Elasticsearch release notes

Review the changes, fixes, and more in each version of Elasticsearch.

To check for security updates, go to Security announcements for the Elastic stack.

Release date: March 25, 2025

Allocation

• Only publish desired balance gauges on master #115383

Authorization

• Add a monitor_stats privilege and allow that privilege for remote cluster privileges #114964
• [Security Solution] Add create_index to kibana_system role for index/DS .logs-endpoint.action.responses-* #115241

CRUD

• Suppress merge-on-recovery for older indices #113462

Data streams

• Adding a deprecation info API warning for data streams with old indices #116447
• Apm-data: disable date_detection for all apm data streams #116995
• Add default ILM policies and switch to ILM for apm-data plugin #115687

Distributed

• Metrics for incremental bulk splits #116765
• Use Azure blob batch API to delete blobs in batches #114566

ES|QL

• Add ES|QL bit_length function #115792
• ESQL: Honor skip_unavailable setting for nonmatching indices errors at planning time #116348 (issue: #114531)
• ESQL: Remove parent from FieldAttribute #112881
• ESQL: extract common filter from aggs #115678
• ESQL: optimise aggregations filtered by false/null into evals #115858
• ES|QL CCS uses skip_unavailable setting for handling disconnected remote clusters #115266 (issue: #114531)
• ES|QL: add metrics for functions #114620
• Esql Enable Date Nanos (tech preview) #117080
• Support partial sort fields in TopN pushdown #116043 (issue: #114515)
• [ES|QL] Implicit casting string literal to intervals #115814 (issue: #115352)
• Add support for BYTE_LENGTH scalar function #116591
• Esql/lookup join grammar #116515
• Remove snapshot build restriction for match and qstr functions #114482

Health

• Increase replica_unassigned_buffer_time default from 3s to 5s #112834

Indices APIs

• Ensure class resource stream is closed in ResourceUtils #116437

Inference

• Add version prefix to Inference Service API path #117095

Infra/Circuit Breakers

• Add link to Circuit Breaker "Data too large" exception message #113561

Infra/Core

• Support for unsigned 64 bit numbers in Cpu stats #114681 (issue: #112274)

Infra/Metrics

• Add ensureGreen test method for use with adminClient #113425

Infra/Scripting

• Add a mustache.max_output_size_bytes setting to limit the length of results from mustache scripts #114002

Ingest Node

• Add postal_code support to the City and Enterprise databases #114193
• Add support for registered country fields for maxmind geoip databases #114521
• Adding support for additional mapping to simulate ingest API #114742
• Adding support for simulate ingest mapping adddition for indices with mappings that do not come from templates #115359
• Support IPinfo database configurations #114548
• Support more maxmind fields in the geoip processor #114268

Logs

• Add logsdb telemetry #115994
• Add num docs and size to logsdb telemetry #116128
• Feature: re-structure document ID generation favoring _id inverted index compression #104683

Machine Learning

• Add DeBERTa-V2/V3 tokenizer #111852
• Add special case for elastic reranker in inference API #116962
• Adding inference endpoint validation for AzureAiStudioService #113713
• Adds support for input_type field to Vertex inference service #116431
• Enable built-in Inference Endpoints and default for Semantic Text #116931
• Increase default queue_capacity to 10_000 and decrease max queue_capacity to 100_000 #115041
• Inference duration and error metrics #115876
• Remove all mentions of eis and gateway and deprecate flags that do #116692
• [Inference API] Add API to get configuration of inference services #114862
• [Inference API] Improve chunked results error message #115807

Network

• Allow http unsafe buffers by default #116115

Recovery

• Attempt to clean up index before remote transfer #115142 (issue: #104473)
• Trigger merges after recovery #113102

Reindex

• Change Reindexing metrics unit from millis to seconds #115721

Relevance

• Add query rules retriever #114855
• Add tracking for query rule types #116357

Search

• Add Search Phase APM metrics #113194
• Add docvalue_fields Support for dense_vector Fields #114484 (issue: #108470)
• Add initial support for semantic_text field type #113920
• Adds access to flags no_sub_matches and no_overlapping_matches to hyphenation-decompounder-tokenfilter #115459 (issue: #97849)
• Better sizing BytesRef for Strings in Queries #115655
• Enable _tier based coordinator rewrites for all indices (not just mounted indices) #115797
• Only aggregations require at least one shard request #115314
• ESQL - Add match operator (:) #116819
• Upgrade to Lucene 10 #114741

Security

• Add refresh .security index call between security migrations #114879

Snapshot/Restore

• Improve message about insecure S3 settings #116915
• Retry S3BlobContainer#getRegister on all exceptions #114813
• Split searchable snapshot into multiple repo operations #116918
• Track shard snapshot progress during node shutdown #112567

