# TiKV v7.5.7

2025-09-04 · [Original release notes](https://github.com/tikv/tikv/releases/tag/v7.5.7) · [Full summary](https://ratatosk.io/en/releases/tikv/v7.5.7)

Security 2 · Breaking changes 2 · Deprecations 1 · Important fixes 63 · Enhancements 11 · Other changes 43

## Security (2)

### Redact log now takes effect in affected scenarios

- Affects: setups with Redact log enabled

In certain scenarios, enabling Redact log had no effect.

> - Fix the issue that enabling Redact log does not take effect in certain scenarios #59279 @ tangenta

https://docs.pingcap.com/tidb/stable/release-7.5.7

### Shortened, redacted hexadecimal keys in SST ingest warning logs

- Affects: SST ingest warning logs (Inferred)

When TiKV emits the `sst ingest is too slow` warning, the log could be excessively long, expose raw keys, and print keys without hexadecimal formatting.

> - tikv/tikv#18549

## Breaking changes (2)

### Historical statistics now default to OFF

- Affects: TiDB setups relying on the default for `tidb_enable_historical_stats` (Inferred)

The default was ON; with no operator-side change, the new OFF default turns historical statistics off.

> - Change the default value of tidb_enable_historical_stats from ON to OFF , which turns off historical statistics to avoid potential stability issues #53048 @ hawkingrei

https://docs.pingcap.com/tidb/stable/release-7.5.7

### Cluster-wide statistics refresh skipped after `ANALYZE`

- Affects: setups that execute `ANALYZE` statements

`ANALYZE` execution time is reduced.

> - Avoid refreshing statistics across the entire cluster after ANALYZE statement execution, reducing the execution time of ANALYZE #57631 @ 0xPoe

https://docs.pingcap.com/tidb/stable/release-7.5.7

## Deprecations (1)

### TiKV automatic-compaction settings deprecated

- Affects: TiKV setups using any of `region-compact-check-interval`, `region-compact-check-step`, `region-compact-min-tombstones`, `region-compact-tombstones-percent`, `region-compact-min-redundant-rows`, or `region-compact-redundant-rows-percent`
- Removed in: not stated

The `gc.auto-compaction` configuration group is the replacement for automatic compaction.

> - TiKV deprecates the following configuration items and replaces them with the new gc.auto-compaction configuration group, which controls automatic compaction behavior #18727 @ v01dstar
> - Deprecated configuration items: region-compact-check-interval , region-compact-check-step , region-compact-min-tombstones , region-compact-tombstones-percent , region-compact-min-redundant-rows , and region-compact-redundant-rows-percent .

https://docs.pingcap.com/tidb/stable/release-7.5.7

## Important fixes (63)

### Mapped allocation per thread metric corrected

- Affects: mapped allocation per thread metrics in use

When allocation-byte rates showed allocations on `sst-importer` and `apply` threads, the metric instead showed high memory usage in `sched-worker-high` and `unified-read-pool`, which were not in use.

> - tikv/tikv#18125
> - Fix incorrect thread memory metrics #18125 @ Connor1996

### Inconsistent SQL digests fixed

- Affects: SQL digest generation (Inferred)

SQL statements containing multiple charset introducers, such as `_charset(xxx)` and `_charset(xxx2)`, could generate different digests.

> - Fix the issue that SQL statements containing _charset(xxx), _charset(xxx2), ... generate different digests #58447 @ xhebox

https://docs.pingcap.com/tidb/stable/release-7.5.7

### Incorrect key range in `ALTER RANGE meta SET PLACEMENT POLICY` fixed

- Affects: use of `ALTER RANGE meta SET PLACEMENT POLICY`

When using `ALTER RANGE meta SET PLACEMENT POLICY`, the operation could use an incorrect key range.

> - Fix the issue of incorrect key range in ALTER RANGE meta SET PLACEMENT POLICY #60888 @ nolouch

https://docs.pingcap.com/tidb/stable/release-7.5.7

### Incorrect Grafana Stats Healthy Distribution panel data fixed

- Affects: Grafana setups using the Stats Healthy Distribution panel (Inferred)

Before this fix, the panel could show incorrect data.

> - Fix the issue that the data in the Stats Healthy Distribution panel of Grafana might be incorrect #57176 @ hawkingrei

https://docs.pingcap.com/tidb/stable/release-7.5.7

### `latin1_bin` comparison behavior fixed

- Affects: SQL comparisons using the `latin1_bin` collation (Inferred)

Comparisons using `latin1_bin` differed from those using `utf8mb4_bin` and `utf8_bin`.

