Improvements
- Add the tag to trace the source of the
split
command #8936 - Support dynamically changing the
pessimistic-txn.pipelined
configuration #9100 - Reduce the impact on performance when running Backup & Restore and TiDB Lightning #9098
- Add monitoring metrics for the ingesting SST errors #9096
- Prevent the leader from being hibernated when some peers still need to replicate logs #9093
- Increase the success rate of the pipelined pessimistic locking #9086
- Change the default value of
apply-max-batch-size
andstore-max-batch-size
to1024
#9020 - Add the
max-background-flushes
configuration item #8947 - Disable
force-consistency-checks
by default to improve performance #9029
Bug Fixes
- Fix the issue that Coprocessor might return wrong results when there are more than 255 columns #9131
- Fix the issue that Region Merge might cause data loss during network partition #9108
- Fix the issue that the
ANALYZE
statement might cause panic when using thelatin1
character set #9082 - Fix the wrong results returned when converting the numeric type to the time type #9031
- Fix a bug that TiDB Lightning fails to ingest SST files to TiKV with the Importer-backend or Local-backend when Transparent Data Encryption (TDE) is enabled #8995
- Fix the invalid
advertise-status-addr
value (0.0.0.0
) #9036 - Fix the issue that an error is returned indicating that a key exists when this key is locked and deleted in a committed transaction #8930
- Fix the issue that the RocksDB cache mapping error causes data corruption #9029
- Fix a bug that Follower Read might return stale data after the leader is transferred #9240