What's new in the YugabyteDB v2024.2 LTS release series

Improvements

YSQL

• Enhances logging for YSQL version mismatch errors and DDL operations. #20084
• Enables control over in-place index updates with yb_enable_inplace_index_update. #20908
• Allows use of UNLOGGED tables syntax, handling it as LOGGED with a warning. #23895, #23889
• Reduces memory usage by compressing large catalog objects in catcache. #21040
• Enables DDL atomicity by default, enhancing schema version consistency. #22097
• Adds a new YSQL view for YCQL statement metrics to aid in query analysis and troubleshooting. #20616
• Displays distinct index columns directly in EXPLAIN output. #20831
• Enables controlled activation of the DDL atomicity feature during upgrades. #21535
• Enhances session management by updating read times more consistently. #21623
• Allows loading only pg_shdepend in YbInitPinnedCacheIfNeeded for cross-database DDLs. #21635
• Enables monitoring of YCQL statements through a new YSQL view, enhancing query analysis. #20616
• Enhances schema version consistency by re-invalidating table cache post-ALTER TABLE with DDL atomicity. #21787
• Adds new monitoring columns to localhost:13000/statements for enhanced tracking. #21735
• Enables logging for global-impact DDL statements that adjust all database catalog versions. #21826
• Sorts YSQL extensions into three specific directories for better organization. #21897
• Fixes schema version mismatch after ALTER TABLE in production builds, ensuring reliable schema updates. #21787
• Introduces the yb_parallel_range_size parameter to customize parallel range sizes. #21928
• Removes the unused keep_order field from YbctidGenerator for streamlined processing. #21944
• Introduces the yb_enable_parallel_append flag to disable parallel append and avoid exposing unannounced features. #21934
• Enhances code clarity and performance with style and logic refinements. #22004
• Enables creation of dummy vector indexes in YSQL with placeholder functionality. #22195
• Disallows move and copy operations for LWFunction to prevent misuse. #22069
• Streamlines and deduplicates logic for ybgin and lsm access methods, enhancing support for new index types. #22195
• Enables CREATE/DROP ACCESS METHOD grammar, supporting custom access methods and extensions. #22364
• Clarifies log messages for DDL atomicity in TableSchemaVerificationTask. #22459
• Eliminates duplicated logic for YSQL hidden column ybrowid handling. #22536
• Adds save and restore functionality for ConsistentReadPoint state. #22597
• Ensures consistent table names in backup/restore scenarios by not renaming DocDB tables during legacy rewrites. #22802
• Disables Python checks on all third-party PG extensions. #23166
• Simplifies coding in PgDml and related classes by streamlining parameters and functions. #23192
• Enables the optimizer to prefer backward scans when FLAGS_use_fast_backward_scan is enabled. #22370
• Adds new flags for role management in YSQL backups and restores. #20972
• Enhances code clarity and documentation for Batch Nested Loop Join logic. #23407
• Introduces a dedicated auth-backend for YSQL to handle connection authentication, enhancing security and streamlining the process. #24147
• Ensures correct error handling by adding a missing InvalidOid check in YSQL. #24147
• Enables authentication via a specialized backend, enhancing security and reliability. #24147, #24427
• Enables toggling authentication method in YSQL using the ysql_conn_mgr_use_auth_backend flag. #24147

YCQL

• Adds error message for unsupported GROUP BY in YCQL; introduces flag to suppress it. #13956
• Resolves timestamp precision mismatches in YCQL, aligning inserts and index scans to millisecond precision. #11052, #23476

