pd-server v2.1.0-beta
Pre-release
Pre-release
huachaohuang
released this
29 Jun 16:20
·
3463 commits
to master
since this release
Improvements
- Enable Raft PreVote between PD nodes to avoid leader reelection when network recovers after network isolation
- Optimize the issue that Balance Scheduler schedules small Regions frequently
- Optimize the hotspot scheduler to improve its adaptability in traffic statistics information jitters
- Skip the Regions with a large number of rows when scheduling
region merge
- Enable
raft learner
by default to lower the risk of unavailable data caused by machine failure during scheduling - Remove
max-replica
frompd-recover
- Add
Filter
metrics
Bug Fixes
- Fix the issue that Region information is not updated after tikv-ctl unsafe recovery
- Fix the issue that TiKV disk space is used up caused by replica migration in some scenarios
Compatibility notes
- Do not support rolling back to v2.0.x or earlier due to update of the new version storage engine
- Enable
raft learner
by default in the new version of PD. If the cluster is upgraded from 1.x to 2.1, the machine should be stopped before upgrade or a rolling update should be first applied to TiKV and then PD