TiDB 2.1 RC1 Release Notes

On August 24, 2018, TiDB 2.1 RC1 is released! Compared with TiDB 2.1 Beta, this release has great improvement in stability, SQL optimizer, statistics information, and execution engine.

TiDB

  • SQL Optimizer
    • Fix the issue that a wrong result is returned after the correlated subquery is decorrelated in some cases #6972
    • Optimize the output result of Explain #7011#7041
    • Optimize the choosing strategy of the outer table for IndexJoin #7019
    • Remove the Plan Cache of the non-PREPARE statement #7040
    • Fix the issue that the INSERT statement is not parsed and executed correctly in some cases #7068
    • Fix the issue that the IndexJoin result is not correct in some cases #7150
    • Fix the issue that the NULL value cannot be found using the unique index in some cases #7163
    • Fix the range computing issue of the prefix index in UTF-8 #7194
    • Fix the issue that result is not correct caused by eliminating the Project operator in some cases #7257
    • Fix the issue that USE INDEX(PRIMARY) cannot be used when the primary key is an integer #7316
    • Fix the issue that the index range cannot be computed using the correlated column in some cases #7357
  • SQL Execution Engine
    • Fix the issue that the daylight saving time is not computed correctly in some cases #6823
    • Refactor the aggregation function framework to improve the execution efficiency of the Stream and Hash aggregation operators #6852
    • Fix the issue that the Hash aggregation operator cannot exit normally in some cases #6982
    • Fix the issue that BIT_AND/BIT_OR/BIT_XOR does not handle the non-integer data correctly #6994
    • Optimize the execution speed of the REPLACE INTO statement and increase the performance nearly 10 times #7027
    • Optimize the memory usage of time type data and decrease the memory usage of the time type data by fifty percent #7043
    • Fix the issue that the returned result is mixed with signed and unsigned integers in the UNION statement is not compatible with MySQL #7112
    • Fix the panic issue caused by the too much memory applied by LPAD/RPAD/TO_BASE64/FROM_BASE64/REPEAT #7171 #7266 #7409 #7431
    • Fix the incorrect result when MergeJoin/IndexJoin handles the NULL value #7255
    • Fix the incorrect result of Outer Join in some cases #7288
    • Improve the error message of Data Truncated to facilitate locating the wrong data and the corresponding field in the table #7401
    • Fix the incorrect result for decimal in some cases #7001 #7113 #7202 #7208
    • Optimize the point select performance #6937
    • Prohibit the isolation level of Read Committed to avoid the underlying problem #7211
    • Fix the incorrect result of LTRIM/RTRIM/TRIM in some cases #7291
    • Fix the issue that the MaxOneRow operator cannot guarantee that the returned result does not exceed one row #7375
    • Divide the Coprocessor requests with too many ranges #7454
  • Statistics
    • Optimize the mechanism of statistics dynamic collection #6796
    • Fix the issue that Auto Analyze does not work when data is updated frequently #7022
    • Decrease the Write conflicts during the statistics dynamic update process #7124
    • Optimize the cost estimation when the statistics is incorrect #7175
    • Optimize the AccessPath cost estimation strategy #7233
  • Server
    • Fix the bug in loading privilege information #6976
    • Fix the issue that the Kill command is too strict with privilege check #6954
    • Fix the issue of removing some binary numeric types #6922
    • Shorten the output log #7029
    • Handle the mismatchClusterID issue #7053
    • Add the advertise-address configuration item #7078
    • Add the GrpcKeepAlive option #7100
    • Add the connection or Token time monitor #7110
    • Optimize the data decoding performance #7149
    • Add the PROCESSLIST table in INFORMMATION_SCHEMA #7236
    • Fix the order issue when multiple rules are hit in verifying the privilege #7211
    • Change some default values of encoding related system variables to UTF-8 #7198
    • Make the slow query log show more detailed information #7302
    • Support registering tidb-server related information in PD and obtaining this information by HTTP API #7082
  • Compatibility
    • Support Session variables warning_count and error_count #6945
    • Add Scope check when reading the system variables #6958
    • Support the MAX_EXECUTION_TIME syntax #7012
    • Support more statements of the SET syntax #7020
    • Add validity check when setting system variables #7117
    • Add the verification of the number of PlaceHolders in the Prepare statement #7162
    • Support set character_set_results = null #7353
    • Support the flush status syntax #7369
    • Fix the column size of SET and ENUM types in information_schema #7347
    • Support the NATIONAL CHARACTER syntax of statements for creating a table #7378
    • Support the CHARACTER SET syntax in the LOAD DATA statement #7391
    • Fix the column information of the SET and ENUM types #7417
    • Support the IDENTIFIED WITH syntax in the CREATE USER statement #7402
    • Fix the precision losing issue during TIMESTAMP computing process #7418
    • Support the validity verification of more SYSTEM variables #7196
    • Fix the incorrect result when the CHAR_LENGTH function computes the binary string #7410
    • Fix the incorrect CONCAT result in a statement involving GROUP BY #7448
    • Fix the imprecise type length issue when casting the DECIMAL type to the STRING type #7451
  • DML
    • Fix the stability issue of the Load Data statement #6927
    • Fix the memory usage issue when performing some Batch operations #7086
    • Improve the performance of the Replace Into statement #7027
    • Fix the inconsistent precision issue when writing CURRENT_TIMESTAMP #7355
  • DDL
    • Improve the method of DDL judging whether Schema is replicated to avoid misjudgement in some cases #7319
    • Fix the SHOW CREATE TABLE result in adding index process #6993
    • Allow the default value of text/blob/json to be NULL in non-restrict sql-mode #7230
    • Fix the ADD INDEX issue in some cases #7142
    • Increase the speed of adding UNIQUE-KEY index operation largely #7132
    • Fix the truncating issue of the prefix index in UTF-8 character set #7109
    • Add the environment variable tidb_ddl_reorg_priority to control the priority of the add-index operation #7116
    • Fix the display issue of AUTO-INCREMENT in information_schema.tables #7037
    • Support the admin show ddl jobs <number> command and support output specified number of DDL jobs #7028
    • Support parallel DDL job execution #6955
  • Table Partition (Experimental)
    • Support top level partition
    • Support Range Partition

