What's new in the YugabyteDB v2026.1 STS release series

Improvements

YSQL

• Enables transactional DDL in YSQL, allowing DDLs to run within a transaction block and interleave with DML operations, while ensuring DDL atomicity. #3109, #27616
• Adds authentication timeouts to enhance security and reliability. #27709
• Enables transactional DDL by default in release builds, allowing mixed DDL and DML operations within transactions. #3109
• Deprecated the flag ysql_conn_mgr_max_phy_conn_percent and introduced new flag ysql_conn_mgr_reserve_internal_conns for reserving connections for internal operations. #28859
• Prevents sending of YBParsePrepareErrorResponse packet for unnamed prepared statements. #29526
• Allows ALTER TABLE ALTER COLUMN SET STORAGE without errors and logs a warning for YB tables. #29547
• Ensures connection stickiness when using pg_replication_origin_session_setup. #29842
• Fixes a memory leak in YSQL authentication pass-through. #29882
• Reduces unnecessary index rebuilds during ALTER TYPE operations that don't require a table rewrite. #24007
• Enables the use of ALTER FUNCTION in YSQL for function definition changes. #2717
• Enables ysql_conn_mgr_pool_timeout for control connection pools. #30315
• Enables XML data type and associated functions in YSQL. #1043
• Deprecates old control connection password flag and enhances security by using environment variables for user and database names. #30825
• Adds metric to track server-side connection establishment time. #28893
• Disables yb_enable_upsert_mode with secondary indexes to prevent data inconsistencies. #27842
• Eliminates unnecessary disk I/O by not writing the relcache init file when catalog preloading is enabled. #31837
• Allows configuring TCP backlog size for Connection Manager with ysql_conn_mgr_socket_listen_backlog. #31824
• Stops index backfill if the CREATE INDEX process is killed. #26918
• Allows configuring the TTL for idle control connections, enhancing cleanup. #31959
• Stops retrying DDL read restarts in the main PostgreSQL loop. #29701
• Adds OpenTelemetry tracing for YSQL query lifecycle stages. #30598
• Limits prep statement accumulation on backends with ysql_conn_mgr_max_prepared_statements. #30433
• Adds distributed tracing for PostgreSQL's extended query protocol. #30672
• Automatically creates indexes on id columns for Apache AGE labels, enhancing query performance. #31132
• Automatically creates tracing spans for each RPC to enhance debugging and monitoring. #30599
• Enhances GUC setting accuracy by forwarding the startup packet to the auth backend. #27723
• Separates default and session-based YSQL configuration parameters into different lists for better management. #27725, #27827
• Reduces catalog cache memory spikes during new connections by introducing a TriggerRelcacheInitConnection RPC that allows only superuser postgres to rebuild the relcache init file. #28758
• Enables automatic catalog version increment for all DDL statements by default. #28253
• Limits relcache init file rebuilding to internal connections only, reducing the memory spike associated with new connections. Adds debug logs to help diagnose errors when rebuilding the relcache init file. #28758
• Allows concurrent DDL operations in YSQL when object locking is enabled, enhancing performance and usability. #27037
• Enhances authentication passthrough, ensuring YSQL connection correctness and reducing latency. #29112
• Automatically enables pg_dist_rag when starting TServer with enable_dist_rag_service=true. #29954
• Introduces backoff retry logic for failed auto analyze attempts. #29437
• Allows clients to wait for servers with the needed version to become available. #30233
• Enables CPU limits at the database level by assigning backend processes to specific cgroups. #30165
• Adds support for Apache AGE graph extension to enable graph database capabilities. #30056
• Enhances protocol tracking for custom parse packets in YugabyteDB. #30848
• Assigns YSQL Connection Manager (Odyssey) to a specified cgroup to ensure CPU isolation. #31183
• Adds a background task to retrigger failed DDL verification tasks automatically. #31430
• Adds OpenTelemetry tracing for SQL queries executed via the SPI path. #30945
• pg_total_relation_size now includes index size for accurate table size reporting. #23181
• Automatically single-quotes YSQL configuration parameter values with spaces to prevent crashes. #29808
• Boosts efficiency by only reading necessary columns during concurrent index builds. #29906
• Enables selective enabling of the MAGE extension via the yb_enable_mage flag. #31358
• Updates the Python driver to support the renamed mage extension. #31361
• Adds a new YSQL function yb_stat_auto_analyze to monitor Auto Analyze stats. #15667
• Re-enables parallel append by default in YSQL. #28064
• Enable or disable Query Plan Management to track YSQL statement performance using yb_pg_stat_plans_track. #28160
• Automatically adds SAOP conditions for certain index columns to optimize merge sorts. #29076
• Adds support for LISTEN, NOTIFY, and UNLISTEN commands in YB. #1872
• Allows custom sort orders in index scans through SAOP merge planning. #29070
• Automatically adds equality conditions for generated columns and indexes based on dependency. #29075
• Enables complex index scans with custom sort orders using SAOP merge execution. #29072
• Expands EXPLAIN (ANALYZE, DIST) output to include transaction type, enabling a more formal way to deduce transaction type. #14804
• Adds variable bloom filter support for queries with multiple keys, enhancing efficiency during scan operations. #28439
• Adds support for docdb_rows_returned column in pg_stat_statements to count rows returned by the storage layer. #28505
• Removes TablespaceMap for simpler and direct data access in YSQL functions. #29028
• Allows pushdown of stable and parameterless SQL Value Functions as storage filters. #28997
• Adds a new test for backing up and restoring roles. #28689
• Adds a new ysql_userid column to Active Session History to track user activity and query performance. #28565
• Expands pg_stat_statements metrics, enabling detailed query analysis with a specific flag. #29194
• Removes redundant checks in PgClientSession to streamline code execution. #28735
• Cleans up YbBindSearchArray, enhancing code efficiency and preparing for future bucketized index work. #28834
• Refactors hash permutations in PgGate to enhance memory management and support more complex query conditions. #28769
• Removes the unused pg_client_service_util.h file. #28929
• Enables Read Committed isolation by default in release builds. #29124
• Enables Early Access to Query Diagnostics from 2025.2. #28959
• Reduces noise in logs for non-geo-partitioned clusters by changing tablespace logs to VLOG. #28979
• Increases timeout for TServer catalog version wait to 30 seconds, reducing connection time outs. #28978
• Allows preloading of RANGEMULTIRANGE to reduce cache misses when pg_range is preloaded. #28991
• Removes the use of tablespace_map_ in object locking, using a more robust LRU-based cache. #29028
• Refined type conversions in index backfill code for better consistency and safety. #29038
• Auto analyze flag now defaults based on user-set cost-based optimizer (CBO) flags, reducing bad combinations. #28944
• Reduces peak memory usage in yb_active_session_history by filtering out rows more quickly. #29089
• Speeds up yb_active_session_history queries by allowing time-range filtering. #28810
• Adds four transaction-specific wait events to Active Session History for better operation tracking. #28792
• Ensures yb_local_tablets displays current schema names after a rename. #28688
• Changes explain plan format in query diagnostics to JSON. #29177
• Enhances SAOP merge execution by using the optimal SAOP in scenarios with varied cardinalities. #29071
• Reduces the need to adjust catalog versions when adding new YSQL tables. #29238
• Adds a check to prevent autonomous DDL transactions during an active regular transaction block. #29239
• Refactors hacky varno logic for clearer Explain output formatting. #29256
• Removes dead code by deleting yb_fdw.c and yb_fdw.h, enhancing maintainability. #29257
• Refactors PgClientSession to streamline read time updates via UpdateReadTime. #29310
• Enhances debugging by logging the last DDL node tag regardless of BEGIN block presence. #29402
• Automatically registers new error codes, preventing conflicts and serialization issues. #29482
• Expands yb_explain_hide_non_deterministic_fields to cover additional EXPLAIN fields. #29551
• Simplifies code by removing redundant PgDocOp::active_op_count_. #29584
• Reduces code complexity and repeated operations in YSQL functions. #29584
• Consolidates secondary index scan logic and streamlines class names for future enhancements. #29656
• Enables RPC stats collection in pg_stat_statements by default. #29749
• Adds a deprecated YSQL configuration for toggling PG snapshot management. #29763
• Reduces PgApiImpl initialization to a single method call. #29768
• Adds support for replication connection authentication in Auth Passthrough mode. #29113
• Streamlines YSQL code by organizing ResultStream classes into a separate module. #29824
• Enables ysql_yb_enable_ddl_savepoint_support as a preview feature, ensuring stability for user trials in pre-production environments. #29825
• Extends the lifetime of fetched data in DocDB for more efficient query processing. #29702
• Replaces ReplaceAll in libpq- with boost::algorithm::replace_all. #29887
• Replaces outdated JsonReader with versatile JsonDocument for enhanced JSON handling. #29944
• Reduces boilerplate code in the PgSession class for streamlined operation. #29932
• Reduces redundant checks for recheck logic in YB index access methods. #19364
• Adds direct JsonDocument support in FetchRow/FetchRows for seamless JSON handling. #30063
• Stabilizes InvalMessageSanityTest by using postgres database. #29983
• Enables creating servers of type federatedYugabyteDB. #29988
• Enables fetching the TServer UUID directly via a new SQL function. #29990
• Enables creation of internal objects for LISTEN/NOTIFY in yb_system database, facilitating system-wide notifications. #1872
• Adds tracking for duplicate server key hash collisions in the connection manager. #30023
• Enables full querying of global views on YugabyteDB using the new flag -remote_pg_query_execution_rpc_timeout_ms. #30279
• Adds an RPC for fetching info of yb_system database tables across all DBs. #1872
• Reduces unnecessary rechecks for hash code keys in scans. #30064
• Adds transaction IDs to CREATE DATABASE logs for easier debugging. #30083
• Enhances handling of CLOSE packets in YSQL to fix sequencing and resource issues. #29574
• Enhances performance by reducing unnecessary PostgreSQL rechecks. #30097
• Allows normal users to set yb_max_saop_merge_streams for bucket-based indexes. #30112
• Refactors replication code to enhance LISTEN/NOTIFY, improving error handling and function availability. #1872
• Enables index scans to return custom-sorted results using a new SAOP merge feature. #29072
• Decouples YbctidReader and DocResultStream for better resource management. #30705
• Identifies the specific tablet housing a row based on its primary key. #30196
• Enables handling of ALTER ROLE commands for consistent YSQL configuration management. #30113
• Renames key identifiers for clarity in YugabyteDB's codebase. #30626
• Simplifies YSQL configuration by replacing three old parameters with one new test parameter. #30586
• Streamlines YSQL index backfill processes and reduces redundancy in code execution. #30288
• Reduces clutter in pg_stat_statements by aggregating BACKFILL INDEX entries under a single query identifier. #30203
• Enhances error hints for scenarios where databases are dropped and recreated. #30319
• Simplifies the PGConn interface by introducing CopyFromStdin for easier COPY FROM STDIN operations. #30240
• Enhances function signatures by adjusting the placement of OUT arguments. #30386
• Enables secondary indexes to return primary keys directly, bypassing main table reads. #3574
• Renames yb_fallback_to_legacy_catalog_read_time to yb_enable_concurrent_ddl for clarity. #30419
• Ensures configuration updates are applied to new database sessions after a SIGHUP signal. #30435
• Adds plan_id to yb_active_session_history to help diagnose query latency issues. #30431
• Enables distributed tracing in YSQL using OpenTelemetry SDK. #30596
• Automatically cleans up LISTEN state when a backend crashes to prevent issues and infinite loops. #1872
• Fixes bugs in LISTEN/NOTIFY operations and ensures correct data handling. #1872
• Reduces RPCs during authentication by enabling prefetching for system tables. #29952, #30409
• Enables selective DEALLOCATE for prepared statements using connection manager. #29707
• Enables runtime updates to YSQL Connection Manager flags for enhanced configuration management. #30537
• Simplifies handling of SKIP LOCKED in YBCLockTuple function. #30600
• Ensures object lock release is synchronous with transaction completion to prevent stale schema issues. #30611
• Moves TableScanDesc out of YbScanDesc to enhance table scan handling. #30664
• Eliminates read restart errors for catalog reads by clamping the uncertainty window. #30627
• Enables read restarts for DMLs on catalog queries, reducing errors. #30552
• Surfaces errors when initializing LISTEN on a YugabyteDB node, ensuring clients receive error messages and clean transaction aborts. #1872
• Enhances stability by relocating catalog snapshot switching logic to a more suitable function. #30668
• Collects YSQL statistics for enhanced telemetry in YugabyteDB. #30781
• Enhances control of PgSession over ExplicitRowLockBuffer for varied flushing scenarios. #30685
• Enables ysql_yb_enable_listen_notify flag as a runtime configurable option. #1872
• Makes ConnectionBuilder compatible with MiniYugabytedCluster in Java tests. #30692
• Introduces flags to adjust notifications polling frequency and enhances signal handling. #1872
• Supports extracting traceparent from SQL comments or using the yb_dist_tracecontext parameter. #30597
• Simplifies memory context management in YSQL by generalizing helper methods. #30874
• Eliminates the storage of query pointers in PgGlobalViewRead, enhancing safety and efficiency. #30860
• Removes ASH metadata from PgQueryAutoAnalyzeRequestPB for cleaner RPCs. #30742
• Removes unnecessary coupling and an obsolete field from YbctidReader and DocResultStream. #30705
• Adds detailed connection logging for successful YSQL authentications. #29817
• Removes unnecessary GUC name lowercasing and options parsing in Auth Passthrough mode. #29178
• Replaces the explainrun framework with a more flexible parameterized query system. #30901
• Adds yb_system database classification as system objects. #1872
• Renames misleading "primary" identifiers to "key" in internal code for clarity. #30937
• Enables faster query processing by preloading catalog lists, reducing master lookups. #30868
• Adds version gate to yb_backup.py for with-statistics option compatibility. #30857
• Enhances row lock management by flushing explicit locks in multiple YSQL operations. #30685
• Ensures connections stay sticky when using LISTEN with the connection manager to prevent missed notifications. #1872
• Optimizes debug logging performance by conditionally building strings. #29338
• Ensures backfill queries in pg_stat_statements retain actual read times for clear indexing history. #31121
• Terminates the slowest listener to avoid blocking notifications in a full async queue. #1872
• Enables yb_db_admin to set backtrace_functions for improved debugging. #31217
• Allows users to load the renamed Apache AGE extension as mage in YugabyteDB. #31299
• Enables per-table auto-analyze settings for more tailored database optimization. #30740
• Enhances new connection reliability by fixing relcache revalidation issues. #31309
• Reverts client encoding source to PGC_S_DYNAMIC_DEFAULT, ensuring proper encoding settings on authentication. #30148
• Ensures GUC updates are visible on newly authenticated clients by correctly updating control backend LCVs on SIGHUP. #31357
• Prioritizes traceparent YSQL configuration over SQL comments for distributed tracing. #31112
• Enables tracing for SQL queries with a traceparent comment at the end. #30946
• Enables distributed tracing for shared memory communications between PostgreSQL and TServer. #31324
• Prevents sending incorrect superuser status and ensures correct default settings in Auth Passthrough mode. #30148
• Resets the readahead buffer properly when it fills up to prevent connection issues. #30148
• Enhances forwarding of NoticeResponse packets during backend startup in YSQL. #30148
• Removes redundant ybhnsw.ef_search GUC, simplifying configurations using standard hnsw.ef_search. #31676
• Terminates the slowest listener to prevent queue overflow and ensure timely notifications. #31511
• Enables predicate pushdown during index backfill, optimizing data scanning. #30263
• Enables batched DocDB lookups for mismatched foreign key data types. #27007
• Disables a foreign key optimization by default for type mismatches. #27007
• Reduces flushes for NOTIFY operations by grouping INSERTs and DELETEs. #31568
• Increases the default limit for DDL invalidation messages to 8192 to reduce costly full catalog refreshes. #31797
• Drops replication slots on tserver restarts or decommissioning to unblock WAL cleanup. #31485
• Enables ALTER ROLE and ALTER DATABASE settings by default for pooled connections. #31854
• Removes support for shared catalog version mode, deprecates ysql_enable_db_catalog_version_mode flag. #28937
• Reduces test runtime for PG wait-events by 70% by combining tests into one cluster. #31955
• Disables parallel query for CTAS and REFRESH MATERIALIZED VIEW by default. #31670
• Increases default yb_parallel_range_size to 16MB for better query efficiency. #32006
• Enables automatic restoration of table statistics with ysql_dump. #28713
• Reduces YSQL backend startup logging to only two useful lines. #29579
• Expands extension development by including necessary YugabyteDB headers in the installation directory. #11491
• Enables Batched Nested Loop joins on expression-based indexes. #17175, #28942
• Enables distinguishing control backends during startup through a new packet field. #31418
• Enables BNL joins by shifting unbatchable conditions to join filters. #31760

