title | date | summary | tags | ||
---|---|---|---|---|---|
Weekly update (May 22 ~ June 05, 2017) |
2017-06-05 |
Last two weeks, we landed 53 PRs in the TiDB repositories and 33 PRs in the TiKV repositories. |
|
Last two weeks, we landed 53 PRs in the TiDB repositories.
-
Refactor expression evaluation:
-
Update stored password in
mysql.user
table to make it consistent with MySQL. -
Speed up the DDL process:
-
Close DDL worker gracefully when shutting down the TiDB server.
Last two weeks, We landed 33 PRs in the TiKV repositories.
- Introduce gRPC, see 1850, 1865, 1868, 1879.
- Report the written keys to PD for better scheduler.
- Balance the hot regions by peer and leader together.
- Record the cluster bootstrap time.
- Add the state for operators.
- Add the metrics for balance.
- Add the disaster recovery tool for PD.
- Output the allocation statistics when receiving SIGUSR1.
- Add the direct IO configuration for RocksDB.
- Check the cluster ID when re-connecting to PD.
- Reuse the timer to avoid spawning thread frequently.
- Remove the realtime signal handler.
- Prevent accessing nil store when recovering cluster.
- Check the environment TZ setting.
- Use the forward mode for scanning data in GC command.
- Check the last apply index to determine handing snapshot.
- Increase the region size from 64 MB to 96 MB.
- Split total region peer cache.
- Use FlatMap instead of HashMap to improve performance.
- Modify the default compression in RocksDB.