PD

  • Features
    • Introduce the version control mechanism and support rolling update of the cluster with compatibility
    • Enable the region merge feature
    • Support the GetPrevRegion interface
    • Support splitting Regions in batch
    • Support storing the GC safepoint
  • Improvements
    • Optimize the issue that TSO allocation is affected by the system clock going backwards
    • Optimize the performance of handling Region heartbeats
    • Optimize the Region tree performance
    • Optimize the performance of computing hotspot statistics
    • Optimize returning the error code of API interface
    • Add options of controlling scheduling strategies
    • Prohibit using special characters in label
    • Improve the scheduling simulator
    • Support splitting Regions using statistics in pd-ctl
    • Support formatting JSON output by calling jq in pd-ctl
    • Add metrics about etcd Raft state machine
  • Bug fixes
    • Fix the issue that the namespace is not reloaded after switching Leader
    • Fix the issue that namespace scheduling exceeds the schedule limit
    • Fix the issue that hotspot scheduling exceeds the schedule limit
    • Fix the issue that wrong logs are output when the PD client closes
    • Fix the wrong statistics of Region heartbeat latency

TiKV

  • Features
    • Support batch split to avoid too large Regions caused by the Write operation on hot Regions
    • Support splitting Regions based on the number of rows to improve the index scan efficiency
  • Performance
    • Use LocalReader to separate the Read operation from the raftstore thread to lower the Read latency
    • Refactor the MVCC framework, optimize the memory usage and improve the scan Read performance
    • Support splitting Regions based on statistics estimation to reduce the I/O usage
    • Optimize the issue that the Read performance is affected by continuous Write operations on the rollback record
    • Reduce the memory usage of pushdown aggregation computing
  • Improvements
    • Add the pushdown support for a large number of built-in functions and better charset support
    • Optimize the GC workflow, improve the GC speed and decrease the impact of GC on the system
    • Enable prevote to speed up service recovery when the network is abnormal
    • Add the related configuration items of RocksDB log files
    • Adjust the default configuration of scheduler_latch
    • Support setting whether to compact the data in the bottom layer of RocksDB when using tikv-ctl to compact data manually
    • Add the check for environment variables when starting TiKV
    • Support dynamically configuring the dynamic_level_bytes parameter based on the existing data
    • Support customizing the log format
    • Integrate tikv-fail in tikv-ctl
    • Add I/O metrics of threads
  • Bug fixes
    • Fix decimal related issues
    • Fix the issue that gRPC max_send_message_len is set mistakenly
    • Fix the issue caused by misconfiguration of region_size