DocDB

• Tracks and analyzes disk I/O by stack trace, enhancing debugging and system understanding. #17993
• Resolves a build failure that arose from a naming conflict in debug flags. #17993
• Blocks writes based on total bytes being flushed, not just memtable count. #22571
• Enables background DNS resolution to prevent RPC delays. #22930, #22311
• Enhances load balancer to prioritize under-replicated tablets first. #20263
• Introduces a flag to toggle between new and legacy memory default settings. #22161
• Logs memory stack traces to help diagnose OOM errors, with configurable throttling. #21396
• Upgrades non-FIPS OpenSSL to 3.0.15 and removes CentOS 7 builds. #24436
• Prevents overloading by correctly categorizing bootstrapping tablets in load balancing. #23487
• Deprecates unused flags to reduce confusion and streamline configurations. #20906
• Ensures debug builds can restart from release builds for easier debugging. #21093
• Streamlines xCluster management by relocating specific table and stream mappings to XClusterSourceManager. #21325
• Enables faster backward scans in DocDB with new flags for optimization. #19352
• Enhances error messages for unset preview flags in allowed_preview_flags_csv. #21484
• Ensures GetChangesForXCluster does not return invalid safe times. #21528
• Adjusts TServer memory recommendations and adds new defaults based on node RAM size. #20664
• Adds usearch and fp16 libraries for enhanced vector indexing. #21830
• Ensures correct error messages are shown when creating tables or tablegroups. #21760
• Enables replication of DDL schemas and user settings across clusters. #21848
• Ensures early, clear error messages on timeouts by introducing a flag are_nodes_safe_to_take_down_timeout_buffer_ms. #21855
• Removes the "No active snapshot" warning log to streamline debugging. #21911
• Ensures DDLs are rerun exactly once by checking replicated_ddls before execution. #21943
• Streamlines xCluster stream creation by centralizing functionality in XClusterClient::CreateXClusterStream<Async>. #22343
• Adds skip_producer_stream_deletion flag to speed up xCluster DR failovers. #22050
• Reduces thread usage by reusing the server messenger for AutoFlags. #22076
• Reduces thread usage by reusing server processes in stateful service clients. #22102
• Enhances validation with SCHECK_PB_FIELDS_NOT_EMPTY to ensure PB fields are non-empty. #22182
• Ensures DROP TABLE operation with atomic DDL waits until the table is fully deleted. #22184
• Enhances extensibility by splitting yb_xcluster_ddl_replication.c into multiple files. #22190
• Removes deprecated xCluster JSON format to streamline functionality. #22219
• Blocks multi-statement DDL queries to prevent replication issues. #22060
• Reduces complexity in catalog_manager by relocating RPC endpoints to MasterClusterHandler. #19715
• Adds read-time option to ysql_dump help for point-in-time database dumps. #21886
• Renames flag to xcluster_consumer_thread_pool_size for clearer usage and efficiency. #20305
• Removes unnecessary locks in DeleteOrHideTabletsAndSendRequests to prevent deadlocks. #22618
• Adds column sorting to stack trace tracking UI for easier data analysis. #22841
• Introduces session-based in-memory table locking for improved concurrency control. #23045
• Uses table IDs instead of names for matching in xCluster DDL replication setup. #23013
• Speeds up backward scans for flat doc reader with packed row V2. #22556
• Reuses TServer yb::client in CDCService to reduce resource usage and fix optimization issues. #22893
• Standardizes xCluster client usage and streamlines client creation in XClusterConsumer. #22908
• Ensures replication setup is successful only if all components are healthy. #22948
• Displays external_hybrid_time in log-dump to detect xCluster target writes. #22918
• Adds SOURCE_UNREACHABLE and SYSTEM_ERROR enums to xCluster for better error reporting. #22996
• Allows requesting xCluster streams by producer table IDs for accurate replication setup. #23013
• Adds easier access and tracking controls to stack trace endpoints. #22842
• Renamed and relocated xCluster components to streamline management and functionality. #23044
• Removes deprecated namespace replication code, enhancing system performance. #23046
• Enhances pggate to recognize fast backward scan capabilities for better query cost estimation. #22937
• Ensures debug builds can restart from release builds by making column ID representations consistent. #21093
• Enhances flags UI by hiding invalid flags and empty categories. #23308
• Enables storing and loading vector index updates in DocDB. #23377
• Streamlines statistics management in IntentAwareIterator. #23420
• Speeds up backward scans by caching keys, enabling quicker and more efficient data retrieval. #22372
• Allows configuring tablet splits based on size ratios with FLAGS_tablet_split_min_size_ratio. #21458
• Simplifies tracking heartbeat processes by relocating code to master_heartbeat_service.cc. #21899, #19715
• Simplifies clone state management by moving retry data to in-memory structure. #22138
• Streamlines heartbeat code management by relocating it to master_heartbeat_service.cc. #21899, #19715
• Enhances code readability by moving heartbeat code to master_heartbeat_service.cc. #21899, #19715
• Moves tcmalloc profiling code to the util directory for better organization. #22258
• Introduces enable_rwc_lock_debugging flag to control RWCLock debug logging. #22807
• Switches CloneStateInfo object handling to std::shared_ptr for future flexibility. #23036
• Ensures schema names are not empty in YBTableName settings. #23371
• Simplifies management by shifting certain managers from Catalog Manager to Master. #22603, #19715
• Enhances visibility of the hidden state in Master/Tserver UI tables. #22521
• Reduces unnecessary reads during CQL INSERT operations with range keys. #23330
• Adds pg_yb_logical_client_version table to manage session configurations. #23871
• Adds an ABORTED state for failed clones in YugabyteDB. #21054

