TiDB 7.1.0 Release Notes

Release date: May 31, 2023

TiDB version: 7.1.0

Quick access: Quick start | Production deployment

TiDB 7.1.0 is a Long-Term Support Release (LTS).

Compared with the previous LTS 6.5.0, 7.1.0 not only includes new features, improvements, and bug fixes released in 6.6.0-DMR, 7.0.0-DMR, but also introduces the following key features and improvements:

Feature details

Performance

• Enhance the Partitioned Raft KV storage engine (experimental) #11515 #12842 @busyjay @tonyxuqqi @tabokie @bufferflies @5kbpers @SpadeA-Tang @nolouch

  TiDB v6.6.0 introduces the Partitioned Raft KV storage engine as an experimental feature, which uses multiple RocksDB instances to store TiKV Region data, and the data of each Region is independently stored in a separate RocksDB instance. The new storage engine can better control the number and level of files in the RocksDB instance, achieve physical isolation of data operations between Regions, and support stably managing more data. Compared with the original TiKV storage engine, using the Partitioned Raft KV storage engine can achieve about twice the write throughput and reduce the elastic scaling time by about 4/5 under the same hardware conditions and mixed read and write scenarios.

  In TiDB v7.1.0, the Partitioned Raft KV storage engine supports tools such as TiDB Lightning, BR, and TiCDC.

  Currently, this feature is experimental and not recommended for use in production environments. You can only use this engine in a newly created cluster and you cannot directly upgrade from the original TiKV storage engine.

  For more information, see documentation.

• TiFlash supports late materialization (GA) #5829 @Lloyd-Pottiger

  In v7.0.0, late materialization was introduced in TiFlash as an experimental feature for optimizing query performance. This feature is disabled by default (the tidb_opt_enable_late_materialization system variable defaults to OFF). When processing a SELECT statement with filter conditions (WHERE clause), TiFlash reads all the data from the columns required by the query, and then filters and aggregates the data based on the query conditions. When Late materialization is enabled, TiDB supports pushing down part of the filter conditions to the TableScan operator. That is, TiFlash first scans the column data related to the filter conditions that are pushed down to the TableScan operator, filters the rows that meet the condition, and then scans the other column data of these rows for further calculation, thereby reducing IO scans and computations of data processing.

  Starting from v7.1.0, the TiFlash late materialization feature is generally available and enabled by default (the tidb_opt_enable_late_materialization system variable defaults to ON). The TiDB optimizer decides which filters to be pushed down to the TableScan operator based on the statistics and the filter conditions of the query.

  For more information, see documentation.

• TiFlash supports automatically choosing an MPP Join algorithm according to the overhead of network transmission #7084 @solotzg

  The TiFlash MPP mode supports multiple Join algorithms. Before v7.1.0, TiDB determines whether the MPP mode uses the Broadcast Hash Join algorithm based on the tidb_broadcast_join_threshold_count and tidb_broadcast_join_threshold_size variables and the actual data volume.

  In v7.1.0, TiDB introduces the tidb_prefer_broadcast_join_by_exchange_data_size variable, which controls whether to choose the MPP Join algorithm based on the minimum overhead of network transmission. This variable is disabled by default, indicating that the default algorithm selection method remains the same as that before v7.1.0. You can set the variable to ON to enable it. When it is enabled, you no longer need to manually adjust the tidb_broadcast_join_threshold_count and tidb_broadcast_join_threshold_size variables (both variables does not take effect at this time), TiDB automatically estimates the threshold of network transmission by different Join algorithms, and then chooses the algorithm with the smallest overhead overall, thus reducing network traffic and improving MPP query performance.

  For more information, see documentation.

• Support load-based replica read to mitigate read hotspots #14151 @sticnarf @you06

  In a read hotspot scenario, the hotspot TiKV node cannot process read requests in time, resulting in the read requests queuing. However, not all TiKV resources are exhausted at this time. To reduce latency, TiDB v7.1.0 introduces the load-based replica read feature, which allows TiDB to read data from other TiKV nodes without queuing on the hotspot TiKV node. You can control the queue length of read requests using the tidb_load_based_replica_read_threshold system variable. When the estimated queue time of the leader node exceeds this threshold, TiDB prioritizes reading data from follower nodes. This feature can improve read throughput by 70% to 200% in a read hotspot scenario compared to not scattering read hotspots.

