Elasticsearch 8.1.0 发布

Geo

• Fields API should return normalize geometries #80649 (issues: #79232, #63739)

Aggregations

• Reenable BooleanTermsIT #83421 (issue: #83351)
• backward compatibility with version 7.17.0 #83715
• ip prefix bucket reduction #83637
• reduce float and half-float values to their stored precision #83213

Allocation

• Fix updateMinNode condition #80403 (issue: #41194)
• Make *.routing.allocation.* list-based setting #80420 (issue: #77773)
• Permit metadata updates on flood-stage-blocked indices #81781
• Reroute after cluster recovery #82856 (issue: #82456)

Authorization

• Capture anonymous roles when creating API keys #81427 (issue: #81024)
• Extend fleet-server service account privileges #82600

Autoscaling

• Fix autoscaling of follower data streams #83302 (issue: #82857)

Client

• Fix for self-suppression in REST client #83568 (issue: #42223)

Distributed

• [GCE Discovery] Correctly handle large zones with 500 or more instances #83785 (issue: #83783)

Engine

• Fork to WRITE thread when failing shard #84606 (issue: #84602)

Geo

• Handle bounds properly when grid tiles crosses the dateline #83348 (issue: #83299)
• GeometryNormalizer should not fail if it cannot compute signed area #84051 (issue: #83946)

ILM+SLM

• Fix PolicyStepsRegistry‘s cachedSteps null handling #84588

Indices APIs

• Allow removing unreferenced composable data stream templates #84376 (issues: #84171, #84188)
• Simplify and speed up ExecutorSelector #83514 (issue: #82450)

Infra/Core

• Always re-run Feature migrations which have encountered errors #83918 (issue: #83917)
• Copy trace.id in threadcontext stash #83218
• Preserve context in ResultDeduplicator #84038 (issue: #84036)
• Registration of SystemIndexMigrationTask named xcontent objects #84192 (issue: #84115)
• Update system index mappings if _meta is null #83896 (issue: #83890)

Infra/REST API

• Do not allow safelisted media types on Content-Type #83448

Infra/Scripting

• Fix duplicated allow lists upon script engine creation #82820 (issue: #82778)
• Fix plumbing in double and keyword runtime fields for the scripting fields API #83392

Ingest

• Fix GeoIpDownloader startup during rolling upgrade #84000
• Short circuit date patterns after first match #83764

Machine Learning

• Allow autoscaling to work when vertical scaling is possible #84242 (issue: #84198)
• Correctly capture min stats for inference.ingest_processors in ML usage #82352
• Fail queued inference requests with cause if the process crashes #81584
• Fix NLP tokenization never_split handling around punctuation #82982
• Fix ZeroShotClassificationConfig update mixing fields #82848
• Fix bug where initial scale from 0→1 could scale too high #84244
• Fix submit after shutdown in process worker service #83645 (issue: #83633)
• Fixes categorize_text parameter validation to be parse order independent #82628 (issue: #82629)
• Record node shutdown start time for each node #84355
• Register the named X-content parser for snapshot upgrade params #84420 (issue: #84419)
• Retry anomaly detection job recovery during relocation #83456
• Return zxx for lang_ident_model_1 if no valid text is found for language identification #82746 (issue: #81933)
• Text structure finder caps exclude lines pattern at 1000 characters #84236 (issue: #83434)
• Validate vocabulary on model deployment #81548 (issue: #81470)
• Wait for model process to stop in stop deployment #83644
• Fix a bug in the tuning of the hyperparameters when training regression classification models #2128
• Improve training stability for regression and classification models #2144, #2147, #2150
• Avoid edge cases in the classification weights calculation to maximize minimum recall which could lead to only a single class being predicted #2194
• Address cause of “[CStatisticalTests.cc@102] Test statistic is nan” log errors #2196
• Address possible causes of “x = NaN, distribution = N5boost4math23students_t_distribution” log errors #2197
• Fix bug restoring data gatherer state for time of day and week anomaly detection functions. This could lead to “No queue item for time” and “Time is out of range. Returning earliest bucket index” log errors #2213

更多 Elasticsearch version 8.1.0 | Elasticsearch Guide [8.1] | Elastic