigArrays
- Guava doesn't explicitly remove entries when clearing the entire cache
- Filter cache size limit not honored for 32GB or over
- Threadpool Info: Allow to serialize negative thread pool sizes
- Core: The ignore_unavailable option should also ignore indices
that are closed
- Remove indicesLifecycle.Listener from IndexingMemoryController
- Store: delete unallocated shards under a cluster state task
- Routing: Restore shard routing
- Scripting: Wrap groovy script exceptions in a serializable
Exception object
Changes in 1.2.3
================
bug fixes
- Indexing: DocumentMissingException is uncaught if thrown during retry
of update request
- Internal:
- Remove unsafe unaligned memory access - illegal on SPARC
- Allow to serialize negative thread pool sizes
- Plugin Manager: Properly quote $JAVA in bin/plugin
- Plugins: bin/plugin removes itself
- Query DSL: Function score without function throws NPE
Changes in 1.2.2
================
enhancements
- Internal:
- Make sure we don't reuse arrays when sending an error back
- Better default size for global index -> alias map
bug fixes
- Aggregations:
- Fix reducing of range aggregations.
- Fix cardinality aggregation when doc values field is empty
- Bulk API: Fix return of wrong request type on failed updates
- Cat API: Fix NullPointerException in cat-recovery API
- Core: The ignore_unavailable option should also ignore indices
that are closed
- Internal:
- [Translog] Better support for partial buffer reads/writes in
translog infrastructure
- [FileSystem] Use XNativeFSLockFactory instead of the buggy
Lucene 4.8.1 version
- Java API:
- Fix source excludes setting if no includes were provided
- BulkRequest#add(Iterable) to support UpdateRequests
- Mapping: Fix possibility of losing meta configuration on field
mapping update
- Mappings: Allow _version to use `disk` as a doc values format.
- Percolator:
- Fix memory leak when percolating with nested documents
- Fix handling of nested documents
- Plugins: Fix github download link when using specific version
- Query DSL: Improved explanation for match_phrase_prefix
- Scripts: exposed _uid, _id and _type fields as stored fields
(_fields notation)
- Search:
- The query_string cache should returned cloned Query instances.
- Replace empty bool queries with match_all to prevent
NullPointerExceptions
- Match query with operator and, cutoff_frequency and stacked tokens
- Snapshot/Restore: Allow deleting of interrupted snapshot
- Test: Fix possible race condition in checksum name generator
regression
- Aggregations: fix JSON response for significant terms
Changes in 1.2.1
================
enhancements
- Networking: Upgrade to Netty 3.9.1
- Cluster: have a dedicated join timeout
bug fixes
- Aggregations: Disable circuit break for aggregations
- Aggregations: The reverse nested aggregator now computes parent
documents correctly
- Indexing: Lower translog flush trigger
- Searching: Search template does not correctly replace a parameter after
an initial failure
- Searching: Fixed issue with nested queries not working when filter cache
is disabled
- Searching: The filter cache setting is now handled correctly when above
32 GB in size
regression
- Routing: Restore shard routing
- Mapping: MapperParsingException with include_in_all
Changes in 1.2.0
================
breaking changes
- Java: Use the maven enforcer plugin to work only with java 1.7
- Scripting: Disable dynamic scripting by default
- Configuration: Change default filter cache to 10% and circuit breaker
to 60% to prevent unwanted OOMs
- Snapshot/Restore API: Added PARTIAL snapshot status
- Gateways: Removed deprecated gateway functionality (in favor
of snapshot/restore)
- Versioning: Version types EXTERNAL & EXTERNAL_GTE test for
version equality in read operation & disallow them in the Update API
- Versioning: A Get request with a version set always validates for equality
- Versioning: Calling the Update API using EXTERNAL and EXTERNAL_GTE
version type throws a validation error
- Aggregations: Changed response structure of percentile aggregations
- Cluster State API: Remove index template filtering
- Nodes Stats API: Add human readable JVM start_time and process
refresh_interval
- Java API: Unified IndicesOptions constants to explain intentions
- Internal: Removed operation threading from broadcast actions
- Internal: Remove search operation threading option
- Internal: Ordinals now start with zero
- Internal: Make IndexingOperation/Create/Delete classes less mutable
- Internal: Refactored and cleaned up REST infrastructure
- Internal: Remove Releasable in favor of Closeable
new features
- Aggregations: Add reverse_nested aggregation
- Aggregations: Added extended_bounds support for date_/histogram aggs
- Aggregations: Added option for a backgroundFilter in significant_terms agg
- Function score: Support multi values
- Function score: Added field_value_factor function, a shortcut for simple
script score calculations
- Suggestions: Added support for contexts in the completion suggester
- cat API: Added /_cat/fielddata endpoint to display fielddata usage
- More like this Query API: Support searching for multiple similar documents
specified by ids
- Doc Values: Added support for binary fields
- Internal: Added FreqTermsEnum which has built-in caching for
more performance
- Lucene: Added support for uppercase token filter
enhancements
- Indexing: Don't lookup version for auto generated id and create immediately
- Indexing: Throttle incoming indexing when Lucene merges fall behind
- Indexing: Change default numeric precision_step
- Function Score API: Parser should throw exception if both options,
functions:[] and single, are specified
- Field data: Added a AppendingDeltaPackedLongBuffer-based storage format
to single value field data
- Field data: Add global ordinals
- Field data: Moved the decision to load p/c fielddata eagerly
to a better place.