  For more information, see documentation.

• Enhance the capability of caching execution plans for non-prepared statements (experimental) #36598 @qw4990

  TiDB v7.0.0 introduces non-prepared plan cache as an experimental feature to improve the load capacity of concurrent OLTP. In v7.1.0, TiDB enhances this feature and supports caching more SQL statements.

  To improve memory utilization, TiDB v7.1.0 merges the cache pools of non-prepared and prepared plan caches. You can control the cache size using the system variable tidb_session_plan_cache_size. The tidb_prepared_plan_cache_size and tidb_non_prepared_plan_cache_size system variables are deprecated.

  To maintain forward compatibility, when you upgrade from an earlier version to v7.1.0 or later versions, the cache size tidb_session_plan_cache_size remains the same value as tidb_prepared_plan_cache_size, and tidb_enable_non_prepared_plan_cache remains the setting before the upgrade. After sufficient performance testing, you can enable non-prepared plan cache using tidb_enable_non_prepared_plan_cache. For a newly created cluster, non-prepared plan cache is enabled by default.

  Non-prepared plan cache does not support DML statements by default. To remove this restriction, you can set the tidb_enable_non_prepared_plan_cache_for_dml system variable to ON.

  For more information, see documentation.

• Support the TiDB Distributed eXecution Framework (DXF) (experimental) #41495 @benjamin2037

  Before TiDB v7.1.0, only one TiDB node can serve as the DDL owner and execute DDL tasks at the same time. Starting from TiDB v7.1.0, in the new DXF, multiple TiDB nodes can execute the same DDL task in parallel, thus better utilizing the resources of the TiDB cluster and significantly improving the performance of DDL. In addition, you can linearly improve the performance of DDL by adding more TiDB nodes. Note that this feature is currently experimental and only supports ADD INDEX operations.

  To use the DXF, set the value of tidb_enable_dist_task to ON:

  SET GLOBAL tidb_enable_dist_task = ON;
  

  For more information, see documentation.

Reliability

• Resource Control becomes generally available (GA) #38825 @nolouch @BornChanger @glorv @tiancaiamao @Connor1996 @JmPotato @hnes @CabinfeverB @HuSharp

  TiDB enhances the resource control feature based on resource groups, which becomes GA in v7.1.0. This feature significantly improves the resource utilization efficiency and performance of TiDB clusters. The introduction of the resource control feature is a milestone for TiDB. You can divide a distributed database cluster into multiple logical units, map different database users to corresponding resource groups, and set the quota for each resource group as needed. When the cluster resources are limited, all resources used by sessions in the same resource group are limited to the quota. In this way, even if a resource group is over-consumed, the sessions in other resource groups are not affected.

  With this feature, you can combine multiple small and medium-sized applications from different systems into a single TiDB cluster. When the workload of an application grows larger, it does not affect the normal operation of other applications. When the system workload is low, busy applications can still be allocated the required system resources even if they exceed the set quotas, which can achieve the maximum utilization of resources. In addition, the rational use of the resource control feature can reduce the number of clusters, ease the difficulty of operation and maintenance, and save management costs.

  In TiDB v7.1.0, this feature introduces the ability to estimate system capacity based on actual workload or hardware deployment. The estimation ability provides you with a more accurate reference for capacity planning and assists you in better managing TiDB resource allocation to meet the stability needs of enterprise-level scenarios.

  To improve user experience, TiDB Dashboard provides the Resource Manager page. You can view the resource group configuration on this page and estimate cluster capacity in a visual way to facilitate reasonable resource allocation.

  For more information, see documentation.

• Support the checkpoint mechanism for Fast Online DDL to improve fault tolerance and automatic recovery capability #42164 @tangenta

  TiDB v7.1.0 introduces a checkpoint mechanism for Fast Online DDL, which significantly improves the fault tolerance and automatic recovery capability of Fast Online DDL. Even if the TiDB owner node is restarted or changed due to failures, TiDB can still recover progress from checkpoints that are automatically updated on a regular basis, making the DDL execution more stable and efficient.

  For more information, see documentation.

