Unmutes IndexPrimaryRelocationIT.testPrimaryRelocationWhileIndexing which was
fixed by LuceneChangesSnapshot using accurate ops history.
Signed-off-by: Nicholas Walter Knize <nknize@apache.org>
Adds default implementation for getEngineFactory in EnginePlugin. The
default just returns Optional.empty(), allowing plugin developers to
implement this plugin without implementing this method.
Signed-off-by: John Mazanec <jmazane@amazon.com>
This commit upgrades the core codebase from Lucene 8.10.1 to
lucene 9.0.0. It includes all necessary refactoring of features and
API changes when upgrading to a new major Lucene release.
Signed-off-by: Nicholas Walter Knize <nknize@apache.org>
Co-authored-by: Andriy Redko <drreta@gmail.com>
Improves the LuceneChangesSnapshot to get an accurate count of recovery
operations using sort by sequence number optimization.
Signed-off-by: Nicholas Walter Knize <nknize@apache.org>
Removes the obsolete types based .addMapping method from CreateIndexRequest and
CreateIndexRequestBuilder. Tests are refactored to use the new source only
setMapping method.
Signed-off-by: Nicholas Walter Knize <nknize@apache.org>
Removes types from the TaskResults internal index along with the getMappings
method from IndexMetadata. This is needed to further remove types from
CreateIndexRequest.
Signed-off-by: Nicholas Walter Knize <nknize@apache.org>
* [Remove] Type from TermsLookUp
Signed-off-by: Suraj Singh <surajrider@gmail.com>
* Fix unit test failure
Signed-off-by: Suraj Singh <surajrider@gmail.com>
* Discrepancy in result from _validate/query API and actual query validity
Signed-off-by: Andriy Redko <andriy.redko@aiven.io>
* Moved the validate() check later into the flow to allow range validation to trigger first
Signed-off-by: Andriy Redko <andriy.redko@aiven.io>
Removes types from UID class along with cleaning up all obsolete MapperService
dependendices. This includes removing UID from the Translog Delete operation
which is no longer needed due to type dependency removal.
Signed-off-by: Nicholas Walter Knize <nknize@apache.org>
Removes multi-type support from IndexTemplateMetadata so that instead of holding
a map of multiple types to mappings, it only returns a single mapping for a
single type. Also removes type from documentMapper() method to avoid any
accidental NullPointerExceptions in the internal mapping retrieval.
Signed-off-by: Nicholas Walter Knize <nknize@apache.org>
This inner class is no longer required because its functionality has been moved to the generic GatedCloseable class.
Signed-off-by: Kartik Ganesh <gkart@amazon.com>
* Reducing duplication in plugins around ref-counted releasable classes
Both AmazonEc2Reference and AmazonS3Reference duplicate the same logic - a subclass of AbstractRefCounted that also implements Releasable. This change centralizes this paradigm into a AbstractRefCountedReleasable class and supports both clients via generics. It also updates all fetching implementations to use the get() method instead of client()
Signed-off-by: Kartik Ganesh <gkart@amazon.com>
* Introduce Reference classes for the Closeable and AutoCloseable interfaces
These classes allow you to wrap a reference instance with an onClose runnable that is executed when close() is invoked. Two separate classes are needed because the close() signatures for the two interfaces are different. This change takes the first step to have implementing classes extend from these generic superclasses, before attempting to remove the subclasses entirely. The get() method is also replaced throughout the code base.
Note that there is also a separate Releasable interface that has a similar access pattern, but is implemented separately. This is used in AbstractRefCountedReleasable introduced in a prior commit
Signed-off-by: Kartik Ganesh <gkart@amazon.com>
* More improvements and refactoring
* Functionality around one-way gating is now moved to a dedicated class - OneWayGate. This replaces duplicate functionality throughout the code.
* The two *Reference classes have been renamed to Gated* since that better represents their functionality
* The AbstractRefCountedReleasable has been improved to no longer be abstract by accepting the shutdown hook. This removes the need for the inner class in ReleasableBytesReference, and further simplifies the plugin subclasses (these could probably be removed entirely).
* Finally, unit tests have been added for some classes
Signed-off-by: Kartik Ganesh <gkart@amazon.com>
* Added tests for GatedCloseable
Also updated the license information in GatedAutoCloseableTests
Signed-off-by: Kartik Ganesh <gkart@amazon.com>
* Fixing license information in new files
Signed-off-by: Kartik Ganesh <gkart@amazon.com>
* Added unit tests for RefCountedReleasable
Signed-off-by: Kartik Ganesh <gkart@amazon.com>
* [Remove] TrimUnsafeCommit logic for legacy 6.x indexes
Multiple txlog commits was introduced in legacy 7.x. Legacy 6.x indexes could
therefore not have a safe commit. Since OpenSearch 2.0 is no longer compatible
with legacy 6.x indexes, the logic to trim these unsafe commits is safely
removed.