- Aggregations: Add global ordinal based implementations for
significant_terms aggregation
- Aggregations: Add include/exclude support to global ordinals in terms
and significant_terms aggregations
- Aggregations: Improve the way sub-aggregations are collected.
- Aggregations: Use collectExistingBucket() if a bucket already exists
- Aggregations: Improve terms aggregation to perform the segment ordinal
to global ordinal lookup post segment collection
- Aggregations: Use segment ordinals as global ordinals if possible
- Aggregations: Instantiate facets/aggregations during the QUERY phase
- Aggregations: Refactor common code for unmapped aggregators into
NonCollectingAggregator.
- Aggregations: Fail queries that have two aggregations with the same name
- Throttling: Change default recovery throttling to 50MB / sec
- More Like This API: Fix behavior on default boost factor for More
Like This.
- More Like This API: Added the ability to include the queried document
for More Like This API.
- Index Settings API: Allow to change concurrent merge scheduling
setting dynamically
- Translog: Use unlimited flush_threshold_ops for translog to prefer
size based flushing
- Translog: Fix visibility in buffered translog
- Java API: Made it mandatory to specify IndicesOptions when calling
MetaData#concreteIndices
- Java API: Unified MetaData#concreteIndices methods into a single method
that accepts indices (or aliases) and indices options
- Cluster state: Added new ClusterStateStatus to indicate cluster
state life cycles
- Cluster state: Unify cluster state update on mapping change
- Cluster state: When sending shard start/failed message due to a cluster
state change, use the master indicated in the new state rather than current
- Parent/Child: Make use of global ordinals in parent/child queries
- Parent/Child: Fix Parent/Child assertions for rewrite reader
- Discovery: Unicast discovery enhancement
- Discovery: MulticastChannel returned wrong channel in shared mode
- Mapping API: Allow to parse lat/lon geo points as strings and coerce them
- Mapping API: Support empty properties array in mappings
- Mapping API: Norms disabling on existing fields
- Scripting: Log script change/add and removal at INFO level
- Scroll Search API: Improve scroll search performance by using Lucene's
IndexSearcher#searchAfter(...)
- REST Clear Scroll API: Return missing (404) if a scroll_id is cleared
that no longer exists
- Index Status API: deprecate index status
- Replication: Fail replica shards locally upon failures
- Configuration: Remove spaces from commented config lines in
elasticsearch.yml and logging.yml
- Circuit breaker: Include name of the field that caused a circuit break
in the log and exception message
- Networking: Use loopback when localhost is not resolved
- Settings: Add getAsRatio to Settings class, allow DiskThresholdDecider
to take percentages
- Query internals: Enforce query instance checking before it is wrapped
as a filter
- Query API: Throw exception if an additional field was placed inside
the query body
- Highlighting API: Java API does not have a way to set global highlighting
settings
- Validate Query API: Query validation did not detect extra JSON properties
after the query property
- Bulk API: Added BulkProcessor.flush() method
- Delete by Query API: Capture and set start time in Delete By Query operations
- cat API: Add configured thread pool sizes to _cat/thread_pool
- Plugins: Allow to check for Lucene version
- Format: Support CBOR data format
- Nodes stats API: Add suggest stats
- cat API: _cat/recovery endpoint now also shows relocations
- Circuit breaker: Track the number of times the CircuitBreaker has
been tripped
- Tribe node: Index level blocks, index conflict settings
- Startup: Add possbility to display full stack trace of exception
- Dependencies: Use T-digest as a dependency and remove built-in code
- Dependencies: Upgrade to Guava 17
- Dependencies: Upgrade to Lucene 4.8.0
- Dependencies: Upgrade to Lucene 4.7.2
- Dependencies: Upgrade to Lucene 4.7.1
- Dependencies: Upgrade to mvel 2.2.0.Final
- Dependencies: Update forbidden-apis to 1.5.1 and remove the relaxed
failOnMissingClasses setting, fix typo
- Dependencies: Update to forbidden-apis 1.4.1
- Dependencies: Update JNA to 4.1.0, properly warn on error, hint
at noexec mount
- Dependencies: Update shade-plugin to 2.3
- Release: Release should fail if test are annotated with @AwaitsFix
- Release: Check for no open issues before build release
- Internal: Remove SoftReferences from StreamInput/StreamOutput
- Internal: RecoveryID in IndicesClusterStateService should not be
a per JVM but per Node
- Internal: Forbid Math.abs(int/long).