• Backup & Restore supports checkpoint restore #42339 @Leavrth

  Snapshot restore or log restore might be interrupted due to recoverable errors, such as disk exhaustion and node crash. Before TiDB v7.1.0, the recovery progress before the interruption would be invalidated even after the error is addressed, and you need to start the restore from scratch. For large clusters, this incurs considerable extra cost.

  Starting from TiDB v7.1.0, Backup & Restore (BR) introduces the checkpoint restore feature, which enables you to continue an interrupted restore. This feature can retain most recovery progress of the interrupted restore.

  For more information, see documentation.

• Optimize the strategy of loading statistics #42160 @xuyifangreeneyes

  TiDB v7.1.0 introduces lightweight statistics initialization as an experimental feature. Lightweight statistics initialization can significantly reduce the number of statistics that must be loaded during startup, thus improving the speed of loading statistics. This feature increases the stability of TiDB in complex runtime environments and reduces the impact on the overall service when TiDB nodes restart. You can set the parameter lite-init-stats to true to enable this feature.

  During TiDB startup, SQL statements executed before the initial statistics are fully loaded might have suboptimal execution plans, thus causing performance issues. To avoid such issues, TiDB v7.1.0 introduces the configuration parameter force-init-stats. With this option, you can control whether TiDB provides services only after statistics initialization has been finished during startup. This parameter is disabled by default.

  For more information, see documentation.

• TiCDC supports the data integrity validation feature for single-row data #8718 #42747 @3AceShowHand @zyguan

  Starting from v7.1.0, TiCDC introduces the data integrity validation feature, which uses a checksum algorithm to validate the integrity of single-row data. This feature helps verify whether any error occurs in the process of writing data from TiDB, replicating it through TiCDC, and then writing it to a Kafka cluster. The data integrity validation feature only supports changefeeds that use Kafka as the downstream and currently supports the Avro protocol.

  For more information, see documentation.

• TiCDC optimizes DDL replication operations #8686 @hi-rustin

  Before v7.1.0, when you perform a DDL operation that affects all rows on a large table (such as adding or deleting a column), the replication latency of TiCDC would significantly increase. Starting from v7.1.0, TiCDC optimizes this replication operation and mitigates the impact of DDL operations on downstream latency.

  For more information, see documentation.

• Improve the stability of TiDB Lightning when importing TiB-level data #43510 #43657 @D3Hunter @lance6716

  Starting from v7.1.0, TiDB Lightning has added four configuration items to improve stability when importing TiB-level data.

  • tikv-importer.region-split-batch-size controls the number of Regions when splitting Regions in a batch. The default value is 4096.
  • tikv-importer.region-split-concurrency controls the concurrency when splitting Regions. The default value is the number of CPU cores.
  • tikv-importer.region-check-backoff-limit controls the number of retries to wait for the Region to come online after the split and scatter operations. The default value is 1800 and the maximum retry interval is two seconds. The number of retries is not increased if any Region becomes online between retries.
  • tikv-importer.pause-pd-scheduler-scope controls the scope in which TiDB Lightning pauses PD scheduling. Value options are "table" and "global". The default value is "table". For TiDB versions earlier than v6.1.0, you can only configure the "global" option, which pauses global scheduling during data import. Starting from v6.1.0, the "table" option is supported, which means that scheduling is only paused for the Region that stores the target table data. It is recommended to set this configuration item to "global" in scenarios with large data volumes to improve stability.

  For more information, see documentation.

SQL

• Support saving TiFlash query results using the INSERT INTO SELECT statement (GA) #37515 @gengliqi

  Starting from v6.5.0, TiDB supports pushing down the SELECT clause (analytical query) of the INSERT INTO SELECT statement to TiFlash. In this way, you can easily save the TiFlash query result to a TiDB table specified by INSERT INTO for further analysis, which takes effect as result caching (that is, result materialization).

