In TiDB 3.1 Beta released on Dec 20, 2019, TiDB introduced Follower Read, a highlight open-source feature that lets any follower replica in a Region serve a read request under the premise of strongly consistent reads. This feature improves the throughput of the TiDB cluster and reduces the load on the Raft leader.
Last week, we published a post that guides you through why we introduced Follower Read, how we implement it, and our future plans for it.
The full post is here:
Doubling System Read Throughput with Only 26 Lines of Code
tidb:
tikv:
Last week, we landed 64 PRs in the TiDB repository, 3 PRs in the TiSpark repository, and 33 PRs in the TiKV and PD repositories.
TiDB:
CAST
push-down switcherpoint get
resultlower()
/upper()
does not work if a string contains the utf8
characterindexJoin
information_schema
.tables
to support AUTO_RANDOM
LogicalMemTable
in the cascades plannerElapse
during Hash Join
WindowExec
by using the sliding windowjson_objectagg
aggregate functionTiSpark:
CHblock
decoding bugTiKV and PD:
yyyy-mm-dd HH:MM:SS
format cannot be parsed to Duration
Yield
for apply
Initializer
of CDC's Endpoint
ChunkEncoder
operator Influence
when sorting storesInfra Meetup No.124 (online) (in Chinese)
Date: February 15th, 2020