YCQL

• Enables clearing the YCQL MetaData cache via the yb-ts-cli tool. #7832
• Enhances yb-ctl status info formatting by using a dictionary. #29216
• Enables JWT-based authentication for YCQL using external IDPs. #29860
• Enables regular expression based identity mapping for YCQL JWT authentication. #29861
• Enables ycql_use_jwt_auth as a preview flag to enhance JWT authentication testing in YCQL. #30276
• Renames ycql_jwt_options to ycql_jwt_conf for familiar configuration naming. #30276
• Handles double quotes in ycql_jwt_conf to prevent server startup failures. #30557
• Enables production use of ycql_use_jwt_auth for OIDC authentication. #31512

DocDB

• Changes default value of ysql_num_tablets to 1 for initial number of tablets in hash partitioned tables. #28060
• Optimizes index scans by adaptively deciding to use next or re-seek, reducing query latency. #28616
• Adds logging to detect and inform if scheduled tasks in the reactor run slower than expected. #24801
• Enables detailed data and checksum dumps for tablets via new commands, aiding in consistency checks and troubleshooting. #29904
• Reworked the load balancer to not hold CatalogManager::mutex_ for each placement run, thus improving efficiency. #29183
• Enables log-dump to read encrypted WAL files using the master_addresses flag. #17816
• Adds the ysql_clone_disable_connections flag to allow YSQL database connections during cloning. #29678
• Adds "DETAIL" prefix to Master and TServer log lines for clarity. #29703
• Automatically scales up transaction status tablets as clusters grow. #29555
• Renames a variable to better reflect that it stores the master leader's cloud info, not the TServer's. #20231
• Enhances the log-dump tool with packed rows support and improves error handling. #29608
• Adds detailed logging for SST file metrics and creation events. #20880
• Adds background checks for TServer connectivity and shares results via RPC. #30039
• Suppresses error logs for missing files on deletion to streamline operations. #29757
• Reduces log frequency to once per second for tablets stuck in bootstrapping. #24566
• Enhances sst_dump to correctly handle and display RocksDB-level tombstone records. #19707
• Defers promotion of peers from PRE_VOTER to VOTER until local bootstrap completes. #29795
• Enhances the log-dump tool output and usage message, adds transaction ID field, and introduces a new flag no_pretty_hybrid_times. #29733
• Logs tserver start time and PID for easier troubleshooting. #14805
• Enhances logging for middle key selection in tablet splitting, aiding in debug and analysis. #27336
• Enables stepping down to a preferred leader zone when removing a replica. #19884
• Tracks and alerts on more RPC rejections due to memory pressure. #20670
• Enables system catalog flush and compaction across all YB-Master peers. #10948
• Adds a tablet_split_candidates metric for tracking unsplitted tablet counts. #28958
• Adds logging for large memory allocations with stack traces. #12500
• Adds retries to FetchData RPC to handle intermittent network issues during remote bootstrapping. #26217
• Enables tracking of memory used by compactions in RocksDB. #26845
• Enhances advisory lock table scalability by including all columns in the hash calculation. #30660
• Ensures xCluster resumes from checkpoints more reliably by reusing known OpIds and reducing LookupOpId calls. #30769
• Increases default tablet split and compaction thread limits for better performance. #29026
• Logs now detail the reasons behind each RocksDB flush for better traceability. #25615
• Makes placement_uuid optional in read replica tablespace creation, auto-filling it from cluster config if omitted. #31192
• Captures origin_id from PostgreSQL sessions, enhancing CDCSDK handling of WAL records for distributed transactions. #29582
• Prevents xCluster replication issues by ensuring new schema versions are current. #29760
• Adds control for vector index inclusion in post-split compaction. #29378
• Adds a flag to control automatic tablet splitting for tables with vector indexes. #29380
• Limits TServer crashes by capping log reader memory usage in xCluster replication. #28124, #22992
• Checks tablet limits before starting the DB cloning process. #22338
• Enables safe activation and deactivation of the skip prefix locks feature and allows for safe upgrades. #28823
• Allows control over the number of tasks running in parallel via the introduction of PriorityThreadPoolToken. #29151
• Control the number of vector index compactions per tserver with new flags, offering control over chunks taken for compaction and limiting running compactions per TServer. #27601
• Promotes object locking preview flag to a regular flag, allowing users to control the feature through the enable_object_locking_for_table_locks flag. #29317
• Marks follower_unavailable_considered_failed_sec as a runtime flag. #30123
• Skips replication of PUBLICATION and SUBSCRIPTION DDLs in xCluster mode. #29340
• Fixes lock order inversion during table creation failures. #30088
• Supports session-level object locks extending lock lifespan across transaction boundaries. #27120
• Filters high-frequency logs to a new DETAIL_LEVEL without changing default log visibility. #30307
• Adds infrastructure for managing cgroups in YugabyteDB. #30441
• Enables per-database CPU resource management on tserver with new flags. #30099, #30162
• Preserves pg stats when cloning databases, reducing manual steps. #30257
• Enables per-database CPU limits by using dedicated thread pools for YSQL requests. #30163
• Allows using multiple thread pools to manage CPU limits more efficiently. #30752
• Ensures DDLs committed together on the source are applied atomically on the target. #29127
• Enables isolated CPU usage for each database using a cgroup hierarchy. #30510
• Enables monitoring of CPU usage and throttling metrics per cgroup for quality of service. #30125
• Simplifies QoS cgroup hierarchy and enhances metric visibility for TServer threads. #31183
• Supports early detection of unsupported table types in cross-cluster replication setup, enhancing error messages. #14906
• Reduces spammy logs in CDC service, TX manager, and RocksDB operations. #29581
• Increases yb-tserver file handling capacity on macOS by using raw POSIX file operations. #30601
• Configure RocksDB keys delta-encoding separately for YCQL and YSQL for faster read path and space efficiency. #28862
• Displays xCluster safe time on yb-master UI page for enhanced debugging. #28910
• Introduces two YSQL configuration parameters to control use of tablespace-local locality, offering workaround for suboptimal performance in certain cases. #28604
• Updates compaction logic to correctly handle and clean tombstoned reverse-mapping vector index keys. #27095
• Reduces CPU overhead and memory allocation for filtered queries by eliminating unnecessary data transformations. #28928
• Allows reads and writes to proceed without waiting for a full quorum acknowledgment, reducing YSQL write and read latencies. #28320
• Disables table locks to prevent potential test failures. #28696
• Allows tablespace-local transactions to write to tables in any contained tablespace without global promotion. #28272
• Adds logging for long shutdown delays to aid in debugging. #28796
• Enables direct encoding of QLValuePB to binary format for faster lookups. #28808
• Calculates region locality on the tserver side for improved region local transactions handling. #28828
• Introduces disk write statistics for vector compaction, which offers insights into IO across vector index libraries. #28842
• Enhances verbose logging for xCluster to track safe time issues. #28846
• Enables skip prefix locks by default for new installation clusters. #28894
• Removes unused DummyANN from DocDB, streamlining code maintenance. #28911
• Adds server-level metrics to track the number of transactions started as global, region-local, and tablespace-local. #28921
• Standardizes thread names for the /perf endpoint to improve flamegraph clarity. #28803
• Speeds up read and write requests by using lightweight protobufs. #29787
• Adds latency metrics for bloom filter operations, controllable via rocksdb_collect_bloom_filter_time_metrics flag. #28617
• Enables configuring read replicas at the tablespace level, enhancing data access flexibility. #12180
• Now captures TCMalloc and MemTrackers stats in logs when root memtracker soft memory limit is exceeded, aiding in memory issues investigation. #29050
• Refactored mutex acquisitions in larger clusters to mitigate contention issues. #29186
• Enables vector index background compactions by default to control parallel execution and limit compactions per tserver. #29212
• Ensures more balanced tablet splits by calculating the middle key from user data only. #29214
• Blocks nextval and setval functions in automatic mode target databases for xCluster replication. #29248
• Reenables LogReader memory limit in xCluster following resolution of TryConsume bug. #29252
• Enhances vector index checkpoints placement for tablet splitting. #29262
• Enhances block navigation accuracy and API access in TwoLevelIteratorState. #29281
• Supports lightweight protobufs for enhanced performance in read/write requests. #29295
• Enhances debugging and fixes a bug in RWCLock logging. #29309, #29364
• Enables expression evaluation using lightweight protobufs. #29371
• Added tablet replica SST and WAL size to master UI and sorted tablets by partition key. #29407
• Added new /snapshots endpoint to tserver UI displaying space overhead information of snapshots, including snapshot ID, time, cumulative required space, exclusive data space, and schedule ID. #29388
• Expands test APIs to support N-way tablet splitting with vector of split keys. #29469
• Enhances tablet splitting APIs to support a list of child IDs and keys, still limited to 2-way splits. #29470
• Automatically deletes vector index directories when tablets are deleted. #29447
• Reduces unnecessary transaction aborts during DDL operations with table locks enabled. #27613
• Prevents false deadlocks by not reusing transactions that have blocked others. #28042
• Enables manual tablet splitting for vector indexes. #29377
• Ensures middle key computation for splits correctly checks SST file count and size. #29627
• Enables multi-way tablet splitting for more flexible and precise data distribution. #29471
• Eliminates redundant logic for child tablet registration, enhancing system consistency. #29643
• Automatically appends throttle messages to YB_LOG_EVERY_N* log outputs. #29661
• Facilitates migration to lightweight protobufs using intermediate typedefs. #29705
• Renames kNumSplitParts to kDefaultNumSplitParts for clarity in split parts count. #29713
• Logs now show the distance function used by the vector index. #29804
• Enables testing N-way tablet splits by updating key request and response protocols. #29815
• Allows specifying the number of split children in MasterAdmin::SplitTablet RPC. #29846
• Allows specifying a split_factor in yb-admin split_tablet for custom tablet splits. #29734
• Logs ysql_dump output on script failure during DB cloning. #29607
• Adds detailed metrics for vector index operations in DocDB. #30004
• Enhances cluster management by using local consensus state instead of ListMasters RPC. #29975
• Enhances xCluster DDL replication testing for various table rewrite scenarios. #29996
• Sets yb_hnsw as the default backend for vector indexes. #29964
• Displays TServer connectivity states in the Web UI. #30179
• Enables Object Lock Manager to ignore conflicts with background transactions. #30186
• Stores both original and pause errors in xCluster DDL for easier debugging. #30156
• Reduces PostgreSQL's dependency on RocksDB by restructuring libraries. #30226
• Enables ScaNN support by updating third-party libraries. #30238
• Reduces header bloat in catalog_manager.h by cleaning up unnecessary includes. #30304
• Allows using double flags with flag comparison validators. #30440
• Sets namespace_id in tablet metadata for clone requests. #30548
• Delays YSQL Connection Manager startup until lease acquisition. #29495
• Consolidates separate block caches into one to prevent memory overrun. #30604
• Enables block-based vector index conversion for hnswlib backends. #30720
• Adds a metric to track ongoing tablet splits in the cluster. #30648
• Adds support for EXPLAIN ANALYZE DEBUG in vector index queries, providing detailed metrics. #26550
• Adds column sorting in the Master/TServer UI /threadz table. #30695
• Displays cgroup information in the /threadz endpoint for better debugging. #30696
• Simplifies shared memory object locking by removing the finalized field. #30755
• Removes EnvMirror to prevent potential memory leaks. #30744
• Removes outdated code for the UpdateTransactionStatusLocation RPC. #30584
• Streamlines xCluster failover by making the API asynchronous and adding status persistence in SysCatalog. #30564
• Increases the default value of FLAGS_num_advisory_locks_tablets to 3. #30823
• Enhances xCluster UI with clickable links and color coding for easier readability. #30938
• Sets the follower_unavailable_considered_failed_sec flag higher than leader intervals to ensure peer stability. #30969
• Switches atomic operations in metrics to std::atomic for enhanced performance on ARM. #30974
• Supports LOG(DETAIL) and similar macros for detailed logging. #31141
• Adds CPU profiling access via a new icon on master and tserver utilities pages. #31152
• Simplifies TaggedThreadPools and fixes ThreadPoolTest failures. #31198
• Enables TabletPeer to use per-db servicer pools improving write operations. #31213
• Ensures clones aren't made to a point before a YSQL upgrade for consistency. #28525
• Enables database count limits when QoS is active using qos_max_db_count flag. #31205
• Adds a /cgroups endpoint to the tserver UI to monitor cgroups. #30732
• Enables database cloning by default in new YugabyteDB universes. #30838
• Deprecates yb_silence_advisory_locks_not_supported_error to prevent inconsistent behaviors. #31385
• Allows yb-ts-cli compaction commands to compact vector indexes by default. #31350
• Displays thread names and detailed CPU statistics for leaf cgroups on the /cgroups endpoint. #31392, #31572
• Ensures the database_name label is always included in cgroup metrics. #31664
• Switches the default vector index backend to hnsw_lib for enhanced performance. #31757
• Ensures pipelined async writes continue smoothly during leader changes via TabletInvoker retries. #31402
• Displays YSQL operation lease status on the tablet servers' HTML status page. #25707
• Enables building with GCC 15 and introduces customizable toolchain options. #31593
• Displays distinct PostgreSQL process names in the /cgroups endpoint. #31979
• Allows yb-admin get_table_hash to hash individual colocated tables accurately. #31952
• Enhances yb-admin get_table_hash with key range parameters to focus on specific data segments. #31951
• Speeds up CDCSDK tests by using the default yugabyte namespace instead of creating a new one. #31965
• Enables error messages to display tablet and table names for Snapshot too old errors. #28841
• Enables creating performance flamegraphs via a new endpoint /perf. #28355
• Renames "load balancer" to "cluster balancer" for clarity and accuracy. #29554
• Enables packaging of profiling tools and running perf during unit tests. #29866
• Eliminates the need for continuous snapshot schedules during xCluster failover. #30389
• Streamlines vector index configuration with a single new flag vector_index_backend. #30379
• Adds a new unit test to ensure database cloning preserves xCluster DDL replication tables. #27930
• Fixes index creation inconsistency by replacing catalog lease with more reliable YSQL lease. #27863
• Updates DocumentDB extension to version 0.109-0 and merges latest changes. #30863
• Enables building on macOS with llvm-installer clang using ./yb_build.sh set to clang21. #30035
• Enables consistent use of HybridTime value types as compile-time constants. #32020
• Enhances yb-admin list_tablets to display partition keys in clearer formats, matching the :9000/tablets UI. #28444

CDC

• Reverts a commit due to a misunderstanding about the effects of failures when writing to the CDC state table. #28014
• Automatically detects and adapts to DDL table rewrites in CDC streaming. #29675
• Limits the number of tablets polled in GetConsistentChanges calls. #29900
• Ensures safe upgrades for CDC with DDLs by checking cdc_enable_dynamic_schema_changes. #30355
• Allows dynamic control of GetChanges response size in yb-client. #30780
• Removes dependency on deprecated field for schema name in CDC records, fetching directly from pg_namespace. #28655
• Supports streaming CDC records for specific table rewrite DDL operations. #15576
• Allows CDC streaming from tables without a primary key using a new flag. #29423
• Enables ysql_enable_pg_export_snapshot by default and exposes it as a YSQL configuration parameter. #27253
• Added runtime flag to control CDC with Savepoints and Rollback feature, and introduced tests for rollback and savepoint behavior. #29773
• Switches CDC snapshot batching to byte-based threshold for better memory management. #29850
• Corrects test to ensure hidden tablets are deleted post-expiry, not just due to restart time adjustments. #15576
• Reduces unnecessary iterations over the cdc_state table by removing reliance on DELETING_METADATA state for stream cleanup. #29360
• Throws an error when pg_current_wal_lsn is called, indicating unsupported feature. #30243
• Reduces unnecessary master load by not populating backend_xmin in pg_stat_get_activity. #29747
• Removes the FLAGS_yb_enable_cdc_consistent_snapshot_streams flag and sets consistent snapshot streams as the default for tests. #23000
• Enables safe upgrades for the logical replication of dynamically added tables. #27686
• Enhances system performance by preventing unnecessary RPCs to the yb-master in pg_stat_get_activity. #29747
• Ensures UPDATE/DELETE operations work with row filters and default replica identity. #22490
• Ensures UPDATE and DELETE with row filters work for tables using REPLICA IDENTITY CHANGE. #22526
• Reduces tserver log spew during YSQL logical replication. #30630
• Adjusts CDC retention barriers for sys_catalog based on record consumption, ensuring data integrity. #26427
• Enables streaming of tables without primary keys in logical replication by default. #29423
• Enhances CDC tests to handle the implicit dynamic table addition mechanism effectively. #30298
• Supports capturing before images for intra-transactional DML operations. #29209
• Enables local RPC calls in Virtual WAL if cdc_enable_local_rpc_in_virtual_wal flag is enabled. #20946
• Reduced logging levels in CatalogManager::FindXReplStreamsMarkedForDeletion when a table is dropped. #29637

yugabyted

• Activates read committed isolation and cost-based optimization by default for newly started clusters. #29169
• Enhances compatibility with Python 3.12 by updating CQLSH in third-party libraries. #30154
• Reduces namespace pollution and enhances compilation safety by replacing SpinLockHolder macro with [[nodiscard]] attribute. #31629
• Eliminates build errors due to overlapping macros between gutil and Abseil. #31673
• yugabyted now stops child processes using SIGTERM to allow cleanup. #29676

