TiDB 2.0 RC3 Release Notes

[zhangyanqing]

On March 23, 2018, TiDB 2.0 RC3 is released. This release has great improvement in MySQL compatibility, SQL optimization and stability.

TiDB:

• Fix the wrong result issue of MAX/MIN in some scenarios
• Fix the issue that the result of Sort Merge Join does not show in order of Join Key in some scenarios
• Fix the error of comparison between uint and int in boundary conditions
• Optimize checks on length and precision of the floating point type, to improve compatibility with MySQL
• Improve the parsing error log of time type and add more error information
• Improve memory control and add statistics about IndexLookupExecutor memory
• Optimize the execution speed of ADD INDEX to greatly increase the speed in some scenarios
• Use the Stream Aggregation operator when the GROUP BY substatement is empty, to increase the speed
• Support closing the Join Reorder optimization in the optimizer using STRAIGHT_JOIN
• Output more detailed status information of DDL jobs in ADMIN SHOW DDL JOBS
• Support querying the original statements of currently running DDL jobs using ADMIN SHOW DDL JOB QUERIES
• Support recovering the index data using ADMIN RECOVER INDEX for disaster recovery
• Attach a lower priority to the ADD INDEX operation to reduce the impact on online business
• Support aggregation functions with JSON type parameters, such as SUM/AVG
• Support modifying the lower_case_table_names system variable in the configuration file, to support the OGG data synchronization tool
• Improve compatibility with the Navicat management tool
• Support using implicit RowID in CRUD operations

PD:

• Support Region Merge, to merge empty Regions or small Regions after deleting data
• Ignore the nodes that have a lot of pending peers during adding replicas, to improve the speed of restoring replicas or making nodes offline
• Fix the frequent scheduling issue caused by a large number of empty Regions
• Optimize the scheduling speed of leader balance in scenarios of unbalanced resources within different labels
• Add more statistics about abnormal Regions

TiKV:

• Support Region Merge
• Inform PD immediately once the Raft snapshot process is completed, to speed up balancing
• Add the Raw DeleteRange API
• Add the GetMetric API
• Reduce the I/O fluctuation caused by RocksDB sync files
• Optimize the space reclaiming mechanism after deleting data
• Improve the data recovery tool tikv-ctl
• Fix the issue that it is slow to make nodes down caused by snapshot
• Support streaming in Coprocessor
• Support Readpool and increase the raw_get/get/batch_get by 30%
• Support configuring the request timeout of Coprocessor
• Support streaming aggregation in Coprocessor
• Carry time information in Region heartbeats
• Limit the space usage of snapshot files to avoid consuming too much disk space
• Record and report the Regions that cannot elect a leader for a long time
• Speed up garbage cleaning when starting the server
• Update the size information about the corresponding Region according to compaction events
• Limit the size of scan lock to avoid request timeout
• Use DeleteRange to speed up Region deletion
• Support modifying RocksDB parameters online