  In v7.1.0, this feature is generally available. During the execution of the SELECT clause in the INSERT INTO SELECT statement, the optimizer can intelligently decide whether to push a query down to TiFlash based on the SQL mode and the cost estimates of the TiFlash replica. Therefore, the tidb_enable_tiflash_read_for_write_stmt system variable introduced during the experimental phase is now deprecated. Note that the computation rules of INSERT INTO SELECT statements for TiFlash do not meet the STRICT SQL Mode requirement, so TiDB allows the SELECT clause in the INSERT INTO SELECT statement to be pushed down to TiFlash only when the SQL mode of the current session is not strict, which means that the sql_mode value does not contain STRICT_TRANS_TABLES and STRICT_ALL_TABLES.

  For more information, see documentation.

• MySQL-compatible multi-valued indexes become generally available (GA) #39592 @xiongjiwei @qw4990 @YangKeao

  Filtering the values of an array in a JSON column is a common operation, but normal indexes cannot help speed up such an operation. Creating a multi-valued index on an array can greatly improve filtering performance. If an array in the JSON column has a multi-valued index, you can use the multi-valued index to filter retrieval conditions in MEMBER OF(), JSON_CONTAINS(), and JSON_OVERLAPS() functions, thereby reducing I/O consumption and improving operation speed.

  In v7.1.0, the multi-valued indexes feature becomes generally available (GA). It supports more complete data types and is compatible with TiDB tools. You can use multi-valued indexes to speed up the search operations on JSON arrays in production environments.

  For more information, see documentation.

• Improve the partition management for Hash and Key partitioned tables #42728 @mjonss

  Before v7.1.0, Hash and Key partitioned tables in TiDB only support the TRUNCATE PARTITION partition management statement. Starting from v7.1.0, Hash and Key partitioned tables also support ADD PARTITION and COALESCE PARTITION partition management statements. Therefore, you can flexibly adjust the number of partitions in Hash and Key partitioned tables as needed. For example, you can increase the number of partitions with the ADD PARTITION statement, or decrease the number of partitions with the COALESCE PARTITION statement.

  For more information, see documentation.

• The syntax of Range INTERVAL partitioning becomes generally available (GA) #35683 @mjonss

  The syntax of Range INTERVAL partitioning (introduced in v6.3.0) becomes GA. With this syntax, you can define Range partitioning by a desired interval without enumerating all partitions, which drastically reduces the length of Range partitioning DDL statements. The syntax is equivalent to that of the original Range partitioning.

  For more information, see documentation.

• Generated columns become generally available (GA) @bb7133

  Generated columns are a valuable feature for a database. When creating a table, you can define that the value of a column is calculated based on the values of other columns in the table, rather than being explicitly inserted or updated by users. This generated column can be either a virtual column or a stored column. TiDB has supported MySQL-compatible generated columns since earlier versions, and this feature becomes GA in v7.1.0.

  Using generated columns can improve MySQL compatibility for TiDB, simplifying the process of migrating from MySQL. It also reduces data maintenance complexity and improves data consistency and query efficiency.

  For more information, see documentation.

DB operations

• Support smooth cluster upgrade without manually canceling DDL operations (experimental) #39751 @zimulala

  Before TiDB v7.1.0, to upgrade a cluster, you must manually cancel its running or queued DDL tasks before the upgrade and then add them back after the upgrade.

  To provide a smoother upgrade experience, TiDB v7.1.0 supports automatically pausing and resuming DDL tasks. Starting from v7.1.0, you can upgrade your clusters without manually canceling DDL tasks in advance. TiDB will automatically pause any running or queued user DDL tasks before the upgrade and resume these tasks after the rolling upgrade, making it easier for you to upgrade your TiDB clusters.

  For more information, see documentation.

Observability

• Enhance optimizer diagnostic information #43122 @time-and-fate

  Obtaining sufficient information is the key to SQL performance diagnostics. In v7.1.0, TiDB continues to add optimizer runtime information to various diagnostic tools, providing better insights into how execution plans are selected and assisting in troubleshooting SQL performance issues. The new information includes:

  • debug_trace.json in the output of PLAN REPLAYER.
  • Partial statistics details for operator info in the output of EXPLAIN.
  • Partial statistics details in the Stats field of slow queries.

  For more information, see Use PLAN REPLAYER to save and restore the on-site information of a cluster, EXPLAIN walkthrough, and Identify slow queries.