Other

• Upgrades Node.js to 22.18.0 and transitions from snowpack to vite. #29504, #29505
• Applies CPU limits at the database level by associating PgClientSession threads with the correct database cgroup when enable_qos is true. #30511
• Enables the SCAN Redis command for enhanced search capabilities, safely handles edge cases with empty memtables, and ensures proper error checking during tablet peer lookups to prevent fatal restart errors. #2

Bug fixes

YSQL

• Ensures Perform requests in YSQL are processed in the order they are issued to avoid errors. #23648
• Fixed the Connection Manager's support for replication connections, ensuring that it independently manages these connections, aligning with PostgreSQL's behavior. #28521
• Eliminates stale hashmap entries on parse errors, enhancing protocol-level prepared statement handling. #28576
• Allows users to preload pg_enum in ysql_catalog_preload_additional_table_list. #28757
• Allows yb-admin ysql_catalog_version command to return per-DB Catalog Version value for a provided database. #28690
• Prevents YSQL major upgrade failure caused by inconsistent namespace mapping during a failed upgrade attempt. #28815
• Disallowed Batched Nested Loop joins when the inner relation is a temporary table to prevent cache lookup errors or assertion failures. #28864
• Implemented statement timeouts and interrupt checking in pggate, providing an upper bound for query or operation time and terminating if deadline is reached. Deprecated the unused gflag pggate_rpc_timeout_secs. #28863
• Enhanced handling of empty strings in CatalogEntityPB to prevent parsing failures during table entry deletion. #29240
• Fixed issue where server connection gets stuck due to full server write buffer after ungraceful client disconnects. #29301
• Implemented a fix to remove stale entries from the server's hashmap during connection manager reset phase. #29527
• Fixed an issue where integer overflow in the new cost model could lead to negative cost estimates, resulting in poor plan choices. #29685
• Adjusted CatalogManager::GetYsqlCatalogConfig to use catalog_version_table_in_perdb_mode_ over FLAGS_ysql_enable_db_catalog_version_mode for accurate per-db mode representation. #28690
• Reduces CPU usage in connection manager by improving epoll_* functions and fixing syntax errors. #29827
• Fixes quoting for unique constraints during schema restores. #29838
• Installing pg_stat_monitor now displays a beta warning. #29321
• Optimizes column fetching during scans by removing redundant work, enhancing performance. #29847
• Enhances upgrade prechecks to detect pgcrypto conflicts, preventing PG15 upgrade errors. #29267
• Removes the ysql_beta_feature_pg_stat_monitor flag for pg_stat_monitor beta warnings. #29321
• Fixes issues with placement wildcards and leader preferences in tablespaces. #29971
• Ensures accurate index updates in INSERT ... ON CONFLICT for partial indexes. #30104
• Sets default pg_upgrade directory to pg_upgrade_data and allows configuration via pg_upgrade_working_dir flag. #29172
• Ensures unique index backfills use consistent timestamps for write and read operations. #30331, #30312
• Enhances replication connections from connection managers by not populating catalog version. #30384
• Ensures fast path writes correctly observe recent distributed transactions. #30551
• Automatically cleans up socket directories after a failed YSQL upgrade rollback. #30420
• Eliminates redundant statistic updates during BACKFILL INDEX to improve performance. #30562, #30120
• Prevents data inconsistency during index backfill by disabling in-place updates. #30653
• Fixes memory leak when reading corrupted pg_stat_statements files. #30827
• Sets default yb_qtext_size_limit to unlimited, preventing NULL query texts. #30775
• Prevents server crash by validating portal existence during the extended query protocol. #29986
• Enables on-demand building of catcache lists for improved function resolution. #31043
• Clarifies logs related to YSQL DDL transaction verification for better user understanding. #27153
• Handles stale transaction retries more effectively. #30895
• Ensures connection routes are not destroyed during garbage collection if active servers exist. #31189
• Enhances accuracy of seek and next count estimations in secondary index scans. #25564
• Corrects inconsistencies in index scan cost calculations for better query performance. #27718, #30490
• Ensures consistent cleanup of parallel contexts after YSQL query failures. #30040
• Fixes a segmentation fault in stored procedures that perform ROLLBACK. #31199
• Reduces startup queries by caching colocation status on TServer. #29647
• Stops control connections from being marked as sticky, allowing reuse for multiple auth attempts. #31904
• Fixes CREATE INDEX hanging issue when enable_ysql_operation_lease is set to false. #31978
• Ensures clients don't hang indefinitely by signaling them when a server is removed. #31988
• Allows group and others read permission on initdb log files by hardcoding the permission. #28435
• Sets Read Committed isolation as disabled by default in release builds. #31686
• Fixed an issue where connection manager could crash during authentication due to a double free error. #29195
• Resolves Bind packet forwarding issue for unnamed prepared statements. #29935
• Fixes a potential segfault when handling auth errors in Auth Passthrough mode. #29477
• Anonymizes database names and adds callhome_ysql_interval_secs flag. #30921
• Disables table locking and transactional DDL in the 2026.1 release. #31079, #31088
• Fixes race conditions in concurrent parallel key range fetching. #31219
• Forces immediate flushing of row locks in YSQL, ensuring lock acquisition before operations. #26307
• Supports multiple analyze commits and concurrent DDL, ensuring low priority lock on pg_yb_catalog_version is retaken during subsequent transactions. #28731
• Fixes a bug that caused yb-master to crash when executing the CREATE INDEX command outside of a transaction block with savepoint for DDL support enabled via TEST_ysql_yb_enable_ddl_savepoint_support. #28955
• Forces loading of index expressions during index creation to prevent missed updates and potential inconsistencies. #28267
• Prevents Postgres crash by avoiding invalidation messages during catalog refresh when prefetcher starts. #28490
• Fixes the issue of transaction self-abort with rollback to savepoint, allowing users to rollback DDL operations without facing transaction expired/conflict error. #28956
• Disables query layer retry of ANALYZE command to prevent potential crashes caused by dirty states. #29025
• Automatically sets default values for dependent flags when the YSQL Cost-Based Optimizer (CBO) is enabled or disabled. #29081
• Fixed issue where unique index data was missing for colocated databases after restore, now preserving implicit tablegroup oid. #28781
• Eliminates an issue causing stuck rollback operations to sub-transactions with alter table. #28957
• Eliminated potential TCP deadlock in YSQL by changing to asynchronous write in od_frontend_remote_client. #29303
• Fixed an issue where running ALTER TYPE on a table with dependent indexes triggered double index creation, causing replication conflicts on xCluster targets. #27741
• Now revalidates stale relcache init files efficiently without full rebuilds if catalog version changes but no other changes. #29260
• Modified od_write function to prevent double free of mm_msg_t objects and enhanced machine_msg_free to handle NULL messages. #29398
• Fixes race condition in ROLLBACK TO SAVEPOINT handling for DDLs. #29414
• Updates process exit behavior on pggate heartbeat failure during initialization. #29242
• Fixed a bug in YugabyteDB's backup and restoration process involving the mismatch of colocation_id in colocated databases. The fix preserves the original colocation_id when implicit indexes are created during the ALTER TABLE statement, ensuring that the colocation ids are consistent after restoration. #29260
• Enhances stability during transaction rollbacks involving index and base table deletions. #29373
• Truncate user and database names exceeding 64 characters to prevent buffer overflow during startup. #29331
• Reduces memory usage when using legacy mode for catalog operations with a fixed pruning logic. #29612
• Ensures replaced trigger functions are used correctly in transactions. #29424
• Ensures that newly created or updated functions are correctly recognized and replace stale versions in YSQL. #29424
• Fixes handling of concurrent DDL and DML operations for accurate snapshot application. #29269
• Prevents null pointer access issues in PgClient after shutdown. #29803
• Maintains correct message order by buffering ParameterStatus updates. #29884
• Prevents crashes during transaction abort when an ALTER TYPE command fails with transactional DDL enabled. #29325
• Fixes memory leaks by disabling unused queue for untracked RPCs. #29955
• Avoids "relation already exists" error during concurrent CREATE INDEX. #29735
• Ensures CREATE INDEX CONCURRENTLY runs smoothly by removing outdated checks. #29987
• Reinstates missing values restoration in backups, adding logs for schema mismatches. #29960
• Eliminates backend crashes during DDL aborts when transaction metadata is not set. #29994
• Simplifies the PgSession::RunAsync method signature for easier future modifications. #26307
• Ensures stats for range data types are restored after backup in YugabyteDB. #30147
• Removes tuple locks to prevent deadlocks during concurrent ALTER TABLE. #30095
• Prevents PostgreSQL backend crash by addressing a dangling reference issue. #30013
• Eliminates a memory leak in YSQL's authentication process. #30275
• Enhances secondary index scans on colocated databases to support merge sort operations. #30221
• Fixes error handling for concurrent DDL transactions to prevent incorrect error messages. #29692
• Preserves table statistics during backup and restore, even when secondary indexes are involved. #30151
• Fixes inconsistency in CREATE INDEX handling with object locking enabled. #30260
• Prevents crashes during ROLLBACK with transactional DDL and savepoints. #30661, #30348
• Prevents premature detaching of logical connections during pipelined operations. #30515
• Ensures catalog reads use the latest snapshot after a read restart, preventing index inconsistencies. #30260
• Prevents server crashes during index creation with concurrent DDL enabled. #31153
• Fixes index inconsistency issues during parallel DML and non-concurrent index creation. #30955
• Ensures rollback to savepoint properly waits for table deletion to finish. #30924
• Reduces memory consumption of the YSQL Connection Manager by releasing unused memory back to the OS. #31246
• Reduces errors in updates by skipping unnecessary partition checks. #31475