Signed-off-by: Nicholas Walter Knize <nknize@apache.org>
* fix assertion typo
Signed-off-by: Nicholas Walter Knize <nknize@apache.org>
* rebase and incorporate pr feedback
Signed-off-by: Nicholas Walter Knize <nknize@apache.org>
* MapperService has to be passed in as null for EnginePlugins CodecService constructor
Signed-off-by: Andriy Redko <andriy.redko@aiven.io>
* Addressing code review comments
Signed-off-by: Andriy Redko <andriy.redko@aiven.io>
* Delayed CodecService instantiation up to the shard initialization
Signed-off-by: Andriy Redko <andriy.redko@aiven.io>
* Added logger (associated with shard) to CodecServiceConfig
Signed-off-by: Andriy Redko <andriy.redko@aiven.io>
* Refactored the EngineConfigFactory / IndexShard instantiation of the CodecService
Signed-off-by: Andriy Redko <andriy.redko@aiven.io>
Removes include_type_name from the high level reset client along with relevant
deprecated methods in IndicesClient. All tests are updated to remove the
parameter from the rest requests along with various toXContent methods that are
no longer required.
Signed-off-by: Nicholas Walter Knize <nknize@apache.org>
* Replace the exclusionary word whitelist with allowlist, and blacklist with denylist, in code commet and internal variable/method/class/package name.
Signed-off-by: Andreas <apre@gmx.at>
* Add '_name' field support to score functions and provide it back in explanation response
Signed-off-by: Andriy Redko <andriy.redko@aiven.io>
* Address code review comments
Signed-off-by: Andriy Redko <andriy.redko@aiven.io>
Removes the deprecated types based get, getMapping, getAsync, and
getMappingAsync methods from IndicesClient. It also removes extra nesting of
mappings belong the deprecated type named object and removes the types based
methods from the affected request classes.
Signed-off-by: Nicholas Walter Knize <nknize@apache.org>
* Reintroduce negative epoch_millis #1991
Fixes a regression introduced with Elasticsearch 7 regarding the date
field type that removed support for negative timestamps with sub-second
granularity.
Thanks to Ryan Kophs (https://github.com/rkophs) for allowing me to use
his previous work.
Signed-off-by: Breno Faria <breno.faria@intrafind.de>
* applying spotless fix
Signed-off-by: Breno Faria <breno.faria@intrafind.de>
* more conservative implementation of isSupportedBy
Signed-off-by: Breno Faria <breno.faria@intrafind.de>
* adding braces to control flow statement
Signed-off-by: Breno Faria <breno.faria@intrafind.de>
* spotless fix...
Signed-off-by: Breno Faria <breno.faria@intrafind.de>
Co-authored-by: Breno Faria <breno.faria@intrafind.de>
Removes the type parameter from Client.prepare(Index,Delete,Update) and
everywhere it's used throughout the codebase except for prepareIndex(index,
type, id) which is removed in a follow up.
Signed-off-by: Nicholas Walter Knize <nknize@apache.org>
Removes type support from DocWrite Request and Response, all derived classes,
and all places used.
Signed-off-by: Nicholas Walter Knize <nknize@apache.org>
* [Remove] types from SearchHit and Explain API
Removes type support from SearchHit and Explain API.
Signed-off-by: Nicholas Walter Knize <nknize@apache.org>
* fix failing tests
Signed-off-by: Nicholas Walter Knize <nknize@apache.org>
* move TYPES_DEPRECATION_MESSAGE from RestExplainAction to FullClusterRestartIT
Signed-off-by: Nicholas Walter Knize <nknize@apache.org>
* fix FullClusterRestartIT
Signed-off-by: Nicholas Walter Knize <nknize@apache.org>
* fix failing tests
Signed-off-by: Nicholas Walter Knize <nknize@apache.org>
* fix testSoftDelete
Signed-off-by: Nicholas Walter Knize <nknize@apache.org>
With soft deletes no longer optional, peer recovery is switched to always use the
lucene index instead of replaying operations from the translog.
Signed-off-by: Nicholas Walter Knize <nknize@apache.org>
* Remove type param from yml and integ test files
Signed-off-by: Suraj Singh <surajrider@gmail.com>
* Remove type mapping specific update API
Signed-off-by: Suraj Singh <surajrider@gmail.com>
Removes type support from the get and mget transport and rest actions along
with removal from ShardGetService.