- Internal: Limit the number of bytes that can be allocated
to process requests.
- Internal: Improved bloom filter hashing
- Internal: Ensure IndexEngine.close() is called under lock in the case
of an engine failure
- Internal: Make StreamOutput.writePrimitive*() and
StreamInput.readPrimitive*() methods public.
- Internal: LongHash add/key not consistent
- Internal: Added releasable bytes output, which is uses
in transport / translog
- Internal: Make Releasable extend AutoCloseable.
- Internal: Fail the engine/shard when refresh failed
- Internal: BytesReference.Helper should never materialize a byte[] array.
- Internal: XFilteredQuery defaults to Query First strategy [OPEN]
- Internal: Internally manipulate the terms execution hint as an enum
instead of a String.
- Lucene: Use non analyzed token stream optimization everywhere
- Lucene: Enable turning on IndexWriter's infoStream
- Lucene: Prevent fsync from creating 0-byte files
- Lucene: Exposed LMSimilarity in Elasticsearch
- Lucene: Remove SerialMergeScheduler
- Lucene: Include thread name when logging IndexWriter's infoStream messages
- Lucene: Fix Lucene's getFiniteStrings() to not consume Java stack
- Transport API: Raise node disconnected exception even if the transport
is stopped
- Cleanup: Some minor cleanups from findbugs report
- Cleanup: Fix code typo in FieldSortBuilder.java
- Cleanup: Cleaning up FileSystemUtils class
- Cleanup: REST layer refactoring phase 2 + recycling in HTTP layer
- Cleanup: Aggregation cleanup
- Cleanup: Aggregations cleanup
- Cleanup: StringBuilder should be used in place of StringBuffer
- Cleanup: Variable renamings to reduce unnecessary variable naming diversity
- Cleanup: Remove unused dump infra
- Cleanup: Field data diet
bug fixes
- Mapping API: stem_exclusion as array not working in language analyzers
- Mapping API: Remove RootMapper.validate and validate the routing
key up-front
- Mapping API: Check if root mapping is actually valid
- Mapping API: geo_point doesn't allow null values as of 1.1.0
- Mapping API: Fix mapping creation on bulk request
- Mapping API: Fix include_in_all for multi fields
- Mapping API: GeoPointFieldMapper did not merge GeoPoint specific properties
- Mapping API: Throw error when updating binary field mapping to be stored
- Mapping API: Fix dynamic_type in dynamic_template
- Query API: Search might not return on thread pool rejection
- Query API: limit filter returns wrong results if deleted document
are present
- Query API (internal): Don't report terms as live if all it's docs are
filtered out
- Query API: Executing match_query to search only for an integer could
result in a NullPointerException
- Query API: Lots of warn logs related to FreeContext due to not
deserializing the full message
- Query API: Allow sorting on nested sub generated field
- Query API: optimize_bbox for geo_distance filters could cause
missing results
- Query API: Range/Term query/filter on dates fail to handle numbers
properly
- Query API: Parse has_child query/filter after child type has been parsed
- Query API: TemplateQueryParser swallowed additional parameters
- Query API: match_phrase_prefix broken in case of single term queries
- Search Template API: Made template endpoint compatible with search endpoint
- Search Template API: search_template does not support ?source=
- Validate Query API: Validate query ignores type filter
- Validate Query API: Validate query ignores alias filters
- Validate Query API: Validate query without a body throws an NPE
- More Like This API: Take size and from in request into account
- Suggest API: tie-break suggestions by term in phrase suggester
- Suggest API: Fix various context suggester issues
- Suggest API: Require precision in mapping for context suggester
geo locations
- Suggest API: Renamed category type to "category" instead of "field"
in context suggester
- Delete by Query API: Disabled parent/child queries in the delete
by query api
- Analyze API: Default analyzer includes stopwords
- cat API: _cat/allocation return values are not consistent for clients nodes
- Scroll API: Reduce phase fails if shard failures occur
- Snapshot/Restore API: Unregistering snapshot repositories causes
thread leaks
- Snapshot/Restore API: Fix hanging aborted snapshot during shutdown
- Snapshot/Restore API: Fix for hanging aborted snapshot during node shutdown
- Snapshot/Restore API: Snapshot Status failing without repository
- Snapshot/Restore API: Add an ability to snapshot relocating primary shards
- Snapshot/Restore API: Switch to shared thread pool for all snapshot