• Switches from 32-bit to 64-bit hashing to minimize prepared statement collisions. #30654
• Enhances handling of in-flight Parse operations with a tracking queue. #30175
• Ensures consistent entry matching in response batches during schema version mismatches. #31204
• Ensures consistency during rollback to savepoint by correctly handling asynchronous DeleteTablet RPCs. #30721
• Corrects tracking of parameter values in QPM for maximum execution times. #30853
• Masks sensitive constants in QPM stored plan texts. #31049
• Redacts sensitive execution parameters in yb_pg_stat_plans by default. #31050
• Fixes crash during prepared statement closure by ensuring transactions are active. #31860
• Enhances query accuracy by ensuring hash permutations stay within target tablet bounds. #31051
• Prevents fatal errors during retries of transaction control commands in DDL operations. #30908
• Ensures consistent read times for batches of DML operations following DDLs in a transaction. #29283
• Resolves hangups in COPY operations by resuming client relay on receiving COPY_IN_RESPONSE. #32102
• Fixes connection issues when dropping collections and databases in DocumentDB extension. #27698
• Fixes an issue where the mage extension crashes during the first cypher call under UBSan. #31404
• Prevents crashes when terminating a DDL transaction using pg_terminate_backend. #31439
• Enables execution of DELETE and (re-)INSERT operations in a single flush for indexed JSONB columns, making updates less expensive. #27948
• Increases default values for parallel_setup/tuple_cost to better balance YB scan node and PG node costs. #28939
• Restores distinct error messages for kAbort and kConflict errors for clearer troubleshooting. #28157
• Displays accurate Storage Flush Requests and Execution Time for non-buffered write operations in the EXPLAIN ANALYZE output. #26963
• Adds synchronization wait time to prevent crashes on startup in clockbound mode. #28377
• Introduces a new gflag ysql_yb_enable_update_reltuples_after_create_index to update reltuples for base table and index after CREATE INDEX, aligning with Postgres behavior. #25394
• Reduces RPCs by using Index Only Scan for INSERT ...ON CONFLICT ...DO NOTHING statements. #28324
• Ensures generated hints correctly support partitioned and child table queries. #28070
• Preserves table locality information for deferred triggers during subtransaction rollbacks. #28725
• Reads placement flags values directly via YBCGetGFlags instead of environment variables. #28697
• Fixes a bug with transactional DDL not considering the yb_force_catalog_update_on_next_ddl flag, preventing crashes during transaction rollback containing an ALTER TABLE statement. #28000
• Resolves issues in TestPgRegressPlpgsql by enabling transactional DDL, reducing the chance of crashes due to many retries when transaction rollback is performed, especially with ALTER TABLE statements. #28000
• Introduces a new flag ysql_auth_method to select the password authentication method for YSQL, ensuring backward compatibility and avoiding user lockouts during upgrades. #28809
• Deprecates the YSQL configuration parameter yb_make_next_ddl_statement_nonincrementing to enable incremental catalog cache refresh by default. #28765
• Adds logic to accurately merge catalog modification aspects in nested ddls when transactional ddl is enabled. #28000, #28791
• Fixes an error where a join condition was incorrectly pushed down in a batched nested loop, preventing crashes during planning. #28112
• Fixes a regression bug that caused an OBJECT_NOT_FOUND error during an index alteration, ensuring smoother table modifications. #28849
• Resets transaction read point for DML statements following a DDL execution in transactional DDL mode. #28871
• Global DDL operations won't be aborted by concurrent (auto)ANALYZE processes in separate databases. #28877
• Prevents dangling pointer issue in YbcFlushDebugContext by properly storing temporary strings. #28913
• Corrects the YB_LAST_USED_OID for the yb_tablet_metadata view. #28902
• Disables transactional ddl and object lock in pg_partman-test due to unsupported transactional ddl. #28879, #28726
• Fixes the test PgHintTableTest.PreparedStatementHintCacheRefresh for transactional DDL, preventing crashes due to many retries when a transaction rollback contains an ALTER TABLE statement. #28000
• Prevents regular DDL within a transaction block from being aborted by auto analyze. #28926
• Accommodates AlmaLinux 9 glibc updates in collation unit tests without failure. #25743
• Enables session defaults to reset correctly during transactions via new packet protocols. #28445, #20603
• Fixes crashes during savepoint rollback for the first DDL in transactions. #29013
• Adds a preflight check to fail upgrades if user-defined roles with the yb_ prefix exist, except yb_superuser. #28977
• Refactors YbBindSearchArray's handling to eliminate potential bugs related to mutating is_column_bound and setting ybScan->all_ordinary_keys_bound, improving result accuracy. #29042
• Refactored index backfill code in YSQL for improved readability and efficiency by standardizing field names and optimizing row reading process. #29052
• Enhances error message clarity when altering the primary key index owner. #29001
• Fixes error in REINDEX INDEX statement execution on partitioned tables with transactional DDL enabled. #29058
• Fixes are introduced to address intra-query memory leaks during long-running write queries, reducing transient memory spikes and improving memory allocation. #29056, #29057
• Prevents unnecessary request dispatches for out-of-bound hash permutations. #29041
• Corrects index-only scans in unbatched INSERT...ON CONFLICT...DO NOTHING queries for multi-column indexes. #29109
• Fixes flaky TestYbQueryDiagnosticsStatus and removes unused test resources. #27998
• Resolves a bug that reset the query id of the walsender process to 7 from 11. #29147
• Enhances log message clarity for AlterTable CHECK failures in PgSchemaChecker to aid debugging in PITR restore scenarios. #28535
• Adds a backend type and libpq option to prevent analyze executions from hanging and ensure smooth CREATE INDEX operations. #29115
• Fixed a bug that caused crashes when arrays of more than two dimensions are passed to a scalar array operation. #29188
• Ensures hash permutations reset correctly for parallel queries on hash-distributed tables. #29198
• Restored legacy mode costing and path creation behavior in YSQL. This is achieved by reverting BNL specific changes and negating a side effect on row count estimation introduced by boolean index fix. #29206
• Enables yb_enable_update_reltuples_after_create_index when the Cost-Based Optimizer (CBO) is turned on using the ysql_yb_enable_cbo flag or yb_enable_cbo YSQL configuration parameter. #25394
• Query ID calculation now excludes database OID for consistency with PostgreSQL. #29848
• Ensures each batch has sufficient PgsqlOps/ReadReqs to avoid segmentation faults. #29343
• Switches to SharedLock in get_ysql_db_catalog_version to prevent unnecessary blocking. #28931
• Fixed an issue with prepared statements on temporary tables when using extended query protocol. #29292
• Fixes vector indexes to support ysql_use_packed_row_v2 flag, preventing data corruption. #29348
• Fixes an issue with empty IN condition lists in parallel queries. #29406
• Limits request and response sizes in YSQL ANALYZE to avoid errors. #29288, #29289
• Changes yb_make_all_ddl_statements_incrementing to a test YSQL configuration parameter to prevent DDL failures. #29405
• Fixed a metadata correctness bug that caused errors when using minimal preloading with negative caching. #29485
• Prevents database crashes by skipping cache invalidation for negative entries. #29606
• Resolves gcc13 compiler warning by using NewTempWritableFile. #29609
• Stabilizes testRelcacheInitConnectionStress by adjusting timeouts and reducing connections. #29552
• Fixes "schema version mismatch" errors after savepoint rollbacks. #29538
• Handles sync timeout more effectively during connection resets. #29173
• Fixes binary search in ASH queries to capture all relevant samples. #29667
• Skips tables in DELETING or DELETED state during rollback to savepoint to prevent errors. #29520
• Fixes InsertIntoContainer template for better type handling and error messaging. #29686
• Fixed read restart error occurring during concurrent execution of CREATE TEMP TABLE. #29704
• Prevents crashes during process shutdown by skipping interrupt checks. #29690
• Fixed a missing build dependency issue in the yb_pgwrapper library that caused a failure when running build with clean-postgres. #29728
• Enhances stability by using separate YbctidReader instances for different components. #29781
• Enhances database session shutdown by reducing wait times and avoiding potential crashes. #29756
• Reduces redundant lock requests in YSQL, speeding up certain DDL operations. #29785
• Ensures all read restart errors display complete debugging information in the detail message. #29762
• Adds GFlag for GUC yb_ignore_bool_cond_for_legacy_estimate, allowing setting via GFlag before upgrading from 2.20. #29831
• Fixes the history file usability in ysqlsh by preserving the header during truncation. #9861
• Enables concurrent materialized view refresh without catalog version increment during upgrades. #29874
• Fixes the shebang path in download_ybc.sh to prevent build errors. #29883
• Ensures object lock acquisition deadlines are computed in microseconds, preventing rounding errors for sub-second timeouts. #29891
• Ensures functions remain accurate after rolling back to a savepoint. #29881
• Restores the mistakenly removed ysql_major_upgrade-test unit test. #29913
• Prevents deadlocks during index creation by optimizing lock handling and analysis processes. #29720
• Introduces separate flags to control DDL catalog bumping and negative caching. #29941
• Fixes crashes during complex row comparisons in YSQL queries. #29942
• Enhances error messages when dropping a table concurrently to be more informative. #28532
• Adds yb_ycql_utils to supported extensions for YSQL major upgrades. #29973
• Reduces cache misses by not using metrics_capture in catalog cache keys. #30030
• Fixes metric recording to ensure all valid events are captured. #28340
• Enhances geolocation cost estimation using cluster configuration when no tablespace is assigned. #29313
• Fixes parallel index scans on hash indexes to handle duplicate column issues. #29021
• Allows successful build on macOS 15.7 with Apple Clang 17. #30022
• Renames the YSQL configuration parameter for clearer metric designation. #30069
• Ensures all clockbound contexts synchronize at startup for consistent timing. #30033
• Enables built-in support for Levenshtein string matching, reducing network overhead. #29299