CDC

• Ensures CDC streams are not deleted when all associated tables are dropped. #21419
• Adds new yb-admin commands to manage CDC streams, including removing tables. #22876, #22773
• Prevents tables with enum array columns from being added to CDC streams. #22897
• yb_enable_cdc_consistent_snapshot_streams flag is now default true and automatic. #22984
• Sets retention barriers on tablets during table creation for PG replication. #21643
• Allows changing and persisting the cdcsdk_publication_list_refresh_interval_secs flag value for safer LSN handling. #21796
• Allows configuring default replica identity for tables using the ysql_yb_default_replica_identity flag. #22326
• Enables toggling dynamic table addition in virtual WAL with cdcsdk_enable_dynamic_table_addition flag. #22406
• Uses replication slot names to differentiate between consumption models. #22810
• Supports creation of old model streams via yb-admin in upgraded environments. #22894
• Allows setting per-table replica identity for CDC streams using the ysql_yb_enable_replica_identity flag. #21314
• Suppresses NOTICE messages when setting yb_read_time from walsender. #22379
• Speeds up tablet bootstrap under CDC by skipping initial WAL segments, controlled by skip_wal_replay_from_beginning_with_cdc flag. #24516

yugabyted

• Ensures RENAME DATABASE accurately follows connection rules. #21284
• Tracks role changes accurately by using role_oid instead of role. #21505
• Enables tracking of single-use YSQL configuration parameters in Connection Manager. #21757
• Enhances session consistency by using role_oid for session_authorization in Connection Manager. #21637
• Enables sticky connections when setting specific YSQL configuration parameters. #22957
• Enhances stability by destroying control connections during configuration reloads. #21516
• Enables better handling of complex YSQL configuration parameters using sticky connections. #21637
• Adds a new /pitr endpoint and UI screen for listing scheduled PITRs. #21355
• Adds upgrade_finalize command to yugabyted CLI for post-upgrade tasks. #21888
• Enables flags directly on yb-master and yb-tserver for better setting accuracy. #22061
• Alerts users about version mismatches between nodes via the UI. #21888
• Deprecates Python 2 in yugabyted and shifts to Python 3. #22072, #21409
• Enables scalable parsing of multi-valued flags in yugabyted without code duplication. #22091
• Corrects CPU usage calculation in the Sankey diagram. #22125
• Ensures startup parameters with spaces parse correctly with YSQL Connection Manager. #22248
• Allows specifying multiple data directories using additional_data_dir flag. #22126
• Enables management of xCluster replication using yugabyted commands. #22349
• Adds permanent_uuid to :7000/api/v1/tablet-servers for more reliable UUID retrieval. #22532
• Ensures accurate CPU usage metrics in the YCQL table by updating prev_ticks_ each iteration. #22910
• Prevents errors during node restart when data_dir is missing from the configuration. #23052
• collect_logs now works even if yugabyted is not running. #23210

