902c37c64e
== 5.1.1 Release Notes === Breaking changes Aliases:: - Validate alias names the same as index names REST:: - Remove lenient stats parsing 5.x === Breaking Java changes Discovery:: - Remove pluggability of ElectMasterService Exceptions:: - Remove `IndexTemplateAlreadyExistsException` and `IndexShardAlreadyExistsException` - Replace IndexAlreadyExistsException with ResourceAlreadyExistsException - Backport: Replace IndexAlreadyExistsException with ResourceAlreadyExistsException Internal:: - Pass executor name to request interceptor to support async intercept calls Network:: - Unguice Transport and friends Plugins:: - Plugins: Remove support for onModule === Deprecations Analysis:: - Deprecating request parameters of _analyze API in 5.x CRUD:: - Deprecate VersionType.FORCE Core:: - Add deprecation logging for users that explicitly opt in for the `default` fs type. Mapping:: - Deprecate `timestamp` and `ttl` on index requests. Query DSL:: - Add deprecation logging for lenient boolean queries - Add deprecation logging message for 'fuzzy' query Search:: - Deprecate ignored type parameter in search_shards api Settings:: - Add deprecation logging for the case that store throttling is used. === New features Analysis:: - Expose Lucenes Ukrainian analyzer CAT API:: - Provides a cat api endpoint for templates. Query DSL:: - Add "all fields" execution mode to simple_query_string query - Add support for `quote_field_suffix` to `simple_query_string`. - Add "all field" execution mode to query_string query Reindex API:: - Add automatic parallelization support to reindex and friends === Enhancements Aggregations:: - Rescorer should be applied in the TopHits aggregation Allocation:: - Balance step in BalancedShardsAllocator for a single shard - Process more expensive allocation deciders last - Separates decision making from decision application in BalancedShardsAllocator - Split allocator decision making from decision application Analysis:: - Remove AnalysisService and reduce it to a simple name to analyzer mapping CAT API:: - Adding built-in sorting capability to _cat apis. - Add health status parameter to cat indices API Cache:: - Do not cache term queries. - Parse alias filters on the coordinating node Cluster:: - Cache successful shard deletion checks Core:: - Reduce memory pressure when sending large terms queries. - Install a security manager on startup - Log node ID on startup - Ensure source filtering automatons are only compiled once - Improve scheduling fairness when batching cluster state changes with equal priority - Do not log full bootstrap checks exception Exceptions:: - Add BWC layer for Exceptions Geo:: - Optimize geo-distance sorting. Ingest:: - add `ignore_missing` option to SplitProcessor Internal:: - Rename ClusterState#lookupPrototypeSafe to `lookupPrototype` and remove "unsafe" unused variant - ShardActiveResponseHandler shouldn't hold to an entire cluster state - Remove unused ClusterService dependency from SearchPhaseController - Remove special case in case no action filters are registered - Use TimveValue instead of long for CacheBuilder methods - Remove SearchContext#current and all it's threadlocals Java REST Client:: - Provide error message when rest request path is null Logging:: - Log failure to connect to node at info instead of debug - Truncate log messages from the end - Logging shutdown hack - Disable console logging Mapping:: - Create the QueryShardContext lazily in DocumentMapperParser. Network:: - Grant Netty permission to read system somaxconn - Lazy resolve unicast hosts - Fix handler name on message not fully read - Handle rejected pings on shutdown gracefully Packaging:: - Add empty plugins dir for archive distributions - Make explicit missing settings for Windows service - Change permissions on config files Plugin Lang Painless:: - Add Debug.explain to painless - Implement the ?: operator in painless - In painless suggest a long constant if int won't do - Support decimal constants with trailing [dD] in painless - Implement reading from null safe dereferences - Painless negative offsets Plugin Repository S3:: - Make the default S3 buffer size depend on the available memory. Plugins:: - Clarify that plugins