> - Fix the issue that the comparison behavior of latin1_bin differs from that of utf8mb4_bin and utf8_bin #60701 @ hawkingrei

https://docs.pingcap.com/tidb/stable/release-7.5.7

### [Silent failure] TiDB `ADD UNIQUE INDEX` data inconsistency fixed

- Affects: TiDB setups that execute `ADD UNIQUE INDEX`

Before this fix, executing `ADD UNIQUE INDEX` might cause data inconsistency.

> - Fix the issue that executing ADD UNIQUE INDEX might cause data inconsistency #60339 @ tangenta

https://docs.pingcap.com/tidb/stable/release-7.5.7

### Non-public indexes no longer appear in TiDB’s statistics system table

- Affects: TiDB setups querying the statistics system table while non-public indexes exist (Inferred)

Previously, the statistics system table showed non-public indexes.

> - Fix the issue that non-public indexes are shown in the statistics system table #60430 @ tangenta

https://docs.pingcap.com/tidb/stable/release-7.5.7

### TiDB row-count estimation fixed for TopN statistics

- Affects: TiDB setups where all collected column statistics are in TopN and subsequent writes occur

The row-count estimate could remain at 0 after subsequent writes.

> - Fix the issue that when the collected column statistics are all in TopN, the row count estimation might remain 0 even after subsequent writes #47400 @ terry1purcell

https://docs.pingcap.com/tidb/stable/release-7.5.7

### TiDB estimated costs in `cost_trace` output corrected

- Affects: TiDB queries using `explain format="cost_trace"`

The estimated cost shown by `explain format="cost_trace"` could be incorrect.

> - Fix the issue that the estimated cost displayed in explain format="cost_trace" might be incorrect #61155 @ hawkingrei

https://docs.pingcap.com/tidb/stable/release-7.5.7

### TiDB row count overestimation across months or years fixed

- Affects: TiDB setups estimating row counts across months or years (Inferred)

Estimates spanning months or years could be significantly overestimated.

> - Fix the issue that row count estimates across months or years can be significantly overestimated #50080 @ terry1purcell

https://docs.pingcap.com/tidb/stable/release-7.5.7

### TiDB JOIN row count estimation accuracy fixed

- Affects: TiDB JOIN row count estimation (Inferred)

When column or index statistics were missing for a JOIN, row count estimates could be highly inaccurate. That estimation is fixed in this release.

> - Fix the issue that row count estimation for JOIN can be highly inaccurate when column or index statistics are missing #61602 @ qw4990

https://docs.pingcap.com/tidb/stable/release-7.5.7

### TiDB statistics fix for virtual generated first columns

- Affects: TiDB tables whose first column is a virtual generated column (Inferred)

Statistics could be incorrect for these tables before this release; the issue is fixed in this release.

> - Fix the issue that statistics might be incorrect when the first column of a table is a virtual generated column #61606 @ winoros

https://docs.pingcap.com/tidb/stable/release-7.5.7

### ADMIN CHECK success-on-failure reporting fixed

- Affects: ADMIN CHECK operations

The issue occurred when internal SQL executions failed during the check.

> - Fix the issue that ADMIN CHECK still returns success even after some internal SQL executions fail #61612 @ joechenrh

https://docs.pingcap.com/tidb/stable/release-7.5.7

### TiKV Grafana dashboard display of two metrics corrected

- Affects: `Ingestion picked level` and `Compaction Job Size(files)` in the TiKV Grafana dashboard

The dashboard showed these values incorrectly when they were viewed.

> - Fix the issue that Ingestion picked level and Compaction Job Size(files) are displayed incorrectly in the TiKV dashboard in Grafana #15990 @ Connor1996

https://docs.pingcap.com/tidb/stable/release-7.5.7

### TiKV time-zone conversion corrected for Brazil and Egypt

- Affects: TiKV workloads that perform time-zone conversions for Brazil or Egypt

When TiKV converted time zones for Brazil or Egypt, it could produce an incorrect result.

> - Fix the issue that TiKV converts the time zone incorrectly for Brazil and Egypt #16220 @ overvenus

https://docs.pingcap.com/tidb/stable/release-7.5.7

### Repeated `split-merge-interval` changes now take effect

- Affects: PD setups that repeatedly change `split-merge-interval`

When the setting was changed repeatedly—for example, from 1s to 1h and back to 1s—a change might not take effect.

> - Fix the issue that the split-merge-interval configuration item might not take effect when you modify its value repeatedly (such as changing it from 1s to 1h and back to 1s ) #8404 @ lhy1024

https://docs.pingcap.com/tidb/stable/release-7.5.7

### Increasing resolved ts lag after TiKV scaling fixed

- Affects: TiKV clusters scaling nodes on the same IP address

With outdated store IDs, scaling TiKV nodes in or out on the same IP address caused resolved ts lag to keep increasing; that behavior is fixed.