Security

• Replace the interface used for querying TiFlash system table information #6941 @flowbehappy

  Starting from v7.1.0, when providing the query service of INFORMATION_SCHEMA.TIFLASH_TABLES and INFORMATION_SCHEMA.TIFLASH_SEGMENTS system tables for TiDB, TiFlash uses the gRPC port instead of the HTTP port, which avoids the security risks of the HTTP service.

• Support LDAP authentication #43580 @YangKeao

  Starting from v7.1.0, TiDB supports LDAP authentication and provides two authentication plugins: authentication_ldap_sasl and authentication_ldap_simple.

  For more information, see documentation.

• Enhance the database auditing feature (Enterprise Edition)

  In v7.1.0, TiDB Enterprise Edition enhances the database auditing feature, which significantly expands its capacity and improves the user experience to meet the needs of enterprises for database security compliance:

  • Introduce the concepts of "Filter" and "Rule" for more granular audit event definitions and more fine-grained audit settings.
  • Support defining rules in JSON format, providing a more user-friendly configuration method.
  • Add automatic log rotation and space management functions, and support configuring log rotation in two dimensions: retention time and log size.
  • Support outputting audit logs in both TEXT and JSON formats, facilitating easier integration with third-party tools.
  • Support audit log redaction. You can replace all literals to enhance security.

  Database auditing is an important feature in TiDB Enterprise Edition. This feature provides a powerful monitoring and auditing tool for enterprises to ensure data security and compliance. It can help enterprise managers in tracking the source and impact of database operations to prevent illegal data theft or tampering. Furthermore, database auditing can also help enterprises meet various regulatory and compliance requirements, ensuring legal and ethical compliance. This feature has important application value for enterprise information security.

  For more information, see user guide. This feature is included in TiDB Enterprise Edition. To use this feature, navigate to the TiDB Enterprise page to get TiDB Enterprise Edition.

Compatibility changes

Behavior changes

• To improve security, TiFlash deprecates the HTTP service port (default 8123) and uses the gRPC port as a replacement

  If you have upgraded TiFlash to v7.1.0, then during the TiDB upgrade to v7.1.0, TiDB cannot read the TiFlash system tables (INFORMATION_SCHEMA.TIFLASH_TABLES and INFORMATION_SCHEMA.TIFLASH_SEGMENTS).

• TiDB Lightning in TiDB versions from v6.2.0 to v7.0.0 decides whether to pause global scheduling based on the TiDB cluster version. When TiDB cluster version >= v6.1.0, scheduling is only paused for the Region that stores the target table data and is resumed after the target table import is complete. While for other versions, TiDB Lightning pauses global scheduling. Starting from TiDB v7.1.0, you can control whether to pause global scheduling by configuring pause-pd-scheduler-scope. By default, TiDB Lightning pauses scheduling for the Region that stores the target table data. If the target cluster version is earlier than v6.1.0, an error occurs. In this case, you can change the value of the parameter to "global" and try again.

• When you use FLASHBACK CLUSTER TO TIMESTAMP in TiDB v7.1.0, some Regions might remain in the FLASHBACK process even after the completion of the FLASHBACK operation. It is recommended to avoid using this feature in v7.1.0. For more information, see issue #44292. If you have encountered this issue, you can use the TiDB snapshot backup and restore feature to restore data.

System variables

Configuration file parameters

Improvements

• TiDB

  • Display the number of distinct values for the corresponding column in the Cardinality column of the SHOW INDEX result #42227 @winoros
  • Use SQL_NO_CACHE to prevent TTL Scan queries from impacting the TiKV block cache #43206 @lcwangchao
  • Improve an error message related to MAX_EXECUTION_TIME to make it compatible with MySQL #43031 @dveeden
  • Support using the MergeSort operator on partitioned tables in IndexLookUp #26166 @Defined2014
  • Enhance caching_sha2_password to make it compatible with MySQL #43576 @asjdf

• TiKV

  • Reduce the impact of split operations on write QPS when using partitioned Raft KV #14447 @SpadeA-Tang
  • Optimize the space occupied by snapshots when using partitioned Raft KV #14581 @bufferflies
  • Provide more detailed time information for each stage of processing requests in TiKV #12362 @cfzjywxk
  • Use PD as metastore in log backup #13867 @YuJuncen