Vector Search

• Add support for bitwise inner-product in painless #116082

Aggregations

• Handle with illegalArgumentExceptions negative values in HDR percentile aggregations #116174 (issue: #115777)

Analysis

• Adjust analyze limit exception to be a bad_request #116325

CCS

• Fix long metric deserialize & add - auto-resize needs to be set manually #117105 (issue: #116914)

CRUD

• Preserve thread context when waiting for segment generation in RTG #114623
• Standardize error code when bulk body is invalid #114869

Data streams

• Add missing header in put_data_lifecycle rest-api-spec #116292

EQL

• Don’t use a BytesStreamOutput to copy keys in BytesRefBlockHash #114819 (issue: #114599)

ES|QL

• Added stricter range type checks and runtime warnings for ENRICH #115091 (issues: #107357, #116799)
• Don’t return TEXT type for functions that take TEXT #114334 (issues: #111537, #114333)
• ESQL: Fix sorts containing _source #116980 (issue: #116659)
• ESQL: fix the column position in errors #117153
• ES|QL: Fix stats by constant expression #114899
• Fix NPE in EnrichLookupService on mixed clusters with <8.14 versions #116583 (issues: #116529, #116544)
• Fix TDigestState.read CB leaks #114303 (issue: #114194)
• Fixing remote ENRICH by pushing the Enrich inside FragmentExec #114665 (issue: #105095)
• Use SearchStats instead of field.isAggregatable in data node planning #115744 (issue: #115737)
• [ESQL] Fix Binary Comparisons on Date Nanos #116346
• [ES|QL] To_DatePeriod and To_TimeDuration return better error messages on union_type fields #114934

Infra/CLI

• Fix NPE on plugin sync #115640 (issue: #114818)

Infra/Metrics

• Make randomInstantBetween always return value in range [minInstant, maxInstant] #114177

Infra/REST API

• Fixed a NullPointerException in _capabilities API when the path parameter is null. #113413 (issue: #113413)

Infra/Settings

• Don’t allow secure settings in YML config (109115) #115779 (issue: #109115)

Ingest Node

• Add warning headers for ingest pipelines containing special characters #114837 (issue: #104411)
• Reducing error-level stack trace logging for normal events in GeoIpDownloader #114924

Logs

• Always check if index mode is logsdb #116922
• Prohibit changes to index mode, source, and sort settings during resize #115812

Machine Learning

• Fix bug in ML autoscaling when some node info is unavailable #116650
• Fix deberta tokenizer bug caused by bug in normalizer #117189
• Hides hugging_face_elser service from the GET _inference/_services API #116664 (issue: #116644)
• Mitigate IOSession timeouts #115414 (issues: #114385, #114327, #114105, #114232)
• Propagate scoring function through random sampler #116957 (issue: #110134)
• Update Deberta tokenizer #116358
• Wait for up to 2 seconds for yellow status before starting search #115938 (issues: #107777, #105955, #107815, #112191)

Mapping

• Change synthetic source logic for constant_keyword #117182 (issue: #117083)
• Ignore conflicting fields during dynamic mapping update #114227 (issue: #114228)

Network

• Use underlying ByteBuf refCount for ReleasableBytesReference #116211

Ranking

• Propagating nested inner_hits to the parent compound retriever #116408 (issue: #116397)

Relevance

• Fix handling of bulk requests with semantic text fields and delete ops #116942

Search

• Catch and handle disconnect exceptions in search #115836
• Fields caps does not honour ignore_unavailable #116021 (issue: #107767)
• Fix handling of time exceeded exception in fetch phase #116676
• Fix leak in DfsQueryPhase and introduce search disconnect stress test #116060 (issue: #115056)
• Inconsistency in the _analyzer api when the index is not included #115930
• Semantic text simple partial update #116478
• Updated Date Range to Follow Documentation When Assuming Missing Values #112258 (issue: #111484)
• Validate missing shards after the coordinator rewrite #116382
• _validate does not honour ignore_unavailable #116656 (issue: #116594)

Snapshot/Restore

• Retry throttled snapshot deletions #113237

Vector Search

• Update Semantic Query To Handle Zero Size Responses #116277 (issue: #116083)

Watcher

• Watch Next Run Interval Resets On Shard Move or Node Restart #115102 (issue: #111433)

Ingest Node

• Fix _type deprecation on simulate pipeline API #116259

Machine Learning

• [Inference API] Deprecate elser service #113216

Mapping

• Deprecate _source.mode in mappings #116689