• Reduces read restart errors when refreshing materialized views during concurrent DMLs. #29084
• Enables faster read operations requiring writes by using write pipelining. #30038
• Enhances ASH visibility for index backfill master orchestration. #30338
• Enhances ASH tracing for better monitoring of PostgreSQL backend synchronization. #30103
• Disallows SAOP merge in parallel queries to ensure result order consistency. #30096
• Fixes error when running parallel queries with append plans. #28920
• Fixes the unique constraint violation for ALTER statements in transactions. #30109
• Fixes a race condition in creating listen/notify objects. #1872
• Ensures docdb_wait_time is now visible in pg_stat_statements for Featurebench workloads. #30194
• Enables concurrent index creation on different tables without errors. #30114
• Sets catalog version for ybc_heap_beginscan to ensure consistency and correctness. #30270
• Enhances reliability in parallel query execution by embedding metric structures directly. #30126
• Reverts a change to prevent crashes during certain table alterations. #30199
• Enables concurrent execution of CREATE INDEX CONCURRENTLY and REFRESH MATERIALIZED VIEW without blocking. #30219
• Fixes a bug to ensure concurrent table creation doesn't fail in YSQL. #30219
• Prevents deadlocks during ANALYZE with concurrent DDL by using separate transactions. #30232
• Logs call stacks in YSQL on critical errors for easier issue diagnosis. #26952
• Ensures catalog version is sent during YbBitmapTableScan to prevent mismatches. #30250
• Stops logging backtraces for FATAL errors due to connection limits. #30323
• Ensures read restart errors include detailed logs during catalog version mismatches. #30343
• Fixes data race issues during metric updates in YSQL. #30432
• Enhances xCluster to support vector indexing, ensuring accurate replication and indexing of data post-backfill. #26679
• Enables YbctidReader to handle SKIP LOCKED with a new locking strategy for batch reads. #23584
• Ensures xCluster setups function correctly with concurrent DDL by using safe read points for catalog operations. #30361
• Adds a NOTICE for index rebuilds during ALTER TYPE operations. #30416
• Enables Query Plan Management by default, tracking all hint-generating queries. #30822
• Eliminates read restart errors in backup and restore during concurrent DDLs. #30178
• Fixes parallel scan issues when using yb_hash_code conditions. #30204
• Adds logging to troubleshoot "schema version mismatch" errors in DDL transactions. #30367
• Fixes misleading EXPLAIN text output for scans under serializable isolation. #30373
• Enables DEBUG in auto_explain with new YSQL configuration parameter. #25067
• Disables top-level retries for CALL or DO statements after a commit. #30459
• Reduces unnecessary RPC calls in QPM by checking valid OIDs. #30532
• Fixes undefined behavior in PostgreSQL code to ensure test stability. #30489
• Resolves timeouts during concurrent index creations by ensuring correct catalog version tracking and reducing unnecessary RPC calls. #30491
• Adds debugging logs to help identify schema version mismatches after dropping indexes. #30367
• Enables aggregate subqueries in global views by using the extended query protocol. #30501
• Stops storing un-normalized query strings in pg_stat_statements. #30592
• Corrects parallel index scan rescan issues for accurate query results. #29736
• Reduces errors in READ_COMMITTED transactions by correctly flushing row locks. #30620
• Adds debug logs for tracking pgstat catalog version updates in PG backends. #30491
• Prevents PostgreSQL shutdown errors by cleaning up dangling pgstat entries on replication slot drops. #30646
• Ensures JSON format in QPM correctly includes the closing bracket. #30617
• Increases the timeout for backend catalog version updates to 900 seconds. #30649
• Restores default value for yb_enable_pg_stat_statements_docdb_metrics. #30655
• Ensures SELECT queries correctly follow transactional DDLs by waiting for schema updates. #30367
• Fixes catalog version conflicts during concurrent database operations. #29856
• Restores the ./yb_build.sh daemons no-odyssey functionality on Linux. #30718
• Increases sleep duration during YSQL replication slot creation to handle clock skew. #29481
• Allows renaming the yugabyte role without impacting the notifications poller. #30722
• Ensures consistent read values within a single SQL statement in legacy mode. #30739
• Disables parallel workers with concurrent DDL to prevent errors. #30749
• Fixes a memory leak in handling PGresult objects. #30765
• Ensures planner stats only appear once in EXPLAIN JSON output. #30768
• Automatically sets read time during parallel query execution to prevent errors. #30588
• Ensures remote servers correctly connect to user-specified databases for global view queries. #30849
• Fixes occasional crashes during concurrent create index tests. #30880
• Now ensures CDC properly cleans up resources after notifications are processed. #30702
• Renames GUC yb_max_saop_merge_streams to yb_max_merge_scan_streams for clarity. #30943
• Enhances handling of duplicate notifications in the async queue to avoid client-side errors. #30947
• Fixes timezone issues for expression pushdown by initializing my_exec_path correctly. #30815
• Ensures stable parallel worker operations by using legacy catalog read mode when yb_enable_concurrent_ddl=true. #31008
• Reduces unnecessary rechecks by folding conditions on the same column. #30525
• Ensures updates to partitioned tables correctly check partition keys to prevent errors. #31214
• Enhances error messages for column issues, including table identity details. #31304
• Adjusts YSQL catalog timeouts based on heartbeat_interval_ms for reliability. #31431
• Switches concurrent DDL to a non-runtime, cluster-wide flag for consistency. #31225
• Resolves OID conflict by assigning new OID 8116 to graphid. #31499
• Handles records with new action types by logging warnings, preventing process crashes. #30703
• Enhances stability by using IndexOptInfo for index expressions, preventing crashes in yb_derive_equal_cond. #31313
• Prevents crashes in PG backend when running empty queries with tracing enabled. #31476
• Resolves a critical caching issue that led to "no partition for row" errors. #31272
• Ensures consistent reads by fixing bug in read restart errors handling during UPDATE operations. #28628
• Resolves packet forwarding race conditions in pipelined queries with a new flag. #31368
• Adds SQL function to check OpenSSL FIPS mode and a flag to disable non-FIPS crypto. #31606
• Prevents crashes by properly handling operation failures in the buffer. #31616
• Now waits for index deletions during rollback to savepoint, preventing errors. #31601
• Prevents premature DDL transaction aborts during ROLLBACK TO SAVEPOINT. #31720
• Reduces test flakiness by adjusting max_clock_skew_usec and enabling pre-loading of catalog tables. #28628
• Ensures consistent reporting of query execution times between PGSS and QPM. #30793
• Handles non-retryable CDC errors by terminating relevant processes to prevent silent failures. #31484
• Reduces connection errors with server socket closure verification in ConnMgr. #31291
• Ensures ASH correctly attributes catalog reads to the proper query in extended protocol. #31325
• Reduces misattribution of wait times in query monitoring by ensuring proper query ID tracking across RPC retries. #31820
• Fixes Batched Nested Loop Join to return correct query results with correlated conditions. #31724
• Fixes race condition error during concurrent DDL and DML operations. #31869
• Rejects empty hint/plan texts in yb_pg_stat_plans_insert to prevent core dumps. #31922
• Resolves false deadlock issues during CREATE INDEX operations by improving session transaction handling. #31594
• Enables consistent foreign key checks when toggling yb_enable_fkey_batched_docdb_lookup_when_types_mismatch. #31861
• Fixes retry logic for DROP TABLE to prevent SMgrRelation errors. #29871
• Allows redundant columns in merge scans to prevent errors and crashes. #31984
• Fixes OOB read errors in single-stream merge scans with duplicate IN lists. #32050
• :YSQL:handle RelabelType in merge scan ybIsClauseEligibleSaop does not account for RelabelType nodes wrapping the LHS of SAOP clauses.This causes the planner to miss merge-scan-eligible clauses for columns with binary-compatible coercions (.varchar,which gets coerced to text).Strip the RelabelType wrapper at the three points in yb_merge_scan.c.[YSQL]Merge scan does not work on varchar columns Jira Link:Description Repro:The planner casts varchar to text.The ybIsClauseEligibleSaop function should strip RelabelType nodes from the expressions.Issue Type kind/bug ybIsClauseEligibleSaop does not account for RelabelType nodes wrapping the LHS of SAOP clauses.This causes the planner to miss merge-scan-eligible clauses for columns with binary-compatible coercions (.varchar,which gets coerced to text).Strip the RelabelType wrapper at the three points in yb_merge_scan.c. #31200
• Enhances FK reference checks to wait during conflicts, aligning with PostgreSQL behavior. #32047
• Enables automatic retries for CALL and DO statements under READ COMMITTED. #31924
• Fixes the error when using the yb-ts-cli dump_tablet_data tool on YSQL tablets. #31597
• Ensures correct logging of yb_max_pools as a number, not a boolean. #30347
• Switches to thread-safe customertime_r in the logger to prevent data races. #28264
• Ensures all PG wire protocol packets are fully received before processing. #32087
• Prevents crashes during failed 'DROP TABLE IF EXISTS ...CASCADE' retries. #31248
• Fixes memory leak in 'CREATE TABLE AS SELECT' by managing memory per-tuple, and reduces memory usage during REFRESH MATERIALIZED VIEW operations. #31962
• Fixes bitmap scan errors where wrong primary key columns were read. #32132, #32123
• Optimizes memory tracking for hash key permutations to prevent potential outages. #32196
• Disables TSAN thread leak reports in Connection Manager. #30745
• Fixes race condition in ConnMgr multi-route pooling by using an atomic counter. #30736
• Stops memory leak reports after lock timeouts by disabling the sanitizer. #30090
• Ensures all tablet servers are registered before initdb, preventing test failures due to startup races. #31029
• Now handles catalog version mismatch and table existence errors more reliably. #31345
• ysql_dump now respects the PGHOST environment variable, prioritizing the -h flag. #29976
• Fixes a rare condition where parallel queries could leave a dangling pointer, reducing crash risks. #31895
• Disables parallel queries for colocated tables by default. #32318
• Ensures consistency in QPM data by locking the table during entry population. #32275