• PD

  • Add a controller that automatically adjusts the size of the store limit based on the execution details of the snapshot. To enable this controller, set store-limit-version to v2 (experimental). Once enabled, you do not need to manually adjust the store limit configuration to control the speed of scaling in or scaling out #6147 @bufferflies
  • Add historical load information to avoid frequent scheduling of Regions with unstable loads by the hotspot scheduler when the storage engine is raft-kv2 #6297 @bufferflies
  • Add a leader health check mechanism. When the PD server where the etcd leader is located cannot be elected as the leader, PD actively switches the etcd leader to ensure that the PD leader is available #6403 @nolouch

• TiFlash

  • Improve TiFlash performance and stability in the disaggregated storage and compute architecture #6882 @JaySon-Huang @breezewish @JinheLin
  • Support optimizing query performance in Semi Join or Anti Semi Join by selecting the smaller table as the build side #7280 @yibin87
  • Improve performance of data import from BR and TiDB Lightning to TiFlash with default configurations #7272 @breezewish

• Tools

  • Backup & Restore (BR)

    • Support modifying the TiKV configuration item log-backup.max-flush-interval during log backup #14433 @joccau

  • TiCDC

    • Optimize the directory structure when DDL events occur in the scenario of replicating data to object storage #8890 @CharlesCheung96
    • Optimize the method of setting GC TLS for the upstream when the TiCDC replication task fails #8403 @charleszheng44
    • Support replicating data to the Kafka-on-Pulsar downstream #8892 @hi-rustin
    • Support using the open-protocol protocol to only replicate the changed columns after an update occurs when replicating data to Kafka #8706 @sdojjy
    • Optimize the error handling of TiCDC in the downstream failures or other scenarios #8657 @hicqu
    • Add a configuration item insecure-skip-verify to control whether to set the authentication algorithm in the scenario of enabling TLS #8867 @hi-rustin

  • TiDB Lightning

    • Change the severity level of the precheck item related to uneven Region distribution from Critical to Warn to avoid blocking users from importing data #42836 @okJiang
    • Add a retry mechanism when encountering an unknown RPC error during data import #43291 @D3Hunter
    • Enhance the retry mechanism for Region jobs #43682 @lance6716

