title | summary |
---|---|
TiDB 4.0.3 Release Notes |
TiDB 4.0.3 was released on July 24, 2020. New features include TiDB Dashboard improvements, TiFlash file encryption, and support for various tools. Improvements were made to TiDB, TiKV, PD, and TiDB Dashboard. Bug fixes were also implemented for TiDB, TiKV, PD, TiDB Dashboard, TiFlash, TiCDC, Backup & Restore, Dumpling, TiDB Lightning, and TiDB Binlog. |
Release date: July 24, 2020
TiDB version: 4.0.3
-
TiDB Dashboard
-
TiFlash
- Implement file encryption in TiFlash proxy
-
Tools
-
Backup & Restore (BR)
- Support compressing backup files using zstd, lz4 or snappy #404
-
TiCDC
-
Dumpling
-
-
TiDB
- Add the
tidb_log_desensitization
global variable to control whether to do desensitization when logging SQL queries #18581 - Enable
tidb_allow_batch_cop
by default #18552 - Speed up canceling a query #18505
- Add a header for the
tidb_decode_plan
result #18501 - Make the configuration checker compatible with earlier versions of the configuration file #18046
- Enable collecting the execution information by default #18518
- Add the
tiflash_tables
andtiflash_segments
system tables #18536 - Move
AUTO RANDOM
out of experimental features and announce its general availability. The improvements and compatibility changes are as follows:- Deprecate
experimental.allow-auto-random
in the configuration file. No matter how this item is configured, you can always define theAUTO RANDOM
feature on columns. #18613 #18623 - Add the
tidb_allow_auto_random_explicit_insert
session variable to control the explicit writes onAUTO RANDOM
columns. The default value isfalse
. This is to avoid the unexpectedAUTO_RANDOM_BASE
update caused by explicit writes on columns. #18508 - Allow defining
AUTO_RANDOM
only onBIGINT
andUNSIGNED BIGINT
columns and restrict the maximum number of shard bits to15
, which avoids the allocatable space being consumed too quickly #18538 - Do not trigger the
AUTO_RANDOM_BASE
update when defining theAUTO_RANDOM
attribute on theBIGINT
column and inserting the negative value into the primary key #17987 - Use the highest bit of an integer for ID allocation when defining the
AUTO_RANDOM
attribute onUNSIGNED BIGINT
columns, which gets more allocable space #18404 - Support updating the
AUTO_RANDOM
attribute in the result ofSHOW CREATE TABLE
#18316
- Deprecate
- Add the
-
TiKV
-
PD
- Support the JSON formatted log #2565
-
TiDB Dashboard
- Improve the Key Visualizer bucket merge for cold logical ranges #674
- Rename the configuration item of
disable-telemetry
toenable-telemetry
for consistency #684 - Show the progress bar when switching pages #661
- Ensure that the slow log search now follows the same behavior as log search when there are space separators #682
-
TiFlash
- Change the unit of the DDL Jobs panel in Grafana to
operations per minute
- Add a new dashboard in Grafana to show more metrics about TiFlash-Proxy
- Reduce IOPS in TiFlash proxy
- Change the unit of the DDL Jobs panel in Grafana to
-
Tools
-
TiDB
- Return an error instead of an empty set for
IndexHashJoin
when an error occurs during execution #18586 - Fix the recurring panic when gRPC transportReader is broken #18562
- Fix the issue that Green GC does not scan locks on offline stores which might cause data incompleteness #18550
- Forbid processing a non-read-only statement using TiFlash engine #18534
- Return the actual error message when a query connection panics #18500
- Fix the issue that the
ADMIN REPAIR TABLE
execution fails to reload the table metadata on the TiDB node #18323 - Fix the data inconsistency issue occurred because the lock of a written and deleted primary key in one transaction is resolved by another transaction #18291
- Make spilling disk work well #18288
- Fix the error reported when the
REPLACE INTO
statement works on the table that contains generated columns #17907 - Return the OOM error when the
IndexHashJoin
andIndexMergeJoin
workers panic #18527 - Fix the bug that the execution of
Index Join
might return wrong results in special cases when the index used byIndex Join
contains the integer primary key #18565 - Fix the issue that when the new collation is enabled on the cluster, the data updated on columns with the new collation in a transaction cannot be read through the unique index #18703
- Return an error instead of an empty set for
-
TiKV
-
PD
- Fix the issue that creating TSO stream might be blocked for a while if the server crashes #2648
- Fix the issue that
getSchedulers
might cause a data race #2638 - Fix the issue that deleting the scheduler might cause deadlocks #2637
- Fix the bug that placement rules are not considered when
balance-leader-scheduler
is enabled #2636 - Fix the issue that sometimes service
safepoint
cannot be set properly, which might make BR and dumpling fail #2635 - Fix the issue that the target store in
hot region scheduler
is incorrectly selected #2627 - Fix the issue that the TSO request might take too long when PD leader is switched #2622
- Fix the issue of stale scheduler after leader change #2608
- Fix the issue that sometimes replicas of a Region cannot be adjusted to the best location when placement rules are enabled #2605
- Fix the issue that the deployment path of the store is not updated according to the change of deployment directory #2600
- Prevent
store limit
from changing to zero #2588
-
TiDB Dashboard
- Fix the TiDB connection error when TiDB is scaled out #689
- Fix the issue that TiFlash instances are not displayed in the log searching page #680
- Fix the issue of metric selection reset after refreshing the overview page #663
- Fix a connection issue in some TLS scenarios #660
- Fix the issue that the language dropdown box is not displayed correctly in some cases #677
-
TiFlash
- Fix the issue that TiFlash crashes after renaming the primary key column
- Fix the issue that concurrent
Learner Read
andRemove Region
might cause deadlocks
-
Tools
-
TiCDC
-
Backup & Restore (BR)
- Fix the issue that the backup time might be negative #405
-
Dumpling
-
TiDB Lightning
- Fix the issue that
--log-file
does not take effect #345
- Fix the issue that
-
TiDB Binlog
- Fix the issue that when TiDB Binlog replicates data to the downstream with TLS enabled, Drainer cannot be started which occurs because TLS is not enabled on the database driver used to update the checkpoint #988
-