TiDB 3.1 Beta.2 Release Notes

Release date: March 9, 2020

TiDB version: 3.1.0-beta.2

TiDB Ansible version: 3.1.0-beta.2

Compatibility Changes

• Tools
  • TiDB Lightning
    • Use the default configurations specified in the TiDB Lightning Configuration for certain items not configured in the configuration file #255
    • Add the --tidb-password CLI parameter to set the TiDB password #253

New Features

• TiDB
  • Support adding the AutoRandom keyword in the column attribute to enable TiDB to automatically assign random integers to the primary key, which avoids the write hot spot caused by the AUTO_INCREMENT primary key #14555
  • Support creating or deleting column store replicas through DDL statements #14537
  • Add the feature that the optimizer can independently select different storage engines #14537
  • Add the feature that the SQL hint supports different storage engines #14537
  • Support reading data from followers by using the tidb_replica_read system variable #13464
• TiKV
  • Raftstore
    • Add the peer_address parameter to connect other nodes to the TiKV server #6491
    • Add the read_index and read_index_resp monitoring metrics to monitor the number of ReadIndex requests #6610
• PD Client
  • Support reporting statistics of local threads to PD #6605
• Backup
  • Replace the RocksIOLimiter flow control library with Rust’s async-speed-limit flow control library to eliminate extra memory copies when backing up a file #6462
• PD
  • Tolerate backslash in the location label name #2084
• TiFlash
  • Initial release
• TiDB Ansible
  • Support deploying multiple Grafana/Prometheus/Alertmanager in one cluster #1143
  • Support deploying the TiFlash component #1148
  • Add monitoring metrics related to the TiFlash component #1152

Bug Fixes

• TiKV
  • Raftstore
    • Fix the issue that the read requests cannot be processed because data is not properly read from Hibernate Regions #6450
    • Fix the panic issue caused by the ReadIndex requests during the leader transfer process #6613
    • Fix the issue that Hibernate Regions are not correctly awakened in some special conditions #6730 #6737 #6972
  • Backup
    • Fix the inconsistent data index during the restoration caused by the backup of the extra data #6659
    • Fix the panic caused by incorrectly processing the deleted values during the backup #6726
• PD
  • Fix the panic occurred because the rule checker fails to assign stores to Regions #2161
• Tools
  • TiDB Lightning
    • Fix the bug that the web interface does not work outside the Server mode #259
  • BR (Backup and Restore)
    • Fix the issue that BR cannot exit in time due to an unrecoverable error it encounters when restoring data #152
• TiDB Ansible
  • Fix the issue that the rolling update command fails because the PD Leader cannot be obtained in some scenarios #1122