YCQL

• Enables retrying YCQL analyze steps for OBJECT_NOT_FOUND errors to refresh cache. #29589
• Prevents crashes by correctly handling NULL index references in YCQL queries. #29754
• Enables batched queries in YCQL and improves ASH joins. #18047
• Ensures YCQL truncate command waits for index tables as well. #30121
• Ensures YCQL backfill correctly checks for newer duplicate entries. #30330
• Enhances handling of exceptions by using toolchain libunwind. #30870

DocDB

• Allows removal of associated intentsdb keys during tablet startup to prevent potential pileup and slower queries. #21831
• Prevents silent data loss by failing the clone when a table is altering instead of cloning without data. #27621
• Solves the reversed condition in the rpc_inbound_calls_failed metric, offering more metric information. #28788
• Ensures tablet split handling on the consumer side does not cause incorrect mappings due to race conditions. #28750
• Add a test for cloning to a time before dropping a materialized view and a column in the base table. #23740
• Changes prevent unnecessary node crashes due to network issues during remote bootstrapping process cleanup. #28830
• Prevents TServer process crash and unusable Postgres connections by correctly handling thread creation failure and falling back to RPC mechanism. #28968
• Fixed deadlock in MasterSnapshotCoordinator by modifying PopulateDeleteRetainerInfoForTableDrop to pass table_info to IsTabletCoveredBySnapshot. #28679
• Adds extra logging to track data inconsistency when durable_wal_write is set to false. #23453
• Disables TryConsume usage by LogReader to prevent user operations from being incorrectly blocked due to a bug. #29095
• Now only live tablet servers are contacted when updating transaction version, enabling deletion of transaction tables even if a tablet server is down. #29067
• Reduces log replay during bootstrap by synchronizing intents DB flushed op ID with regular DB. #29184
• Enhances rocksdb_bytes_read metric to include iterator read bytes. #26992
• Prevents duplication of restoration requests in tablet metadata. #27183
• Enhances error messaging for database cloning operations outside allowed timeframes. #21866
• Ensures xCluster stability by using table_id to populate namespace_id in metadata, preventing issues from namespace renames. #29901
• Changes error message to "Unable to find the leader" for clarity. #30000
• Reduces latency by returning immediately after the first successful RPC in GetMasterEntryForHosts. #28844
• Fixes transactions to default to region-based locality during upgrades involving tablespace locality support. #29645
• Prevents data loss on followers during tablet splits by using MVCC for split operations. #30117
• Fixes tablet split validation to handle 0x0000 as an empty key for hash partitions. #30062
• Enhances sst_dump to correctly handle packed rows without schema information. #29729
• Fixes automatic tablet splitting when default_time_to_live is set to 0. #29710
• Prevents thread exhaustion on tserver during remote bootstrap failures. #26813
• Ensures successful cloning of altered databases by managing connection permissions during the operation. #28141
• Enhances error messages to indicate that cached errors reset only upon TServer restart. #25904
• Ensure file safety during checkpoints by explicitly flushing key files like MANIFEST and CURRENT. #30146
• Ensures unique index creation checks respect transaction write order correctness. #30332, #30312
• Resets full report sequence number on tserver re-registration to avoid data inconsistencies. #30169
• Ensures consistent data by handling write errors during shutdown properly. #30093
• Reduces false positives during tablet splits by not reapplying transactions already in the regular database. #30334
• Ensures xCluster replication DDL cleanup by using accurate schema names. #30529
• Fixes batch limits in Raft followers' catch-up process. #30437
• Correctly detects transactions aborted due to deadlocks. #30641
• Prevents data loss during tablet splits by ensuring transactions commit correctly to child tablets. #30555
• Ensures data integrity during compaction by safely merging updates. #30837
• Fixes the "thread pool not ready" error during large transaction processing. #30772
• Enhances child tablet bootstrap speed and reduces memory use by copying bootstrap state during splits. #30803
• Avoids confusion by dynamically fetching current DB names in list_snapshot_schedules. #26274
• Prevents tserver crashes during shutdown by improving transaction handling. #29908
• Prevents infinite WAL segment growth during disk space issues. #31061
• Prevents crashes during high xCluster churn by fixing null pointer issues in metrics aggregation. #31316
• Prevents crashes on upgrade by ensuring schema version lists are properly initialized and compatible. #31533
• Prevents cluster-wide outages by rejecting UpdateConsensus RPCs early when tablets enter a hard write stop. #30728
• Resets has_faulty_drive status automatically when a TServer reregisters. #31332
• Fixes a bug where scans may miss rows due to stale bloom filter pointers. #31688
• Ensures stable data handling during partial compactions, preventing errors like "duplicate key". #28314
• Ensures MetaCache correctly refreshes to include child tablets after a split. #31936
• Eliminates spurious warnings when releasing xCluster DDL queue advisory locks. #31963
• Ensures xCluster non-transactional safe-time logic only sets valid times, reducing false warnings. #31967
• Enables ysql_dump with read-time option for consistent point-in-time backups. #20480
• Replicates DDLs with session variables to ensure consistency across clusters. #28895
• Allows namespace_id to be correctly copied in TableInfo constructors to prevent empty return values. #28478
• Eliminates race conditions between snapshot creation and concurrent DDL operations. #29474
• Enhances xCluster by ensuring indexes are replicated with associated table DDLs. #29699, #29133
• Fixes vector index searches to return complete results post-tablet splits. #29543
• Prevents failures in simultaneous table creations across different databases in the same replication group. #29797
• Ensures correct syntax in CREATE INDEX statements with special options like colocation_id. #29337
• Uses full hybrid time for xCluster DDL queue polling, enhancing replication accuracy. #30019
• Adjusts cgroup v2 setup for system services, fixing build failures on alma9 and ubuntu22. #30609
• Blocks nextval and setval in time travel sessions when yb_read_time is set. #30881
• Blocks advisory lock acquisitions during time travel sessions. #30929
• Adds a session-level advisory lock for ddl_queue_handler to prevent multiple handlers running concurrently. #27186
• Adds a new enable_object_locking_infra flag and YSQL configuration parameter to activate table-level locks only after upgrading all nodes. Both this flag and enable_object_locking_for_table_locks must be true to activate the feature. This flag is tagged as PGC_INTERNAL preventing users from toggling it. #28875
• Allows Parallel Query feature to work correctly with object locking by limiting object lock and finish transaction calls to the leader worker. #28878
• Prevents fatal errors in transactions by relaxing subtransaction checks. #29023
• Reduces pg_locks transient errors by verifying tablet leaders during lookup and increasing retry attempts. #28180
• Fixes a TServer crash issue during hybrid scans caused by incorrect read requests from Postgres. #29347
• Prevents race conditions during YSQL lease re-acquisition by synchronizing it with lock requests. #27829
• Ensures batched COPY operations securely re-acquire table and index locks, reducing partial copy failures. #29327
• Fixes bugs in Prometheus metrics scraping for tables with special characters in their names. #29351
• Prevents shutdown errors by checking for active tasks during registry closure. #29593
• Switched from using SIGQUIT to SIGINT to stop the PostgreSQL postmaster, preventing core dump generation at TServer startup. #29670
• Adds transaction intent record logging for easier debugging of crashes. #14641
• Enhances xCluster DDL replication by including schema names in relation maps. #29598
• Ensures all tablets have updated schemas for xCluster replication, improving error handling and consistency. #28326, #28946
• Introduces min_replay_txn_first_write_ht to address transaction data loss that occurred due to incorrect filtering of transactions on bootstrap. #29642
• Eliminates false conflicts in fastpath object locking by replicating PostgreSQL's matrix. #29793
• Automatically cleans up discarded checkpoints to prevent disk full issues during remote bootstraps. #29909
• Enhances xCluster replication by skipping unnecessary tuple generation in ATRewriteTable. #29594
• Enables hot certificate reloading for client-to-node encryption. #29631
• Adds logging for database creation and deletion events in ysql_db_catalog_version_map. #29476
• Allows xCluster DDL replication setup with materialized views. #29970
• Fixes ysql_dump to correctly handle vector indexes in colocated databases. #28631
• Prevents deadlocks by adjusting lock handling during simultaneous operations. #29476
• Increases the RPC timeout to 1 hour for AddTableToXClusterTarget tasks. #30072
• Prevents xCluster replication from halting when dropping an invalid index. #26633
• Ensures consistent lock behavior across serializable transactions when toggling skip_prefix_locks. #30014
• Prevents lock leaks by ensuring masters grant new leases instead of refreshing expired ones. #30183
• Fixes a specific deadlock during post-split compaction on vector indexed tables. #30290
• Prevents OID collisions by invalidating tserver cache after ImportSnapshot. #29335
• Handles backfill requests for deleted tables without crashing. #29830
• Ensures YCQL unique index backup and restore functions correctly without schema version errors. #29789
• Fixes error during backward scans by checking key bounds before decoding. #30607
• Fixes decoding issues in tables with vector indexes using packed rows v2. #30762
• Preserves clone sequence numbers to ensure cloning idempotency after PITR operations. #30958
• Enables reliable creation of vector indexes by fixing MetaCache handling and validation for tablet splits. #26381
• Reduces read restart errors for autonomous DDLs by using current ht instead of safe time. #29139
• Fixes cloning failure for sequence tables due to old snapshot errors. #31278
• Adds use_cgroups_cpu flag to respect CPU limits in containerized environments. #29844
• Prevents tserver crashes during shutdown with in-flight split or clone operations. #28463
• Enhances cloning for databases with multiple colocated parent tables. #31285
• Enables cloning operations to succeed after a PITR that hard deletes tables. #31491
• Fixes corruption error on Master UI for vector index partitions. #27674
• Ensures successful restore of YSQL backups with manually split base tables. #31185
• Disables periodic retriggering of DDL verification to prevent crashes. #31815, #31430
• Fixes a crash loop issue caused by incorrect PREPARING state of vector indexes in clones. #32031
• Sets namespace_id in tablet metadata at creation to skip later backfill. #29982
• Fixes TServer crash during xCluster replication after node scale-down. #30991
• Ensures flag descriptions are clean by removing Jira links. #31136
• Refactors background task handling for clarity and maintainability. #28787