Bug fixes

YSQL

• Fixes error for index with bool NULLS LAST in YSQL. #22121
• Fixes YSQL upgrade exception in single connection mode. #22283
• Allows YB Admins without superuser status to run pg_locks. #23266
• Ensures safe upgrades by verifying pggate compatibility before sending Scanned Rows. #21229
• Disables in-place index updates by default to address index inconsistency issues. #24672
• Fixes out-of-bounds memory error in HashAggregate with duplicate hash keys. #21122
• Eliminates memory leaks in the YSQL cache system for improved database stability. #22262
• Fixes memory leak affecting catalog cache refresh during upgrades. #22262
• Fixes memory leaks in CacheMemoryContext to stabilize cache memory size after refresh. #22262
• Documents the limitation in retry logic for multi-statement queries in ysqlsh. #21804
• Prevents YSQL DDL operations from returning control until related rollbacks complete. #20033
• Enhances index scan efficiency by modeling remote index filters in cost calculations. #20635
• Fixes issues with colocation in table rewrites and partitions. #20302, #20914
• Renames ddl_rollback_enabled to yb_ddl_rollback_enabled for clarity. #21480
• Adds network latency to startup costs and fixes bug for small table costs. #20898
• Disables bitmap scans by default to prevent regressions with CBO. #21479
• Reduces unnecessary logging when catalog_version_table_in_perdb_mode is set to true. #21481
• Correctly identifies unbatchable filters in indexpath creation. #21292
• Enables control of optimizer statistics via the ysql_yb_enable_optimizer_statistics flag. #21650
• Fixes buffer overflow during ALTER TABLE SET TABLESPACE operation. #21655
• Resolves deadlock issues when deleting a table and its index concurrently in YB-Master. #21663
• Ensures consistent query results with JSONB columns by fixing recheck criteria in index scans. #21451
• Fixes query accuracy by ensuring RHS values fit the LHS data type in index conditions. #21758
• Blocks query layer retries for multi-statement queries to prevent unintended effects. #21361
• Fixes TServer startup to correctly maintain per-database catalog version mode. #21850
• Fixes a bug in index tuple width calculation for cost estimation. #21892
• Ensures RowCompareExpressions correctly enforce non-null column conditions. #21847
• Reduces schema version mismatch errors during consecutive DDL statements with atomicity enabled. #21706
• Enables consistent DDL execution across databases by fixing flag behavior. #21706
• Introduces ysql_min_new_version_ignored_count flag to prevent delays in TServer crashes when versions mismatch. #21776
• Resolves PostgreSQL process crashes on the "show all" command by correcting text description locations for the yb_enable_ddl_atomicity_infra parameter. #21947
• Fixes restarts of UPDATE ...RETURNING queries in debug builds. #22010
• Reverts naming from "Storage SQL" to "Remote SQL" in postgres_fdw.c. #22070
• Ensures ALTER TABLE operates correctly after rules are dropped with old rewrite off. #22064
• Ensures pg_locks displays correct table OID instead of relfilenode after table rewrites. #22081
• Resolves flag clashes by setting YB_AT_REWRITE_ALTER_PRIMARY_KEY to 0x8000. #22086
• Fixes PostgreSQL crashes triggered by enabling yb_debug_log_catcache_events. #22139
• Enables backward parallel scans in YSQL for more flexible data querying. #21633
• Stops unnecessary CPU use and log warnings when YSQL is disabled. #22213
• Corrects log messages for successfully dropped columns to indicate "roll forward" instead of "rollback". #22243
