Doc Menu

TiDB Operator 1.1.6 Release Notes

Release date: October 16, 2020

TiDB Operator version: 1.1.6

Compatibility Changes

  • With #3342, the spec.pd.config will be migrated from YAML format to TOML format automatically; however, if the following parameters are configured in the spec.pd.config, the migration cannot be done after upgrading TiDB Operator to v1.1.6. Therefore, please edit the TidbCluster CR to change the value of the parameter from string format to bool format, for example, from "true" to true.

    • replication.strictly-match-label
    • replication.enable-placement-rules
    • schedule.disable-raft-learner
    • schedule.disable-remove-down-replica
    • schedule.disable-replace-offline-replica
    • schedule.disable-make-up-replica
    • schedule.disable-remove-extra-replica
    • schedule.disable-location-replacement
    • schedule.disable-namespace-relocation
    • schedule.enable-one-way-merge
    • schedule.enable-cross-table-merge
    • pd-server.use-region-storage

Rolling Update Changes

  • If tidb.pingcap.com/sysctl-init: "true" is set for spec.tidb.annotations or spec.tikv.annotations, the TiDB or TiKV cluster will be rolling updated after upgrading TiDB Operator to v1.1.6 due to #3305.
  • If TiFlash is deployed, the TiFlash cluster will be rolling updated after upgrading TiDB Operator to v1.1.6 due to #3345.

New Features

  • Add spec.br.options to the Backup and Restore CR to support customizing arguments for BR (#3360, @lichunzhu)
  • Add spec.tikv.evictLeaderTimeout to TidbCluster CR to make TiKV evict leader timeout configurable (#3344, @lichunzhu)
  • Support monitoring multiple TiDB clusters with one TidbMonitor CR when TLS is disabled. spec.clusterScoped is added to the TidbMonitor CR and needs to be set to true to monitor multiple clusters (#3308, @mikechengwei)
  • Support specifying resources for all initcontainers (#3305, @shonge)
  • Support deploying heterogeneous TiDB clusters (#3003 #3009 #3113 #3155 #3253, @mikechengwei)

Improvements

Bug Fixes

  • Fix the problem that may bootstrap multiple PD clusters (#3365, @lichunzhu)