repositories
- Field data: Fix setting of readerGen in BytesRefOrdValComparator
on nested documents
- Field data: Provide meaningful error message if field has no fielddata type
- Field data: Add a dedicated field data type for the _index field mapper
- Field data: Remove ScriptDocValues.EMPTY
- Field data: Eager fielddata loading not working
- Field data: Global ordinals cause ClassCastExceptions if used with
a bounded fielddata cache
- Search Scroll API: Throw better error if invalid scroll id is used
- Search Scroll API: Missing scroll ID no longer returns exception
- Clear Scroll API: Clear scroll should accept scroll_id in body
- Percolation API: Percolator doesn't reduce CircuitBreaker stats
in every case.
- Percolation API: The percolator needs to take deleted percolator
documents into account.
- Percolation API: Propagate percolate mapping changes to cluster state
- Percolation API: Fix assertion in percolation with nested docs
- Aggregations: Parsing is now more strict
- Aggregations: DateHistogram.Bucket should return the date key in UTC
- Aggregations: A nested nested aggregation falls outside of its parent
nested aggregation bounds
- Aggregations: Add shard_min_doc_count parameter to terms aggregation
- Aggregations: Add shard_min_doc_count parameter for significant terms
similar to shard_size
- Aggregations: Lower the initial sizing of sub aggregations
- Aggregations: Remove abstraction in the percentiles aggregation.
- Allocation: Failed shards could be re-assigned to the same nodes
if multiple replicas failed at once
- Allocation: Fix format string for DiskThresholdDecider reroute explanation
- cat API: ElasticsearchIllegalStateException when invoking
/_cat/plugin endpoint
- cat API: Node version sometimes empty in _cat/nodes endpoint
- Transport: Fixed NPE when initializing an accepted socket in NettyTransport.
- Transport: A master node operation can timeout in a rare race condition
when a master has a connection issue which is quickly restored
- Recovery: Do not start a recovery process if the primary shard is currently
allocated on a node which is not part of the cluster state
- Recovery API: Percent bytes recovered greater than 100%
- REST Recovery API: Empty HTTP body returned from _recovery API on empty
cluster
- Throttling: Store IO throttling throttles far more than asked
- Merging: Ensure pending merges are updated on segment flushes
- Time-to-live: TTL Purge Thread might bring back already deleted index
- Tribe node: Fix possible NullPointerException if index doesn't exist
- Cluster State: Added the cluster name as its identifier
- Function score: Fixing questionable PNRG behavior in random score
- Discovery: Count latch down if sendsPing throws exception
- Bulk API: Ensure that specific failures do not affect whole request
- Search API: Search might not return on thread pool rejectio
- Cluster State: Added tighter check for cluster state change on master
node operations
- Build: Mustache dependency not shaded
- Internal: Restore read/write visibility in PlainShardsIterator.
- Internal: Fix bug in PropertyPlaceholder and add unit tests
- Internal: MetaData#concreteIndices now throws exception with a single
index argument if allowNoIndices == false
- Internal: Closing an IndexReader on an already relocated / closed shard
can cause memory leaks
- Internal: Make sure successful operations are correct if second search
phase is fast
- Internal: Aggregations could keep pages in case of errors
- Internal: Fixed possible NullPointerException in PagedBytesReference
by taking stream position into account
- Internal: Replace InternalSearchResponse#EMPTY
with InternalSearchResponse#empty()
- Internal: JsonXContentGenerator#writeRawField produced invalid JSON
if raw field was first field in the json object
- Internal: IndexShardRoutingTable might barf due to an overflow if it
has handled lots of searches
- Internal: Searcher might not be closed if store handle could
not be obtained
- Cleanup: Fix some warnings reported by findbugs
regression
- Merging: Switch back to ConcurrentMergeScheduler as the default
- Versioning: Indexing a document fails when setting version=0 &
version_type=external
Changes in 1.1.2
================
enhancements
- Dependencies: Upgrade to mvel 2.2.0.Final, supporting java 8
bug fixes
- Percolation API: Fix assertion in percolation with nested docs
- Percolation API: Fix assertion in percolation with nested docs
- Percolation API: The percolator needs to take deleted percolator
documents into account.