• Enhances stability by modifying batch operation constraints for indexed queries. #21954
• Reduces crashes and allocation errors in specific join operations. #21878
• Reduces errors by giving BNL hashtable its own expression context. #21266
• Re-enables rechecking for RowCompareExpressions to enhance accuracy with NULL inputs. #22075
• Corrects style declaration of YbDdlRollbackEnabled to eliminate GCC errors. #22334
• Fixes an issue with invalid pointer usage in relation data loading. #22342
• Enhances yb_get_range_split_clause to handle decoding errors gracefully, preventing backup issues. #22356
• Fixes memory leaks in ybcFetchNextHeapTuple by properly freeing YBCStatus. #22396
• Stops the YSQL webserver properly on SIGTERM to prevent coredumps. #18948
• Adds warnings and errors for system catalog table misuse. #22520
• Fixes primary key inheritance when attaching partitions with ALTER TABLE. #22562
• Prevents OID collision by excluding OID 65535 for database allocation. #22598
• Skips relfilenode checks for parent partition tables in schema verification. #22625
• Enhances accuracy in backward prefix-based scans by removing kGroupEnd markers. #22615
• Removes an unused function to reduce confusion. #22649
• Ensures bitmap scans correctly recheck results using all necessary columns, preventing incorrect query results. #22622
• Corrects handling of IN expressions on single columns to prevent incorrect tuple path execution. #22704
• Corrects data type error in tuple range checks during scans. #22800
• Ensures creating shared relations during YSQL upgrade increments catalog version globally. #22830
• Adds a function to reset statistics collected by the ANALYZE command. #22028
• Fixes a bug to prevent crashes and errors by ensuring correct column numbers are used for recheck. #22832
• Fixes a deadlock issue during DDL operations by adjusting lock handling. #22882
• Eliminates false log errors in DDL operations by ensuring callback invocation correctness. #22882
• Adjusts handling of INCLUDE columns in index scans to ensure proper sorting. #22822
• Enhances the EXPLAIN command to accurately report subquery distinct index scans. #22923
• Enhances sequence cache uniqueness by using database and sequence OIDs together. #22935
• Fixes crash in Values Scan by repairing SubPlans in multi-row VALUES lists. #22967
• Enables faster inserts into tables with identity columns by optimizing single-shard transactions. #22837
• Fixes log formatting and memory deallocation in yb_pclose_check. #23057
• Resumes drop column operations if alterations miss the initial completion step. #23100
• Eliminates unnecessary wait for concurrent transaction completion in DEFERRABLE READ ONLY transactions. #23120
• Prevents TServer crashes by altering type lookup for certain SAOPs. #23287
• Blocks REINDEX and TRUNCATE on system tables using table rewrite. #24356
• Ensures parent partitioned tables correctly rewrite during operations, preventing backup/restore errors. #24458
• Assumes indexes contain 1000 rows when statistics are unavailable, ensuring better query planning. #24496
• Fixes foreign key checks to avoid errors with unique indexes. #24663
• Adds error prevention for pushed-down expressions that access the catalog cache. #23401
• Increases YSQL connection manager's thread stack size to 512 KB to prevent crashes on Alma 9 machines. #24979
• Fixes memory leaks in YSQL Connection Manager for improved stability. #10065
• Ensures xCluster YBClients connect using specified addresses by setting skip_master_flagfile. #23145
• Disables index update optimization by default to fix ON CONFLICT UPDATE issues. #25075