can be closed - Plugins: Convert custom discovery to pull based plugin - Removing plugin that isn't installed shouldn't trigger usage information - Remove pluggability of ZenPing - Make UnicastHostsProvider extension pull based Query DSL:: - Using ObjectParser in MatchAllQueryBuilder and IdsQueryBuilder - Expose splitOnWhitespace in `Query String Query` - Throw error if query element doesn't end with END_OBJECT - Remove `lowercase_expanded_terms` and `locale` from query-parser options. Reindex API:: - Make reindex-from-remote ignore unknown fields Scripting:: - Wrap VerifyError in ScriptException - Support binary field type in script values - Mustache: Add - Expose `ctx._now` in update scripts Search:: - Add indices and filter information to search shards api output - remove pointless catch exception in TransportSearchAction - Optimize query with types filter in the URL (t/t/_search) - Makes search action cancelable by task management API Search Templates:: - Add profile and explain parameters to template API Snapshot/Restore:: - Abort snapshots on a node that leaves the cluster Stats:: - Remove load average leniency - Strengthen handling of unavailable cgroup stats - Add basic cgroup CPU metrics Task Manager:: - Add search task descriptions Tribe Node:: - Add support for merging custom meta data in tribe node === Bug fixes Aggregations:: - Rewrite Queries/Filter in FilterAggregationBuilder and ensure client usage marks query as non-cachable - Percentiles bucket fails for 100th percentile - Thread safety for scripted significance heuristics Allocation:: - Allow master to assign primary shard to node that has shard store locked during shard state fetching Analysis:: - Can load non-PreBuiltTokenFilter in Analyze API CAT API:: - Consume `full_id` request parameter early Cache:: - Fix the request cache keys to not hold references to the SearchContext. Circuit Breakers:: - ClusterState publishing shouldn't trigger circuit breakers Cluster:: - Remove cluster update task when task times out Core:: - Add a StreamInput#readArraySize method that ensures sane array sizes - Use a buffer to do character to byte conversion in StreamOutput#writeString - Fix ShardInfo#toString - Protect BytesStreamOutput against overflows of the current number of written bytes. - Return target index name even if _rollover conditions are not met - .es_temp_file remains after system crash, causing it not to start again Dates:: - Fix time zone rounding edge case for DST overlaps Engine:: - Die with dignity on the Lucene layer - Fix `InternalEngine#isThrottled` to not always return `false`. - Retrying replication requests on replica doesn't call `onRetry` Highlighting:: - Fix FiltersFunctionScoreQuery highlighting - Fix highlighting on a stored keyword field Index APIs:: - Validate the `_rollover` target index name early to also fail if dry_run=true Index Templates:: - Fix integer overflows when dealing with templates. Ingest:: - fix trace_match behavior for when there is only one grok pattern - Stored scripts and ingest node configurations should be included into a snapshot Inner Hits:: - Skip adding a parent field to nested documents. Internal:: - Rethrow ExecutionException from the loader to concurrent callers of Cache#computeIfAbsent - Restore thread's original context before returning to the ThreadPool - Fix NPE in SearchContext.toString() - Source filtering should treat dots in field names as sub objects. Java API:: - Transport client: Fix remove address to actually work - Add a HostFailureListener to notify client code if a node got disconnected - Fix InternalSearchHit#hasSource to return the proper boolean value - Null checked for source when calling sourceRef - ClusterAdminClient.prepareDeletePipeline method should accept pipeline id to delete Java REST Client:: - Rest client: don't reuse the same HttpAsyncResponseConsumer across multiple retries Logging:: - Do not prematurely shutdown Log4j - Assert status logger does not warn on Log4j usage - Fix logger names for Netty Mapping:: - Fail to index fields with dots in field names when one of the intermediate objects is nested. - Uncommitted mapping updates should not efect existing indices Network:: - DiscoveryNode