> - Fix the issue that the resolved ts lag keeps increasing after scaling in or out TiKV nodes on the same IP address because of outdated store IDs #12162 @ 3AceShowHand

https://docs.pingcap.com/tidb/stable/release-7.5.7

### Per-CF timestamp filtering added to SST restores

- Affects: `sst_importer` restores with timestamp-overlapping SST versions

When SST data contained stale or unfinished versions before a boundary, or versions beyond the target restore point, restores could reintroduce already-applied changes or break consistency.

> - tikv/tikv#18399
> - The Download API of TiKV supports filtering out data within a certain time range when downloading backup files, which avoids importing outdated or future data versions during restore #18399 @ 3pointer

### Raftstore slow-log message index reporting fixed

- Affects: TiKV deployments that emit raftstore slow logs for `StoreFsm::handle_msgs` (Inferred)

When slow logs occurred during `StoreFsm::handle_msgs`, the log could identify a slow `StoreMsg::CompactEvent` as `StoreMsg::Unreachable`, producing misleading debugging output.

> - tikv/tikv#18561
> - Fix misleading descriptions in StoreMsg log entries in slow logs #18561 @ LykxSassinator

### False slow-store detection on other TiKV nodes after recovery fixed

- Affects: clusters with TiKV failure recovery

After one TiKV failure recovered, another node was detected as a slow store, its leaders were evicted, and reported QPS dropped by 50%.

> - tikv/tikv#18605

### In-progress manual compaction jobs can be canceled during graceful shutdown

- Affects: TiKV nodes that run manual compaction jobs

During graceful shutdown, an in-progress job could keep running and delay a rolling restart or shutdown.

> - tikv/tikv#18396
> - Fix the issue that TiKV fails to terminate ongoing manual compaction tasks during graceful shutdown #18396 @ LykxSassinator

### TiKV Raft peers stuck in `busy apply` after restart fixed

- Affects: TiKV Raft peers

After a TiKV restart, some peers could remain in `busy apply` for two minutes despite having no Raft log lag, causing `Server is busy` errors and preventing PD from transferring leaders to that instance.

> - tikv/tikv#18233
> - Fix the issue that the unexpected Server is busy error occurs after TiKV restarts #18233 @ LykxSassinator

### Slow Azure Blob uploads from non-memory sources fixed

- Affects: Azure Blob backups that read upload data from a non-memory source

A reported 400 MiB flush took about 17 minutes and fully used one CPU core.

> - tikv/tikv#18410
> - Fix the issue that log backup uploads to Azure Blob Storage are slow when transferring large volumes of data #18410 @ YuJuncen

### Raftstore deadlock during CPU profiling fixed

- Affects: TiKV deployments on AWS arm using CPU profiling

During CPU profiling, a deadlock could block the `raftstore` thread and stop it from sending heartbeats to PD, making TiKV appear disconnected.

> - tikv/tikv#18474
> - Fix the issue that a deadlock might be triggered during CPU profiling #18474 @ YangKeao

### Repeated `Task::Compact` compactions fixed

- Affects: large regions whose MVCC versions belong to one row

When `cleanup_worker` had pending work while many regions needed compaction, an unsplittable large region could repeatedly enqueue `Task::Compact` requests. The duplicate manual compactions did no useful work and blocked other tasks that needed compaction.

> - tikv/tikv#18451

### Wrong results with `utf8mb4_unicode_ci` fixed

- Affects: the `utf8mb4_unicode_ci` collation (Inferred)

> - tikv/tikv#18784

### Online Unsafe Recovery blocked by TiFlash replicas fixed

- Affects: clusters using TiFlash with online unsafe recovery and automatic failed-store detection

Certain TiFlash replicas could block recovery from advancing the commit index, causing online unsafe recovery to exceed its timeout.

> - tikv/tikv#18197
> - Fix the issue that Online Unsafe Recovery might be blocked by certain TiFlash replicas, preventing the commit index from advancing #18197 @ v01dstar

### BR supports IMDSv2 for Amazon S3 backups

- Affects: TiDB deployments on Amazon EC2 where BR accesses Amazon S3 using the instance's IAM role

Previously, full-database backups to S3 could fail to find AWS credentials even when the TiKV pod had S3 access. BR can now use the EC2 instance's associated IAM role to access S3 when the instance is configured to allow it.

> - When deploying TiDB on Amazon EC2, BR supports AWS Instance Metadata Service Version 2 (IMDSv2). You can configure your EC2 instance to allow BR to use the IAM role associated with the instance for appropriate permissions to access Amazon S3 #16443 @ pingyu

https://docs.pingcap.com/tidb/stable/release-7.5.7

### Shared KV request data races fixed in `IndexMerge` and `IndexLookUp`

- Affects: query pushdown using shared KV requests in `IndexMerge` or `IndexLookUp` (Inferred)

When queries were pushed down, shared KV requests in these operators could cause data races.