YCQL

• Stops automatic regeneration of deleted cassandra role on cluster restart. #21057

DocDB

• Enhances read latency measurement during remote bootstrap with compression adjustments. #20848
• Prevents data corruption by handling blind writes in heartbeat processing. #21836
• Avoids crashes from parsing invalid timestamps in LTO builds by updating exception handling. #22191
• Eliminates unnecessary flush during snapshot deletion, preventing write blocks. #22369
• Resolves deadlocks in xCluster setup by standardizing lock order. #22376
• Ensures scans honor set timeouts to avoid indefinite read operations. #21829
• Enables restores to times just before the oldest snapshot in a schedule. #21269
• Ensures full xCluster error info is sent to masters post-failover. #22624
• Adds validation to prevent deletion of snapshots used in restores and vice versa. #23055
• Prevents server crashes during table truncation by safely shutting down iterators before releasing locks. #23243
• TServers now reject remote bootstrap requests for previously deleted tablets. #24574
• Sets a default 10MB max size for RocksDB MANIFEST files, now configurable via flags. #11117
• Ensures creating an index fails if it can't be added to transactional replication. #21732
• Enables ysql_skip_row_lock_for_update as an auto-flag to ensure safe upgrades. #22057
• Eliminates inconsistencies by not including inactive tablets in OpenTable operations. #24860
• Ensures consistent memory usage estimation for write operations at bootstrap and runtime. #21254
• Now truncates error messages over 1KB in AsyncRpc::Failed to prevent crashes. #21402
• Ensures hidden tables and tablets are excluded from snapshot generation to prevent cloning issues. #21631
• Prevents crashes during leader changes by clearing old leader data properly. #21808
• Renames flag min_secustomerent_size_to_rollover_at_flush to specify units. #21691
• Adjusts the class of the enable_automatic_tablet_splitting flag to ensure xCluster compatibility. #22088
• Switches from scoped_refptr to std::shared_ptr to safely manage memory in TabletInfo, preventing use-after-free errors. #18257, #21139
• Ensures cotable IDs match in snapshots and restores for colocated tablets. #23047
• Ensures replication consistency for new tables in colocated databases after T-server restarts. #23188
• Switches pg_cron back to using standard SIGTERM handling for cleaner shutdowns. #24658
• Ensures consistent tablegroup assignments for colocated table backups without using explicit tablespaces. #24809
• Eliminates fatal errors during tabletPB creation by ensuring atomic operations. #21340
• Prevents race conditions during WAL closure and consensus destruction. #21564
• Adds TSAN suppression for boost::regex_match to eliminate false positives. #21585
• Restores compilation on almalinux8 fastdebug gcc11. #21536
• Fixes metric attribute collisions and ensures correct metric types for tablets. #21608
• Avoids segmentation faults in yb-master by checking for null pointers before use. #21648
• Reduces logging redundancy during checkpoint operations to VLOG(1). #21658
• Blocks DML on databases exclusively under STANDBY xCluster mode. #21245
• Extends the lifecycle of the session in CDC State Table callback to prevent crashes. #21103
• Prevents duplicate connection destruction attempts during concurrent failures. #21738
• Enables viewing RPC bind addresses in the master leader UI. #21959
• Prevents node crash during WAL rotation if the log is closed. #21769
• Prevents crashes by ensuring background tasks aren't deleted prematurely during callbacks. #21773
• Ensures graceful shutdown by waiting for all callback completions in PollTransactionStatusBase. #21773
• Fixes compilation error on AlmaLinux 8 with gcc11 by initializing prev_op. #21811
• Fixes an incorrect TSAN suppression path for boost::regex_match. #21585
• Speeds up load balancer actions by allowing tablet additions on TS with pending deletes. #21806
• Enhances stability by handling RPC shutdowns and task aborts more effectively. #21917
• Prevents crashes during remote bootstrap with non-leader peers by enhancing session handling. #22007
• Delays min_running_ht initialization until after tablet bootstrap completes. #22099
• Ensures skip_table_tombstone_check is set for colocated tables. #22115
• Prevents crashes during catalog reloads by safely handling null cluster_config_. #21775
• Reduces the interval for tablet server metrics heartbeat to prevent misreporting leaderless tablets. #22189
• Reduces resource usage by reusing server yb_client in xCluster Consumer. #22845
• Fixes pg_locks to handle lock queries without specified transactions correctly. #22181
• Clarifies flag descriptions for memory division as a percentage of the hard memory limit. #22423
• Restores functionality of CI builds on GCC 12, debug AlmaLinux 9. #22501
• Ensures xCluster DDL replication captures the actual user's role, not the superuser. #22514
• Enhances logging for MemTable flushes to track memory limits more effectively. #22737
• Prevents corrupted characters in Prometheus metric descriptions during scrapes. #22767
• Fixes column drop cascading when dependent objects are dropped. #22874
• Enhances YSQL version upgrade process by refining master branch coding. #23079
• Adds flag to disable intent filtering during bootstrap to prevent data corruption. #23184
• Allows large byte requests on RateLimiter by breaking them into smaller chunks. #23173
• Restores the missing home icon in the master UI navigation bar. #23275
• Disables the read-time option in ysql_dump help to ensure correct DDL handling. #23299
• Enables yb-admin to correctly accept snapshot retention settings. #23332
• Resolves a build failure by removing unreachable code. #23375
• Renaming YSQL migration script from V60 to V59.1 ensures proper versioning and script ordering. #24618
• Fixes load balancer logic to correctly handle leader stepdown tasks. #21834
• Enhances system thread management for faster performance. #21929
• Deprecates the enable_pg_savepoints flag to prevent incorrect behaviors and enhance transaction reliability. #29018
• Prevents "schema version mismatch" errors after using DROP INDEX. #22637
• Ensures hidden split parent tablets are correctly represented as leaderless. #21371
• Ensures xCluster setup fails if the stream state update to ACTIVE fails. #22601
• Returns an empty response from pg_replication_slots when ysql_yb_enable_replication_commands is false. #23096
• Eliminates unnecessary warning logs about cluster configuration versions. #23119
• Prevents FATAL errors after multiple splits and PITR restores. #25103

