TiDB 4.0.8 released, distributed NewSQL database

TiDB 4.0.8 has been released. The specific updates of this version are as follows:

new function

• TiDB

  • Support aggregate functions  APPROX_PERCENTILE #20197

• TiFlash

  • Support  CAST function pushdown

• Tools

  • TiCDC

    • Support snapshot level consistent replication  #932

Optimization and promotion

• TiDB

  • In the greedy algorithm that selects the index combination to calculate the expression selection rate, the index with the low selection rate is preferred  #20154
  • Record more RPC information in Coprocessor running status  #19264
  • Optimize the efficiency of reading slow logs to improve slow query performance  #20556
  • When selecting the execution plan, the optimizer will wait for the timeout execution plan in the Plan Binding phase to record more Debug information.  #20530
  • Increase statement retry time in slow query and slow log  #20495  #20494
  • Add system table  table_storage_stats #20431
  • As  INSERT/ UPDATE/ REPLACE run record statements relating to the RPC message  # 20430
  • EXPLAIN FOR CONNECTION Added operator information in  the result of the statement  #20384
  • Adjust the client connection establishment/disconnection log level to #20321 in the TiDB log DEBUG 
  • Add monitoring information of Coprocessor Cache  #20293
  • Record more pessimistic lock related parameters in runtime information  #20199
  • Add two new time-consuming information in runtime information and Trace function  #20187
  • Add runtime information of transaction commit in the slow log  #20185
  • Close Index Merge Join  #20599
  • Add ISO 8601 and time zone support for temporary string constants  #20670

• TiKV

  • Add  Fast-Tune  monitoring page to assist performance diagnosis  #8804
  • Add  security.redact-info-log configuration to delete user data from logs  #8746
  • Modify the metafile format of error code  #8877
  • Enable dynamic modification  pessimistic-txn.pipelined configuration  #8853
  • The memory profile function is enabled by default  #8801

• PD

  • Metafile that generates error  #3090
  • Add more useful information to the operator log  #3009

• TiFlash

  • Add monitoring of Raft log
  • Add cop monitoring of  task memory usage
  • Make min/ max index more accurate in the presence of deleted data 
  • Improve query performance under small batch data
  • Add  error.toml files to support standard error codes

• Tools

  • Backup & Restore (BR)

    •  Speed ​​up recovery by adding  split and  ingestpipeline  #427
    • Scheduler that supports manual recovery of PD  #530
    • Change the PD scheduler interface to suspend scheduler  #551

  • TiCDC

    • Periodically output statistics in the MySQL sink  #1023

  • Dumpling

    • Support directly exporting data to S3 storage  #155
    • Support exporting View  #158
    • Support exporting data tables that only contain generated columns  #166

  • TiDB Lightning

    • Support multi-byte CSV delimiter and separator  #406
    • Speed ​​up import by disabling some PD schedulers  #408
    • Use the GC-TTL interface on the v4.0 cluster to prevent GC errors in the checksum phase  #396

Bug fix

• TiDB

  • Fix the problem of unexpected Panic when using partition table  #20565
  • When fixing the outer join, if the outer table has filter conditions, the result of Index Merge Join is sometimes incorrect  #20427
  • Fix   the problem of  BIT incorrectly returning due to type length overflow when the type is converted  #20363NULL
  • Fix   the problem that the default value of the ALTER TABLE ... syntax change  BITtype may cause the default value to be wrong  #20340
  • Repair  BIT convert type  INT64 when the length of the overflow could lead to the wrong question  # 20312
  • Fix the problem that mixed types of columns may cause incorrect results when performing conditional propagation optimization  #20297
  • Fix the issue of Panic when the Plan Cache stores expired execution plans.  #20246
  • Repair  FROM_UNIXTIME and  UNION ALL when used together, the result is returned incorrectly truncates question  # 20240
  • Fix   the problem that the type may cause incorrect results when Enum converted to  Floattype  #20235
  • Fix  RegionStore the problem of Panic under certain conditions  #20210
  • Fix  BatchPointGet the problem of wrong results when requesting to sort the maximum value of unsigned integers  #20205
  • Fix  Enum/ Set Type When mixed Collation's type determination, the result that Collation may be incompatible with MySQL  #20364
  • Fix  the problem that the result is incompatible with MySQL when 0 converting other types  to  YEARtypes  #20292
  • Fix   the issue that the reported result is incorrect when KV Duration included store0in the  monitoring indicators  #20260
  • Fix the problem Float that when writing  type data, it is out of range still written incorrectly after the error is  prompted due to length overflow  #20252
  • Fix the problem that the  NOT NULL attribute generation column allows to write NULL values  in some cases  #20216
  • Fix  YEAR the problem that the error message is inaccurate when the type data is written beyond the allowable range  #20170
  • Fix invalid auto-id the problem  that an error will be reported when a pessimistic transaction is retried in some cases  #20134
  • Fix   the problem that the repetitive constraint check is not performed when ALTER TABLE changing  Enum/ Settype  #20046
  • Fix the problem that the running information of Coprocessor Task is incorrect when some operators are executed concurrently  #19947
  • Read-only system variables can not be repaired as explicitly Session-level variables  SELECT problem  # 19944
  • Fix the problem that repeated  ORDER BY conditions sometimes cause the execution plan selection to be sub- optimal  #20333
  • Fix the problem of failure due to font exceeding the maximum allowed value when generating Metric Profile  #20637

• TiKV

  • Fix the problem that the lock conflict in the encryption function causes the pd-worker to process the heartbeat slowly  #8869
  • Fix the problem of incorrectly generating memory profile  #8790
  • Fix the problem of error when specifying GCS storage class (storage class) during backup  #8763
  • Fixed the problem that the Leader can't be found on the Learner node of a restart or new Split  #8864

• PD

  • Fixed the bug that TiDB Dashboard caused PD panic in some scenarios  #3096
  • Fixed an error that may cause PD panic after a PD store is offline for more than ten minutes  #3069

• TiFlash

  • Fixed the problem of wrong timestamp in log information
  • Fixed an issue where the wrong capacity when using multi-disk deployment caused the creation of TiFlash copy to fail
  • Fixed the problem that TiFlash may prompt that the data file is damaged after restarting
  • Fixed an issue where damaged files may remain on the disk after TiFlash crashes
  • Fixes the wait index duration problem of slow query due to the fact that the status in the Raft Learner protocol cannot be updated in time when the write traffic is small 
  • Fixed an issue where the proxy would write a large amount of Region state information to the key-value engine when replaying expired Raft logs

• Tools

  • Backup & Restore (BR)

    • Fix send on closed channel panic issue  that may occur during Restore  #559

  • TiCDC

    • Fix the problem that the owner failed to exit due to failure to update GC safepoint  #979
    • Fix unexpected task information update  #1017
    • Fix unexpected empty Maxwell message  #978

  • TiDB Lightning

    • Fix the problem of column information error  #420
    • Fix the problem of infinite loop when obtaining Region information in Local mode  #418

Update instructions: https://docs.pingcap.com/zh/tidb/stable/release-4.0.8