- Query API: limit filter returns wrong results if deleted document
are present
- Query API: Executing match_query to search only for an integer could
result in a NullPointerException
- Query API: Allow sorting on nested sub generated field
- Query API: Search might not return on thread pool rejection
- Query API: optimize_bbox for geo_distance filters could cause
missing results
- Query API: Range/Term query/filter on dates fail to handle numbers properly
- Query API: TemplateQueryParser swallowed additional parameters
- Query API: Parse has_child query/filter after child type has been parsed
- Validate Query API: Validate query ignores type filter
- Validate Query API: Validate query ignores alias filters
- Validate Query API: Validate query without a body throws an NPE
- Snapshot/Restore API: Fix for hanging aborted snapshot during node shutdown
- Snapshot/Restore API: Fix for hanging aborted snapshot during node shutdown
- Snapshot/Restore API: Fix hanging aborted snapshot during shutdown
- Aggregations: Fail queries that have two aggregations with the same name
- Mapping API: stem_exclusion as array not working in language analyzers
- Mapping API: Remove RootMapper.validate and validate the routing
key up-front
- Build: Mustache dependency not shaded
- Transport: Fixed NPE when initializing an accepted socket in NettyTransport.
- Internal: Restore read/write visibility in PlainShardsIterator.
- Internal: Fix bug in PropertyPlaceholder and add unit tests
- Field data: Fix setting of readerGen in BytesRefOrdValComparator
on nested documents
- Field data: Provide meaningful error message if field has no fielddata type
- Analyze API: Default analyzer includes stopwords
- cat API: _cat/allocation return values are not consistent for clients nodes
- Delete by Query API: Disabled parent/child queries in the delete
by query api
- Search API: Search might not return on thread pool rejection
Changes in 1.1.1
================
enhancements
- Upgrade to Lucene 4.7.2
- Geo Point Mapping: Allow to parse lat/lon as strings and coerce them
- Test cluster: Moved wipe* methods, randomIndexTemplate &
ensureEstimatedStats to TestCluster
bug fixes
- Mapping API: Fix include_in_all for multi field
- Mapping API: Throw error when updating binary field mapping to be stored
- Mapping API: geo_point doesn't allow null values as of 1.1.0
- Mapping API: Ensure mapping is propagated to master at all times, even
if index creation failed
- Mapping API: Ensure that TTL cleanup thread does not accidentally create
indices again
- Mapping API: Propagate percolate mapping changes to cluster state
- Snapshot/Restore API: Snapshot status failing without repository
- Snapshot/Restore API: SnapshotMetaData.fromXContent does not match
toXContent, but throws an Exception
- Internal: Ensure pending merges are updated on segment flushes
- Internal: Replace InternalSearchResponse#EMPTY with
InternalSearchResponse#empty() to prevent out-of-scope modification
of that instance
- Internal: Return to use ConcurrentMergeScheduler
- Internal: Closing an IndexReader on an already relocated / closed shard
could cause memory leaks
- cat API: _cat plugin endpoint throws an exception
- Allocation: Fix format string for DiskThresholdDecider reroute explanation
- Allocation: Failed shards could be re-assigned to the same nodes if
multiple replicas failed at once
- Aggregations: A nested nested aggregation falls outside of its parent
nested aggregation bounds
- Field data: Percolator doesn't reduce CircuitBreaker stats in every case
- Field data: Fixed eager fielddata loading
- Bulk API: Ensure that specific failures do not affect whole request
- Search API: Prevent ArrayIndexOutOfBoundsSelection on counter which selects
a random shard in case of a lot of searches
- Search Template API: endpoint now supports source parameter
- REST Scroll API: Clear scroll should accept scroll_id in body
- REST Scroll API: Throw useful error when invalid scroll_id is specified
- Search stats: Make sure successful operations are counted correct if
second search phase is fast
- Scripting: ScriptDocValues.EMPTY doesn't implement getValue, resulting
in different behaviour if no document in a segment had a value
- Cluster State: Clusterstate misses the cluster name as it's identifier
- Tribe node: Fixing errors on start up
- Index Template API: Fixed issue with dynamic mapping in Elasticsearch
3cf72e3a3a