TiDB 2.1.18 Release Notes

Release date: November 4, 2019

TiDB version: 2.1.18

TiDB Ansible version: 2.1.18

TiDB

• SQL Optimizer
  • Fix the issue that invalid query ranges might appear when split by feedback #12172
  • Fix the issue that the privilege check is incorrect in point get plan #12341
  • Optimize execution performance of the select ... limit ... offset … statement by pushing the Limit operator down to the IndexLookUpReader execution logic #12380
  • Support using parameters in ORDER BY, GROUP BY and LIMIT OFFSET #12514
  • Fix the issue that IndexJoin on the partition table returns incorrect results #12713
  • Fix the issue that the str_to_date function in TiDB returns a different result from MySQL when the date string and the format string do not match #12757
  • Fix the issue that outer join is incorrectly converted to inner join when the cast function is included in the query conditions #12791
  • Fix incorrect expression passing in the join condition of AntiSemiJoin #12800
• SQL Engine
  • Fix the incorrectly rounding of time (for example, 2019-09-11 11:17:47.999999666 should be rounded to 2019-09-11 11:17:48) #12259
  • Fix the issue that the duration by sql_type for the PREPARE statement is not shown in the monitoring record #12329
  • Fix the panic issue when the from_unixtime function handles null #12572
  • Fix the compatibility issue that when an invalid value is inserted as the YEAR type, the result is NULL instead of 0000 #12744
  • Improve the behavior of the AutoIncrement column when it is implicitly allocated, to keep it consistent with the default mode of MySQL auto-increment locking ("consecutive" lock mode): for the implicit allocation of multiple AutoIncrement IDs in a single-line Insert statement, TiDB guarantees the continuity of the allocated values. This improvement ensures that the JDBC getGeneratedKeys() method will get the correct results in any scenario #12619
  • Fix the issue that the query is hanged when HashAgg serves as a child node of Apply #12769
  • Fix the issue that the AND and OR logical expressions return incorrect results when it comes to type conversion #12813
• Server
  • Fix the issue that the SLEEP() function is invalid for the KILL TIDB QUERY statements #12159
  • Fix the issue that no error is reported when AUTO_INCREMENT incorrectly allocates MAX int64 and MAX uint64 #12210
  • Fix the issue that the slow query logs are not recorded when the log level is ERROR #12373
  • Adjust the number of times that TiDB caches schema changes and corresponding changed table information from 100 to 1024, and support modification by using the tidb_max_delta_schema_count system variable #12515
  • Change the query start time from the point of "starting to execute" to "starting to compile" to make SQL statistics more accurate #12638
  • Add the record of set session autocommit in TiDB logs #12568
  • Record SQL query start time in SessionVars to prevent it from being reset during plan execution #12676
  • Support ? placeholder in ORDER BY, GROUP BY and LIMIT OFFSET #12514
  • Add the Prev_stmt field in slow query logs to output the previous statement when the last statement is COMMIT #12724
  • Record the last statement before COMMIT into the log when the COMMIT fails in an explicitly committed transaction #12747
  • Optimize the saving method of the previous statement when the TiDB server executes a SQL statement to improve performance #12751
  • Fix the panic issue caused by FLUSH PRIVILEGES statements under the skip-grant-table=true configuration #12816
  • Increase the default minimum step of applying AutoID from 1000 to 30000 to avoid performance bottleneck when there are many write requests in a short time #12891
  • Fix the issue that the failed Prepared statement is not print in the error log when TiDB panics #12954
  • Fix the issue that the COM_STMT_FETCH time record in slow query logs is inconsistent with that in MySQL #12953
  • Add an error code in the error message for write conflicts to quickly locate the cause #12878
• DDL
  • Disallow dropping the AUTO INCREMENT attribute of a column by default. Modify the value of the tidb_allow_remove_auto_inc variable if you do need to drop this attribute. See System Variables for more details. #12146
  • Support multiple uniques when creating a unique index in the Create Table statement #12469
  • Fix a compatibility issue that if the foreign key constraint in CREATE TABLE statement has no schema, schema of the created table should be used instead of returning a No Database selected error #12678
  • Fix the issue that the invalid list index error is reported when executing ADMIN CANCEL DDL JOBS #12681
• Monitor
  • Add types for backoff monitoring and supplement the backoff time that is not recorded before, such as the backoff time when committing #12326
  • Add a new metric to monitor Add Index operation progress #12389

PD

• Improve the --help command output of pd-ctl #1772

Tools

• TiDB Binlog
  • Fix the issue that ALTER DATABASE related DDL operations cause Drainer to exit abnormally #770
  • Support querying the transaction status information for Commit binlog to improve replication efficiency #761
  • Fix the issue that a Pump panic might occur when Drainer's start_ts is greater than Pump's largest commit_ts #759

TiDB Ansible

• Add two monitoring items "queue size" and "query histogram" for TiDB Binlog #952
• Update TiDB alerting rules #961
• Check the configuration file before the deployment and upgrade #973
• Add a new metric to monitor index speed in TiDB #987
• Update TiDB Binlog monitoring dashboard to make it compatible with Grafana v4.6.3 #993