Bug fixes

• TiDB

  • Fix the issue that there is no prompt about manually executing ANALYZE TABLE after reorganizing partitions #42183 @CbcWestwolf
  • Fix the issue of missing table names in the ADMIN SHOW DDL JOBS result when a DROP TABLE operation is being executed #42268 @tiancaiamao
  • Fix the issue that Ignore Event Per Minute and Stats Cache LRU Cost charts might not be displayed normally in the Grafana monitoring panel #42562 @pingandb
  • Fix the issue that the ORDINAL_POSITION column returns incorrect results when querying the INFORMATION_SCHEMA.COLUMNS table #43379 @bb7133
  • Fix the case sensitivity issue in some columns of the permission table #41048 @bb7133
  • Fix the issue that after a new column is added in the cache table, the value is NULL instead of the default value of the column #42928 @lqs
  • Fix the issue that CTE results are incorrect when pushing down predicates #43645 @winoros
  • Fix the issue of DDL retry caused by write conflict when executing TRUNCATE TABLE for partitioned tables with many partitions and TiFlash replicas #42940 @mjonss
  • Fix the issue that there is no warning when using SUBPARTITION in creating partitioned tables #41198 #41200 @mjonss
  • Fix the incompatibility issue with MySQL when dealing with value overflow issues in generated columns #40066 @jiyfhust
  • Fix the issue that REORGANIZE PARTITION cannot be concurrently executed with other DDL operations #42442 @bb7133
  • Fix the issue that canceling the partition reorganization task in DDL might cause subsequent DDL operations to fail #42448 @lcwangchao
  • Fix the issue that assertions on delete operations are incorrect under certain conditions #42426 @tiancaiamao
  • Fix the issue that TiDB server cannot start due to an error in reading the cgroup information with the error message "can't read file memory.stat from cgroup v1: open /sys/memory.stat no such file or directory" #42659 @hawkingrei
  • Fix the Duplicate Key issue that occurs when updating the partition key of a row on a partitioned table with a global index #42312 @L-maple
  • Fix the issue that the Scan Worker Time By Phase chart in the TTL monitoring panel does not display data #42515 @lcwangchao
  • Fix the issue that some queries on partitioned tables with a global index return incorrect results #41991 #42065 @L-maple
  • Fix the issue of displaying some error logs during the process of reorganizing a partitioned table #42180 @mjonss
  • Fix the issue that the data length in the QUERY column of the INFORMATION_SCHEMA.DDL_JOBS table might exceed the column definition #42440 @tiancaiamao
  • Fix the issue that the INFORMATION_SCHEMA.CLUSTER_HARDWARE table might display incorrect values in containers #42851 @hawkingrei
  • Fix the issue that an incorrect result is returned when you query a partitioned table using ORDER BY + LIMIT #43158 @Defined2014
  • Fix the issue of multiple DDL tasks running simultaneously using the ingest method #42903 @tangenta
  • Fix the wrong value returned when querying a partitioned table using Limit #24636
  • Fix the issue of displaying the incorrect TiDB address in IPv6 environment #43260 @nexustar
  • Fix the issue of displaying incorrect values for system variables tidb_enable_tiflash_read_for_write_stmt and tidb_enable_exchange_partition #43281 @gengliqi
  • Fix the issue that when tidb_scatter_region is enabled, Region does not automatically split after a partition is truncated #43174 #43028 @jiyfhust
  • Add checks on the tables with generated columns and report errors for unsupported DDL operations on these columns #38988 #24321 @tiancaiamao
  • Fix the issue that the error message is incorrect in certain type conversion errors #41730 @hawkingrei
  • Fix the issue that after a TiDB node is normally shutdown, DDL tasks triggered on this node will be canceled #43854 @zimulala
  • Fix the issue that when the PD member address changes, allocating ID for the AUTO_INCREMENT column will be blocked for a long time #42643 @tiancaiamao
  • Fix the issue of reporting the GC lifetime is shorter than transaction duration error during DDL execution #40074 @tangenta
  • Fix the issue that metadata locks unexpectedly block the DDL execution #43755 @wjhuang2016
  • Fix the issue that the cluster cannot query some system views in IPv6 environment #43286 @Defined2014
  • Fix the issue of not finding the partition during inner join in dynamic pruning mode #43686 @mjonss
  • Fix the issue that TiDB reports syntax errors when analyzing tables #43392 @guo-shaoge
  • Fix the issue that TiCDC might lose some row changes during table renaming #43338 @tangenta
  • Fix the issue that TiDB server crashes when the client uses cursor reads #38116 @YangKeao
  • Fix the issue that ADMIN SHOW DDL JOBS LIMIT returns incorrect results #42298 @CbcWestwolf
  • Fix the TiDB panic issue that occurs when querying union views and temporary tables with UNION #42563 @lcwangchao
  • Fix the issue that renaming tables does not take effect when committing multiple statements in a transaction #39664 @tiancaiamao
  • Fix the incompatibility issue between the behavior of prepared plan cache and non-prepared plan cache during time conversion #42439 @qw4990
  • Fix the wrong results caused by plan cache for Decimal type #43311 @qw4990
  • Fix the TiDB panic issue in null-aware anti join (NAAJ) due to the wrong field type check #42459 @AilinKid
  • Fix the issue that DML execution failures in pessimistic transactions at the RC isolation level might cause inconsistency between data and indexes #43294 @ekexium
  • Fix the issue that in some extreme cases, when the first statement of a pessimistic transaction is retried, resolving locks on this transaction might affect transaction correctness #42937 @MyonKeminta
  • Fix the issue that in some rare cases, residual pessimistic locks of pessimistic transactions might affect data correctness when GC resolves locks #43243 @MyonKeminta
  • Fix the issue that the LOCK to PUT optimization leads to duplicate data being returned in specific queries #28011 @zyguan
  • Fix the issue that when data is changed, the locking behavior of the unique index is not consistent with that when the data is unchanged #36438 @zyguan

