pd-server v5.2.0

Improvements

• Add more QPS dimensions for hot Region scheduling, and support adjusting the priority of the scheduling #3869
• Support hot Region balance scheduling for the write hotspot of TiFlash #3900

Bug Fixes

• Fix the issue that the expected scheduling cannot be generated due to scheduling conflicts among multiple schedulers #3807 #3778

pd-server v5.1.1

Bug fixes

• Fix the issue the expected scheduling cannot be generated when the conflict occurs due to multiple schedulers running at the same time #3807 #3778
• Fix the issue that the scheduler might appear again even if the scheduler is already deleted #2572

pd-server v4.0.14

Improvements

• Reduce the conflicts among multiple schedulers that run at the same time #3858 #3854

Bug fixes

• Fix a bug that the scheduler might reappear after executing the delete operation #2572
• Fix the data race issue that might occur when the scheduler is started before the temporary configuration is loaded #3771
• Fix a PD panic issue that might occur during the Region scattering operation #3761
• Fix the issue that the priority of some operators is not set correctly #3703
• Fix a PD panic issue that might occur when deleting the evict-leader scheduler from a non-existent store #3660
• Fix the issue that the PD Leader re-election is slow when there are many stores #3697

pd-server v5.0.3

Improvements

• Update TiDB Dashboard to v2021.06.15.1 #3798

Bug Fixes

• Fix the data race issue that occurs when loading TTL configurations after the scheduler is started #3771
• Fix a bug that the is_learner field of the TIKV_REGION_PEERS table in TiDB is incorrect #3372 #24293
• Fix the issue that when all TiKV nodes in a zone are offline or down, PD does not schedule replicas to other zones #3705
• Fix the issue that PD might get panic after the scatter Region scheduler is added #3762

pd-server v5.1.0

Improvements

• Avoid unexpected statistics after adding the scatter region scheduler #3602
• Solve multiple scheduling issues in the scaling process
• Optimize the generation process of replica snapshots to solve slow scheduling issues during scaling:
• Solve slow scheduling issues caused by heartbeat pressure due to traffic changes
• Reduce the space discrepancies of large clusters due to scheduling, and optimize the scheduling formula to prevent the bursting issue (which is similar to heterogeneous space clusters) caused by large compression rate discrepancies #3592 #10005

Bug Fixes

• Fix the issue that the PD Leader re-election is slow when there are many stores #3697
• Fix the panic issue that occurs when removing the evict leader scheduler from a non-existent store #3660
• Fix the issue that the statistics are not updated after offline peers are merged #3611

Pre Release v5.2.0-alpha

schedulers: avoid recalculate tolerantRatio (#3745)

* schedulers: avoid recalculate tolerantRatio

Signed-off-by: disksing <i@disksing.com>

* fix test

Signed-off-by: disksing <i@disksing.com>

Co-authored-by: Ti Chi Robot <ti-community-prow-bot@tidb.io>

pd-server v5.0.2

Bug Fixes

• Fix the issue that the PD Leader re-election is slow when there are many stores #3697
• Fix the panic issue that occurs when removing the evict leader scheduler from a non-existent store #3660
• Fix the issue that the statistics are not updated after offline peers are merged #3611

Release v5.1.0-alpha

Release v5.1.0-alpha

pd-server v4.0.13

Improvements

• Optimize the metrics of TSO processing time to help users determine whether the TSO processing time at the PD side is too long #3524
• Update the dashboard version to v2021.03.12.1 #3469

pd-server v5.0.1

Improvements

• Modify the Region score calculator to better satisfy isomerous stores #3605
• Avoid unexpected statistics after adding the scatter region scheduler #3602