• Fixes a regression that allowed restoring to a time earlier than the maximum retention window, potentially causing database instability. #28699
• Prevents yb-master crash by making MessageDifferencer thread-safe and limiting its use to debug builds. #28729
• Allows deletion of hidden tablets during point-in-time restores that are no longer in use. #28704
• Allows AutoAnalyze to work consistently, regardless of table level locks feature, by always acquiring the acquire_for_update_lock lock. #28783
• Reduces Unable to find schema name for YSQL table errors and optimizes the retrieval of YSQL Table Oid. #28793, #27590
• Eliminates shutdown deadlock in CQL service, enabling consistent and graceful server shutdown. #28438
• Ensures transactions start with the correct locality during object lock acquisitions. #28317
• Allows table-level locks to respect the lock_timeout setting, returning an error if lock is not obtained in time. #27677
• Redundant finish transaction calls no longer request a new transaction when releasing object locks if a DocDB transaction doesn't exist. #28761
• Enables "SELECT FOR KEY SHARE" to avoid crashes during cluster restart by starting each run with a clean state. #28881
• Adjusts advisory lock release timing to prevent handler blockages during shutdown in time-sensitive builds. #28889
• Reduces master_rpc_timeout_ms to 1.5s to lessen outages by quickly detecting dead/unavailable master nodes. #28924
• Prevents destructor calls on partially initialized services. #28925
• Enhances stability by handling upper bound updates correctly across all scenarios in HybridScanChoices. #28945
• Modifies fast path transaction criteria to exclude object locking writes. #28922
• Prevents vector index compaction failure when all vectors are filtered out during compaction. #29016
• Excludes invalid indices from snapshots to prevent backup restore failures. #28986
• Adds object lock auto-flag checks to YbMaybeLockMasterCatalogVersion and AcceptInvalidationMessages for upgrade safety. #29054
• Supports successful backups during tablet splits. #29002
• Adjusts default replication factor in ExternalMiniClusterOptions to match num_masters, potentially fixing some tests. #29098
• Implemented root memtracker adjustment to limit deviation from TCMalloc stats, avoiding overestimation and false triggering of memory limits. #29094
• Resolves deadlock detection issues in PgLibPqTest.PagingReadRestart by temporarily disabling it. #28891
• Ensures DDL operations wait when DML is in progress with object locking. #29150
• Enables object and table locking features by default in release builds for enhanced QA validation. #29224
• Eliminates unexpected system exits when DDL transactions enter the tablet wait-queue in a colocated database with index. #29215
• Fixes bug where ALTER TABLE statements would not roll back in case of master crash during the ALTER TABLE DDL operation. #27021
• Ensures PG connections handle restarts correctly with object locking. #29130
• Disables inline waiting for PG connections on cluster start with object locking. #29236
• Adds debug logs to assist in debugging issues related to concurrent DDL execution without object locking. #29080
• Resolved a race condition between server shutdown and tablet cleanup by moving startup tablet cleanup to a background thread and adding a check to avoid crashes if shutdown is in progress. #29022
• Fixes a process crash that could occur during unique index checks due to incorrect functioning of DBIter's fast next mode with backward iteration. #28960
• Ensures accurate schema during backups amid concurrent schema alterations and snapshot creations. #28766
• Ensures yb_index_check correctly acquires object locks, preventing debug assertion failures. #29389
• Switched from using SIGKILL to SIGQUIT and SIGINT for stopping postmaster process during lease loss and tserver shutdown respectively. #28670
• Adjusts object locking sanity checks to transaction level granularity. #29390
• Fixes transaction-related DFATAL issues by enabling concurrent updates. #29512
• Ensures consistent behavior across sessions during upgrades by caching table lock flag values. #29515
• Fixes crash when reading Postgres tables via CQL API due to incorrect null handling. #29498
• Enhances isolation test coverage by adding partition-drop-index-locking and adjusting timeouts to align with PostgreSQL behavior. #21705
• Disables tablet splits during table hide operations to prevent state inconsistencies and lock leaks. #29559
• Ensures correct resetting of the wait state in the WAL appender thread. #29586
• Ensures PostgreSQL works with node_to_node_encryption_use_client_certificates. #7335, #26138
• Ensures vector index searches function correctly across sessions. #29445
• Reduces TServer does not have a live lease errors for pg clients by adding retry logic. #29600
• Automatically resets stale transaction IDs on transaction failure to prevent lock leaks. #29628
• Disables intent filtering for bootstrap due to transaction filtering issues with use_bootstrap_intent_ht_filter set to true. #29664
• Adds logging for RestoreSnapshotSchedule RPCs, aiding in debugging. #28707
• Eliminates data race in VectorLSMInsertRegistryBase shutdown handling. #29745
• Prevents potential double decrement errors in object lock state management. #29790
• Reduces lock overflow issues by treating redundant lock requests as no-ops. #29774
• Prevents restore failures during schema merges in backups taken during DDL operations. #29765
• Fixes function calls to accurately interpret is_using_table_locks. #29821
• Enables use of LTO for prof_gen builds, as issues with older compiler versions are resolved. #15093
• Fixes hashing split keys to respect the partition's lower bound. #29888
• Adds a null check to prevent indefinite wait states in WAL appender threads. #29586
• Prevents fatal errors by storing status tablet data for global object lock requests. #29323
• Fixes yb-admin corruption errors for tablet partitions by using the correct schema. #29992
• Reduces unnecessary VI_STATS logging in pg_client_session.cc. #30070
• Restores superclass tserver flags in PgCboDependentFlagsTest to prevent brittleness. #30077
• Ensures PRIMARY KEY indexes process before UNIQUE indexes during ALTER operations to prevent double index creation. #30055
• Ensures load balancer removes old read replicas when tablets move to a non-replica tablespace. #30141
• Skips parent tablet in data dumps for colocated tables. #30220
• Adds a tool to convert text protos to YugabyteDB binary format. #30124
• Fixes a race condition in kPgSession transactions to better detect deadlocks. #30252
• Improves error messages when yb-ctl timeouts occur during startup. #30278
• Removes undefined GetAtomicFlag and SetAtomicFlag functions for better reliability. #30321
• Ensures unique index backfill doesn't encounter SnapshotTooOld errors. #30329
• Reverts an Abseil enablement on Mac due to failing performance tests. #30339
• Fixes static initialization order bugs for more reliable value constructions. #30375
• Ensures failed transactions trigger early cleanup to prevent issues. #30499
• Relaxes namespace name checks during the cloning process to allow cloning from renamed namespaces. #30313
• Ensures skip prefix lock mode consistency across transaction operations. #30587
• Ensures accurate tracking of all RPC rejections due to memory pressure. #30616
• Ensures correct replay of operations for vector indexes added or removed during tablet bootstrap. #30684
• Escapes stack traces in /threadz to prevent HTML rendering issues. #30689
• Ensures txn status table additions are saved correctly to prevent issues after failovers. #30652
• Uses the pgwrapper::PqEscapeLiteral function to escape literals in clone scripts. #30694
• Prevents object locks from becoming orphaned during PITR restoration, ensuring proper DDL operation continuity. #30639
• Ensures yb-ctl uses Python 3 in releases to avoid failures. #30713
• Reduces crashes by copying shared pointers locally before releasing locks during metrics aggregation. #30758
• Fixes a bug in COPY implementation to ensure data consistency. #30782
• Reduces false conflicts during Object Lock Manager bootstrap by sorting lock entries. #30795
• Prevents cloning into the inconsistent "dead zone" after a PITR restore. #30794
• Reduces spam by correcting leader stepdown logging. #30788
• Marks the flag for selecting vector index backend as hidden and advanced. #30855
• Fixes top-level intent key size calculation for colocated tables. #30862
• Ensures valid leader preference settings with wildcard placements in clusters. #30821
• Prevents TSAN race conditions by using ANNOTATE_UNPROTECTED_WRITE in tests. #30902
• Ensures error codes are correctly set for RemoteExec RPCs in global views path. #30981
• Fixes a crash issue when ReleaseObjectLocksGlobal RPC times out during TServer restarts. #30986
• Fixes specific TSAN test failures by adjusting flag validators and test configurations. #31044
• Corrects flag description display in the documentation. #31054
• Fixes use-after-free errors in transaction RPC callbacks, enhancing stability. #29146
• Enables ysql_dump DB clone to ignore non-existing roles using the dump-role-checks option. #26555
• Ensures replication setup does not fail by preserving colocation_id for vector indexes during backup. #31216
• Ensures tablet shuts down smoothly by modifying operation types and adjusting storage destruction order. #30770
• Prevents crashes by handling cgroup initialization failures more robustly. #31425, #31443
• Enhances database cloning with new ysql_dump options for renaming databases and owners. #31482
• Fixes race condition issues during cgroup setup in PostgreSQL backends. #31595
• Adds flags to configure memory reset thresholds for aggregate calculations. #31660
• Corrects the typo in the ASH wait state code kConflictResolution_ResolveConflicts. #31902
• Ensures arc lint works correctly by updating the linting type to flake8. #31949
• Ensures high-priority TServer threads run in the correct SystemHighCgroup, enhancing response times and stability. #31710
• Enables static linking for libxml2 and dependencies in LTO releases. #31851
• Sets PostgreSQL processes to use the default TServer cgroup // instead of /. #31712
• Adds tablet ID to MinRunningHybridTime ASH samples for better tracking. #31909
• Enhances xCluster replication by reconnecting after a PostgreSQL restart. #31964
• Ensures async writes during tablet splits are correctly rejected, preventing database inconsistencies. #31912
• Adjusts pg_client_use_shared_memory flag rules based on object locking settings. #31307
• Reverts to using glibc backtrace for stability in the 2026.1.0 release. #32197
• Ensures all threads are correctly assigned to their designated cgroups under QoS settings. #31751
• Reimplements libunwind for safer and more reliable stack trace collection across all builds. #31453
• Resolved rounding errors in StopWatch by changing calculation order. #29205
• Stops empty sample blocks from being incorrectly counted in analytic calculations. #29203
• Ensures transactions default to region-local during upgrades when table locks are active. #30244
• Reduces log clutter by removing repetitive connection messages. #30581
• Stops creation of vector index block cache when RocksDB cache is disabled. #30763
• Eliminates race conditions and crashes in YBThreadPool by ensuring worker memory safety. #28297
• Fixes shutdown delays caused by remote bootstrap completion checks. #31255
• Logs now show when a master marks a tserver as unresponsive. #30261

