pd-server v2.0.0

• Support Region Merge, to merge empty Regions after deleting data [experimental]
• Support Raft Learner [experimental]
• Optimize the scheduler
  • Make the scheduler to adapt to different Region sizes
  • Improve the priority and speed of restoring data during TiKV outage
  • Speed up data transferring when removing a TiKV node
  • Optimize the scheduling policies to prevent the disks from becoming full when the space of TiKV nodes is insufficient
  • Improve the scheduling efficiency of the balance-leader scheduler
  • Reduce the scheduling overhead of the balance-region scheduler
  • Optimize the execution efficiency of the the hot-region scheduler
• Operations interface and configuration
  • Support TLS
  • Support prioritizing the PD leaders
  • Support configuring the scheduling policies based on labels
  • Support configuring stores with a specific label not to schedule the Raft leader
  • Support splitting Region manually to handle the hotspot in a single Region
  • Support scattering a specified Region to manually adjust Region distribution in some cases
  • Add check rules for configuration parameters and improve validity check of the configuration items
• Debugging interface
  • Add the Drop Region debugging interface
  • Add the interfaces to enumerate the health status of each PD
• Statistics
  • Add statistics about abnormal Regions
  • Add statistics about Region isolation level
  • Add scheduling related metrics
• Performance
  • Keep the PD leader and the etcd leader together in the same node to improve write performance
  • Optimize the performance of Region heartbeat

pd-server v2.0.0-rc.4

• Support splitting Region manually to handle the hot spot in a single Region
• Fix the issue that the label property is not displayed when pdctl runs config show all
• Optimize metrics and code structure

pd-server v2.0.0-rc.3

• Support Region Merge, to merge empty Regions or small Regions after deleting data
• Ignore the nodes that have a lot of pending peers during adding replicas, to improve the speed of restoring replicas or making nodes offline
• Fix the frequent scheduling issue caused by a large number of empty Regions
• Optimize the scheduling speed of leader balance in scenarios of unbalanced resources within different labels
• Add more statistics about abnormal Regions

pd-server v2.0.0-rc.1

• Optimize the logic of code used to check the Region status to improve performance
• Optimize the output of log information in abnormal conditions to facilitate debugging
• Fix the monitor statistics that the disk space of TiKV nodes is not enough
• Fix the wrong reporting issue of the health interface when TLS is enabled
• Fix the issue that concurrent addition of replicas might exceed the threshold value of configuration, to improve stability

pd-server v1.1.0-beta

• Add the Drop Region debug interface
• Support setting priority of the PD leader
• Support configuring stores with a specific label not to schedule Raft leaders
• Add the interfaces to enumerate the health status of each PD
• Add more metrics
• Keep the PD leader and the etcd leader together as much as possible in the same node
• Improve the priority and speed of restoring data when TiKV goes down
• Enhance the validity check of the data-dir configuration item
• Optimize the performance of Region heartbeat
• Fix the issue that hot spot scheduling violates label constraint
• Fix other stability issues

pd-server v1.0.8

• Reduce the lock overheat of the region heartbeats
• Fix the issue that a hot region scheduler selects the wrong Leader

pd-server v1.1.0-alpha.1

• Add more APIs
• Support TLS
• Add more cases for scheduling Simulator
• Schedule to adapt to different Region sizes
• Fix some bugs about scheduling

pd-server v1.0.7

• Fix possible scheduling loss issue in abnormal conditions
• Fix the compatibility issue with proto3
• Add the log

pd-server v1.0.6

• Fix store selection problem of hot-region scheduler

pd-server v1.0.5

• Fix the issue that the leaders stop balancing under some circumstances.
  • 869
  • 874
• Fix potential panic during bootstrapping.