CDC

• Prevents memory leaks in the GetChanges RPC by deleting its MemoryContext. #22328
• Adds more debug logs to ListReplicationSlots for better debugging of CDC state issues. #21652
• Prevents new indexes and materialized views from being added to existing CDC streams. #22808
• Prevents non-eligible tables from being added to CDC streams, reducing resource usage. #22876, #22835, #22773
• Introduces new flag cdcsdk_enable_identification_of_non_eligible_tables for automatic identification of non-eligible tables in CDC streams, eliminating the need for manual master restarts for cleanup control. #22876, #22835, #22773
• Ensures accurate server clock timestamps in logical replication XLogData. #22929
• Enhances accuracy of transaction start time logging for better consistency in CDC. #24521
• Fixes memory leaks in the walsender's handling of transactions and file nodes. #24788
• Enhances memory management in walsender by implementing deep freeing of record batches. #21530
• Adds debug logs in walsender for better issue investigation. #21465
• Reduces memory leaks in CDC streaming by managing record batches in a separate memory context. #21530
• Logs errors as warnings when cleaning up virtual WAL after logical replication ends. #21651
• Fixes directory creation for serialized transactions during replication. #21765
• Disables loading replication slots from disk on startup to prevent crashes. #21841
• Adds enhanced debug logs for easier CDC stress run troubleshooting. #21780
• Reduces unnecessary RPC calls for non-distributed transaction rollbacks. #21519
• Enhances restart LSN calculation accuracy by updating transaction tracking logic. #21950
• Removes table_name and table_id attributes from CDCSDK metrics to prevent TServer crashes. #22142
• Fixes segmentation faults in walsender by updating replica identities for dynamically added tables. #22273
• Ensures omitted values in logical replication are preserved by serializing the yb_is_omitted array. #21946
• Fixes premature tablet exclusion in CDC streams based on timeout settings. #22383
• Fixes a race condition in CDC dynamic table creation by checking tablet initialization. #22408
• Fixes incorrect removal of the BEGIN record in CDC streams when no DML changes occur. #21646
• Eliminates startup issues by always using the record's commit time for Walsender operations. #22398
• Adds support for CDC with dynamic OID data types like hstore and enum arrays. #23179
• Clears cached schema on re-fetch to ensure correct DDL records delivery. #20698