Signed-off-by: Nicholas Walter Knize <nknize@apache.org>
Default mappings were deprecated in Legacy version 6x and cannot be added to
indexes created in legacy 7+ or any version of OpenSearch. All support for
default mappings are removed for OpenSearch 2+.
Signed-off-by: Nicholas Walter Knize <nknize@apache.org>
With types deprecation the type support is removed from internal search API
(SearchRequest and QueryShardContext).
Signed-off-by: Nicholas Walter Knize <nknize@apache.org>
* Support unordered non-overlapping intervals
This commit exposes Intervals.unorderedNoOverlaps (LUCENE-8828).
Signed-off-by: Matt Weber <matt@mattweber.org>
* Replace ordered and overlap flags with enum.
- add IntervalMode enum
- deprecate ordered
- bwc between IntervalMode and ordered flag
- update tests
- update rest tests
Signed-off-by: Matt Weber <matt@mattweber.org>
* fix annotated text plugin test
Signed-off-by: Matt Weber <matt@mattweber.org>
* bwc to 2x only
Signed-off-by: Matt Weber <matt@mattweber.org>
* BWC rest tests for ordered flag
Signed-off-by: Matt Weber <matt@mattweber.org>
* Bump com.diffplug.spotless from 5.6.1 to 6.2.0
Bumps com.diffplug.spotless from 5.6.1 to 6.2.0.
---
updated-dependencies:
- dependency-name: com.diffplug.spotless
dependency-type: direct:production
update-type: version-update:semver-major
...
Signed-off-by: dependabot[bot] <support@github.com>
* spotlessApply
Signed-off-by: Nicholas Walter Knize <nknize@apache.org>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
Co-authored-by: Nicholas Walter Knize <nknize@apache.org>
* Remove doc type specific indexing APIs and relevant changes
Signed-off-by: Suraj Singh <surajrider@gmail.com>
* Remove type param from yml and integ test files
Signed-off-by: Suraj Singh <surajrider@gmail.com>
* [Refactor] MapperService to QueryShardContext in valueFetcher
QueryShardContext encapsulates MapperService along with other mechanisms useful
for analyzers, highlighters, and readers. This commit refactors the valueFetcher
to use the QueryShardContext as a parameter instead of direct access to the
MapperService. This is an API change. Behavior of the method is unaffected.
Signed-off-by: Nicholas Walter Knize <nknize@apache.org>
* change unnecessary Optional usage
Signed-off-by: Nicholas Walter Knize <nknize@apache.org>
* spotlessApply
Signed-off-by: Nicholas Walter Knize <nknize@apache.org>
* Add regexp interval source
Add a regexp interval source provider so people can use regular
expressions inside of intervals queries.
Signed-off-by: Matt Weber <matt@mattweber.org>
* Fixes
- register regexp interval in SearchModule
- use fully-qualified name for lucene RegExp
- get rid of unnecessary variable
Signed-off-by: Matt Weber <matt@mattweber.org>
Lucene 9 changes from BigEndian to LittleEndian in DataInput and DataOutput.
The use of Lucene's Data IO classes are refactored in this commit to use
StreamInput which maintains the same method calls while preserving the byte
order of the data.
Signed-off-by: Nicholas Walter Knize <nknize@apache.org>
Lucene 9 removed CodecReader#ramBytesUsed and all file formats that no longer
consume large amounts of memory. As a result RAM estimation for segments is no
longer possible and is removed by this commit. backwards compatibility is retained
for the tranport layer.
Signed-off-by: Nicholas Walter Knize <nknize@apache.org>
Fix composite aggregation failed test cases introduce by missing_order parameter by using MissingOrder to decide null values's order in InternalBucket.
Signed-off-by: Peng Huo <penghuo@gmail.com>
Remove deprecated max_merge_at_once_explicit parameter since it has been
removed in lucene 9 and deprecated in opensearch 1.3
Signed-off-by: Nicholas Walter Knize <nknize@apache.org>
Lucene 9 removes the ability to define an explicit version on an analysis
component so remove the the version parameter and deprecation warning
prohibiting users from explicitly setting versioned analyzers.
Signed-off-by: Nicholas Walter Knize <nknize@apache.org>
* Added timeout to ensureGreen() for testClusterGreenAfterPartialRelocation
Signed-off-by: Owais Kazi <owaiskazi19@gmail.com>
* Fixed spotless imports
Signed-off-by: Owais Kazi <owaiskazi19@gmail.com>
Remove the deprecated sync flush API which was replaced by sequence
number and retention lease mechanisms and no longer used in 2.0.0.