• TiKV

  • Fix the issue that when you enable tidb_pessimistic_txn_fair_locking, in some extreme cases, expired requests caused by failed RPC retries might affect data correctness during the resolve lock operation #14551 @MyonKeminta
  • Fix the issue that when you enable tidb_pessimistic_txn_fair_locking, in some extreme cases, expired requests caused by failed RPC retries might cause transaction conflicts to be ignored, thus affecting transaction consistency #14311 @MyonKeminta
  • Fix the issue that encryption key ID conflict might cause the deletion of the old keys #14585 @tabokie
  • Fix the performance degradation issue caused by accumulated lock records when a cluster is upgraded from a previous version to v6.5 or later versions #14780 @MyonKeminta
  • Fix the issue that the raft entry is too large error occurs during the PITR recovery process #14313 @YuJuncen
  • Fix the issue that TiKV panics during the PITR recovery process due to log_batch exceeding 2 GB #13848 @YuJuncen

• PD

  • Fix the issue that the number of low space store in the PD monitoring panel is abnormal after TiKV panics #6252 @HuSharp
  • Fix the issue that Region Health monitoring data is deleted after PD leader switch #6366 @iosmanthus
  • Fix the issue that the rule checker cannot repair unhealthy Regions with the schedule=deny label #6426 @nolouch
  • Fix the issue that some existing labels are lost after TiKV or TiFlash restarts #6467 @JmPotato
  • Fix the issue that the replication status cannot be switched when there are learner nodes in the replication mode #14704 @nolouch

• TiFlash

  • Fix the issue that querying data in the TIMESTAMP or TIME type returns errors after enabling late materialization #7455 @Lloyd-Pottiger
  • Fix the issue that large update transactions might cause TiFlash to repeatedly report errors and restart #7316 @JaySon-Huang

• Tools

  • Backup & Restore (BR)

    • Fix the issue of backup slowdown when a TiKV node crashes in a cluster #42973 @YuJuncen
    • Fix the issue of inaccurate error messages caused by a backup failure in some cases #43236 @YuJuncen

  • TiCDC

    • Fix the issue of TiCDC time zone setting #8798 @hi-rustin
    • Fix the issue that TiCDC cannot automatically recover when PD address or leader fails #8812 #8877 @asddongmen
    • Fix the issue that checkpoint lag increases when one of the upstream TiKV nodes crashes #8858 @hicqu
    • Fix the issue that when replicating data to object storage, the EXCHANGE PARTITION operation in the upstream cannot be properly replicated to the downstream #8914 @CharlesCheung96
    • Fix the OOM issue caused by excessive memory usage of the sorter component in some special scenarios #8974 @hicqu
    • Fix the TiCDC node panic that occurs when the downstream Kafka sinks are rolling restarted #9023 @asddongmen

  • TiDB Data Migration (DM)

    • Fix the issue that latin1 data might be corrupted during replication #7028 @lance6716

  • TiDB Dumpling

    • Fix the issue that the UNSIGNED INTEGER type primary key cannot be used for splitting chunks #42620 @lichunzhu
    • Fix the issue that TiDB Dumpling might panic when --output-file-template is incorrectly set #42391 @lichunzhu

  • TiDB Binlog

    • Fix the issue that an error might occur when encountering a failed DDL statement #1228 @okJiang

  • TiDB Lightning

    • Fix the performance degradation issue during data import #42456 @lance6716
    • Fix the issue of write to tikv with no leader returned when importing a large amount of data #43055 @lance6716
    • Fix the issue of excessive keys within region is empty, skip doIngest logs during data import #43197 @D3Hunter
    • Fix the issue that panic might occur during partial write #43363 @lance6716
    • Fix the issue that OOM might occur when importing a wide table #43728 @D3Hunter
    • Fix the issue of missing data in the TiDB Lightning Grafana dashboard #43357 @lichunzhu
    • Fix the import failure due to incorrect setting of keyspace-name #43684 @zeminzhou
    • Fix the issue that data import might be skipped during range partial write in some cases #43768 @lance6716

Performance test

To learn about the performance of TiDB v7.1.0, you can refer to the TPC-C performance test report and Sysbench performance test report of the TiDB Cloud Dedicated cluster.

Contributors

We would like to thank the following contributors from the TiDB community:

• blacktear23
• ethercflow
• hihihuhu
• jiyfhust
• L-maple
• lqs
• pingandb
• yorkhellen
• yujiarista (First-time contributor)