CDC

• Ensures CDC state table entries are properly deleted when databases and streams are dropped. #23497
• Fixed issue in PeerMessageQueue::ReadReplicatedMessagesForConsistentCDC where WAL Ops were missed due to bugs in record reading and handling of approaching deadlines. #29060
• Adds a new yb-ts-cli command to release CDC retention barriers on specific tablets. #31187
• Allows UPAM to update certain barriers even with "before image not found" errors. #31187
• Ensures data consistency by correctly updating operation IDs only after quorum commits. #30181
• Handles error propagation from CDC to client for missing footer in segments. #28592
• Optimizes the WalSndWakeup function by returning early to prevent unnecessary spin lock acquisitions and core dump occurrences. #28379
• Allows large transactions for unqualified tables to stream correctly, preventing the virtual WAL from getting stuck. #28744
• Fixed issue where child tablets wrongly inherited parent tablet's checkpoint after a tablet split, causing CDC stream failures. #28840
• Adds support to filter aborted subtransactions (rolled-back savepoints) in CDC streaming, making it consistent with the apply path. #22817
• Fixes CDC issue with virtual WAL restart after a DDL, preventing record decoding failure. #29136
• Fixes the issue where UPAM gets stuck for colocated tablets with indexes, ensuring continuity of CDC operations. #30067
• Prevents SIGSEGV in CDC operations by ensuring LogReader is not null. #30174
• Excludes sys catalog entries from active stream checks in logical replication. #30382
• Handles multi-level tablet splits correctly in CDC operations. #30210
• Prevents data loss in CDC by retaining rewritten colocated tables. #30521, #30167
• Fixes CDCSDK crash on ALTER TABLE with colocated tables, ensures data streaming continuity. #30521, #30167
• Disables cleanup of expired tables in hash range-based CDC when the flag is true. #31487
• Prevents the PG connector crash during upgrades and rollbacks in replication scenarios. #27177
• Prevents cleanup of expired tables in hash range-based CDC streams to avoid initialization failures. #31487
• Stops applying the LogReader memory limit to CDC, a setting intended only for xCluster. #29066
• Adds "Log index cache entry for op index not found" error to retryable errors in cdcsdk_virtual_wal. #29061
• Enables CDC to distinguish between INSERT and UPDATE for packed rows using the ysql_mark_update_packed_row flag. #16682
• Fixed error in UpdatePeersAndMetrics for Sys Catalog tablet, improving handling of pg_publication_rel and pg_class tables. #29140
• Fixes endpoint verification failures by using DNS hostnames for local connections. #29669
• Now deletes associated streams when a database is dropped to prevent errors. #27835
• Enables cdc_enable_savepoint_rollback_filtering by default, enhancing CDC support. #29396
• Reduces lingering CDC metrics after stream deletion by clearing cached entries. #30771
• Corrects resource retention by blocking CDC state table entries for ineligible tables. #30773
• Eliminates empty BEGIN+COMMIT pairs in multi-shard transactions after tablet splits. #30980
• Fixes missing COMMIT in CDC stream for dropped colocated indexes. #31101
• Reduces log spam by expiring stale entries for logical replication streams. #31235
• Enhances CDC restart time handling to progress when there's no new data to stream. #31091
• Preserves xrepl_origin_id in CDC messages during transaction resends. #30116
• Prevents errors in follower masters when they become leaders by ensuring sys_catalog retention barriers are updated. #30445
• Ensures new replication origins are visible immediately after creation. #30068
• Reduces connector crashes during DDL operations in CDC workflows. #31784
• Prevents data loss by disallowing CDC table rewrites on colocated tables. #31908
• Logs a message if child tablets lack state table entries due to workflow races. #23567
• Fixes potential use-after-free error in CDC state table post-shutdown. #15507
• Enables enable_table_rewrite_for_cdcsdk_table by default, resolving a deadlock in DeleteTableInMemory. #32029