Signed-off-by: Nicholas Walter Knize <nknize@apache.org>
Lucene 9 removes the ability to define an explicit version on an analysis
component. The version parameter is deprecated at parse time and a warning is
issued
to the user through the deprecation logger.
Signed-off-by: Nicholas Walter Knize <nknize@apache.org>
max_merge_at_once_explicit is removed in lucene 9 so the index setting is
deprecated for removal in the next major release.
Signed-off-by: Nicholas Walter Knize <nknize@apache.org>
Adds support for _first and _last parameters for missing bucket ordering in composite aggregation.
By default, if order is asc, missing_bucket at first, if order is desc, missing_bucket at last. If
missing_order is "_first" or "_last", regardless order, missing_bucket is at first or last respectively.
Signed-off-by: Peng Huo <penghuo@gmail.com>
Logically delay the request size calculation until it's absolutely required by the
indexing pressure framework.
Signed-off-by: Saurabh Singh <sisurab@amazon.com>
Add support for setting the max expansions on a wildcard interval.
The default value is still 128 and the max value is bounded by
`BooleanQuery.getMaxClauseCount()`.
Signed-off-by: Matt Weber <matt@mattweber.org>
Soft Deletes have been enabled by default since Legacy version 7.0 and made
mandatory in Version 2.0.0. This commit refactors the InternalEngine to always
use soft-deletes. It is a follow on to making soft deletes mandatory in 2.0.0.
Signed-off-by: Nicholas Walter Knize <nknize@apache.org>
* [BUG] Wait for outstanding requests to complete in LastSuccessfulSettingsUpdate test
Signed-off-by: Suraj Singh <surajrider@gmail.com>
* [BUG] Wait for outstanding requests to complete in LastSuccessfulSettingsUpdate test
Signed-off-by: Suraj Singh <surajrider@gmail.com>
When some node or set of nodes is excluded (based on some cluster setting)
BalancedShardsAllocator iterates over them in breadth first order picking 1 shard from
each node and repeating the process until all shards are balanced. Since shards from
each node are picked randomly it's possible the p and r of shard1 is relocated first
leaving behind both p and r of shard2. If the excluded nodes were to go down the
cluster becomes red.
This commit introduces a new setting "cluster.routing.allocation.move.primary_first"
that prioritizes the p of both shard1 and shard2 first so the cluster does not become
red if the excluded nodes were to go down before relocating other shards. Note that
with this setting enabled performance of this change is a direct function of number
of indices, shards, replicas, and nodes. The larger the indices, replicas, and
distribution scale, the slower the allocation becomes. This should be used with care.
Signed-off-by: Ankit Jain <jain.ankitk@gmail.com>
* [BWC] Ensure 2.x compatibility with Legacy 7.10.x
This commit fixes TransportHandshaker to send a spoofed Legacy 7.10.2 mincompat
version to ensure OpenSearch 2.x nodes can join a Legacy 7.10.x cluster for
rolling upgrade support. Without this change 7.10.x and OpenSearch 2.x mixed
cluster bwc tests would fail.
Signed-off-by: Nicholas Walter Knize <nknize@apache.org>
* fix AbstractSimpleTransportTestCase failures
Signed-off-by: Nicholas Walter Knize <nknize@apache.org>
* change v3 to local static
Signed-off-by: Nicholas Walter Knize <nknize@apache.org>
* make V_3_0_0 package private and reuse in test
Signed-off-by: Nicholas Walter Knize <nknize@apache.org>
This test fails due to equality check of majors between OpenSearch Version 1.x and LegacyESVersion 7.x. They should be equivalent but in this case the check fails and LegacyVersion.v6.8.x is chosen which was removed in 52508d5.
This change fixes the comparison logic by using `Version.compareMajor` method instead.
Signed-off-by: Rabi Panda <adnapibar@gmail.com>
* Concurrent Searching (Experimental): modify profiling implementation to support concurrent data collection
Signed-off-by: Andriy Redko <andriy.redko@aiven.io>
* Addressing code review comments
Signed-off-by: Andriy Redko <andriy.redko@aiven.io>
Nick Knize
John Mazanec
Andriy Redko
Suraj Singh
Kartik
Kartik
Daniel Doubrovkine (dB.)
Subhobrata Dey
aponb
Breno Faria
Matt Weber
dependabot[bot]
Ankit Jain
Peng Huo
Andrey Pleskach
Lukáš Vlček
Owais Kazi
Saurabh Singh
Rishikesh Pasham
Venkata Jyothsna Donapati
Sarat Vemulapalli
Rabi Panda
sharp-pixel