and TransportAddress should preserve host information - Die with dignity on the network layer - Prevent double release in TcpTransport if send listener throws an exception Packaging:: - Set vm.max_map_count on systemd package install - Export ES_JVM_OPTIONS for SysV init - Debian: configure start-stop-daemon to not go into background - Generate POM files with non-wildcard excludes Plugin Lang Painless:: - Test fix for def equals in Painless - Fix a VerifyError bug in Painless - Fix Lambdas in Painless to be Able to Use Top-Level Variables Such as params and doc Plugin Mapper Attachment:: - NPE is raised when defining a non existing type within attachments type Query DSL:: - Fixes date range query using epoch with timezone - Allow overriding all-field leniency when `lenient` option is specified - Max score should be updated when a rescorer is used REST:: - Strict level parsing for indices stats - The routing query string param is supported by mget but was missing from the rest spec - fix thread_pool_patterns path variable definition - ensure the XContentBuilder is always closed in RestBuilderListener - XContentBuilder: Avoid building self-referencing objects Reindex API:: - Ignore IllegalArgumentException with assertVersionSerializable - Bump reindex-from-remote's buffer to 200mb - Fix reindex-from-remote for parent/child from <2.0 Search:: - Fix match_phrase_prefix on boosted fields - Respect default search timeout - Remove LateParsingQuery to prevent timestamp access after context is frozen Search Templates:: - SearchTemplateRequest to implement CompositeIndicesRequest Settings:: - Handle spaces in `action.auto_create_index` gracefully - Fix settings diff generation for affix and group settings - Don't reset non-dynamic settings unless explicitly requested Snapshot/Restore:: - Fixes shard level snapshot metadata loading when index-N file is missing - Ensures cleanup of temporary index-* generational blobs during snapshotting - Fixes get snapshot duplicates when asking for _all - Keep snapshot restore state and routing table in sync (5.x backport) Task Manager:: - Task cancellation command should wait for all child nodes to receive cancellation request before returning Tribe Node:: - Add socket permissions for tribe nodes === Upgrades Core:: - Upgrade to lucene-6.3.0. Dates:: - Update Joda Time to version 2.9.5 Logging:: - Upgrade Log4j 2 to version 2.7 Network:: - Upgrade to Netty 4.1.6 Plugin Ingest Attachment:: - Update to Tika 1.14 == 5.0.2 Release Notes === Enhancements Core:: - Install a security manager on startup Exceptions:: - Add BWC layer for Exceptions Logging:: - Truncate log messages from the end Scripting:: - Wrap VerifyError in ScriptException Snapshot/Restore:: - Abort snapshots on a node that leaves the cluster === Bug fixes Allocation:: - Allow master to assign primary shard to node that has shard store locked during shard state fetching Cluster:: - Remove cluster update task when task times out Core:: - Add a StreamInput#readArraySize method that ensures sane array sizes - Use a buffer to do character to byte conversion in StreamOutput#writeString Engine:: - Die with dignity on the Lucene layer - Fix `InternalEngine#isThrottled` to not always return `false`. Index Templates:: - Fix integer overflows when dealing with templates. Ingest:: - fix trace_match behavior for when there is only one grok pattern Internal:: - Rethrow ExecutionException from the loader to concurrent callers of Cache#computeIfAbsent - Fixes potential NullPointerException on shard closing Java API:: - Transport client: Fix remove address to actually work - Add a HostFailureListener to notify client code if a node got disconnected Logging:: - Do not prematurely shutdown Log4j Network:: - Die with dignity on the network layer Plugin Lang Painless:: - Fix a VerifyError bug in Painless - Fix Lambdas in Painless to be Able to Use Top-Level Variables Such as params and doc Search:: - Respect default search timeout Settings:: - Don't reset non-dynamic settings unless explicitly requested Tribe Node:: - Add socket permissions for tribe nodes |
||
|---|---|---|
| .. | ||
| files | ||
| patches | ||
| DESCR | ||
| distinfo | ||
| Makefile | ||
| PLIST | ||