> - Fix the issue that shared KV requests in IndexMerge and IndexLookUp operators cause data races when pushing down queries #60175 @ you06

https://docs.pingcap.com/tidb/stable/release-7.5.7

### Hash Aggregation goroutine leak fixed

- Affects: Setups using the Hash Aggregation operator (Inferred)

> - Fix a potential goroutine leak issue in the Hash Aggregation operator #58004 @ xzhangxian1008

https://docs.pingcap.com/tidb/stable/release-7.5.7

### Frequent Region merges blocking TTL job startup fixed

- Affects: setups running TTL jobs (Inferred)

Frequent Region merges could prevent TTL jobs from starting.

> - Fix the issue that frequent Region merges prevent TTL jobs from starting #61512 @ YangKeao

https://docs.pingcap.com/tidb/stable/release-7.5.7

### TiFlash query inconsistency after lossy DDL fixed

- Affects: TiFlash queries (Inferred)

After a lossy DDL statement was executed, TiFlash query results could be inconsistent.

> - Fix the issue that TiFlash query results are inconsistent after executing a lossy DDL statement #61455 @ Lloyd-Pottiger

https://docs.pingcap.com/tidb/stable/release-7.5.7

### DDL operations getting stuck after schema-version update failures fixed

- Affects: setups with metadata locking (MDL) disabled

In this setup, a DDL operation could get stuck if it failed to update the schema version.

> - Fix the issue that after disabling metadata locking (MDL), DDL operations get stuck after failing to update the schema version #61210 @ wjhuang2016

https://docs.pingcap.com/tidb/stable/release-7.5.7

### TiDB session crash fixed

- Affects: setups with `Fix Control #44855` enabled

A TiDB session might crash when `Fix Control #44855` was enabled.

> - Fix the issue that a TiDB session might crash when Fix Control #44855 is enabled #59762 @ winoros

https://docs.pingcap.com/tidb/stable/release-7.5.7

### TiDB in-memory statistics deletion on task timeout fixed

- Affects: TiDB statistics exception handling

When background tasks timed out, improper exception handling for statistics could mistakenly delete in-memory statistics.

> - Fix the issue that improper exception handling for statistics causes in-memory statistics to be mistakenly deleted when background tasks time out #57901 @ hawkingrei

https://docs.pingcap.com/tidb/stable/release-7.5.7

### TiDB Hash Join v1 panic recovery fixed

- Affects: TiDB setups using the Hash Join v1 operator

When its `Close()` method encountered a panic, it failed to recover from it.

> - Fix the issue that the Close() method of the Hash Join v1 operator fails to recover from a panic #60926 @ xzhangxian1008

https://docs.pingcap.com/tidb/stable/release-7.5.7

### TiFlash shallow-copy crash and incorrect-results issue fixed

- Affects: TiDB setups using TiFlash

A shallow copy of `PhysicalExchangeSender.HashCol` could make TiFlash crash or produce incorrect results.

> - Fix the issue that shallow copy of PhysicalExchangeSender.HashCol causes TiFlash to crash or generate incorrect results #60517 @ windtalker

https://docs.pingcap.com/tidb/stable/release-7.5.7

### TiDB latest-statistics update failure fixed

- Affects: TiDB clusters running `ANALYZE` statements

In extreme cases, when an `ANALYZE` statement's disk-spilling operation took too long, other TiDB nodes might fail to update the latest statistics.

> - Fix the issue that when the disk-spilling operation of an ANALYZE statement takes too long in extreme cases, other TiDB nodes might fail to update the latest statistics #54552 @ 0xPoe

https://docs.pingcap.com/tidb/stable/release-7.5.7

### TiDB circular foreign key infinite-loop issue fixed

- Affects: TiDB setups using circular foreign key definitions (Inferred)

Before this fix, circular foreign key definitions could cause infinite loops.

> - Fix the issue that circular foreign key definitions cause infinite loops #60985 @ hawkingrei

https://docs.pingcap.com/tidb/stable/release-7.5.7

### TiDB Plan Cache execution-plan error fixed

- Affects: TiDB setups using Plan Cache (Inferred)

Previously, Plan Cache could store an incorrect execution plan, causing execution errors.

> - Fix the issue that Plan Cache stores an incorrect execution plan, causing execution errors #56772 @ dash12653

https://docs.pingcap.com/tidb/stable/release-7.5.7

### TiDB TopN pushdown sort-item generation fixed

- Affects: queries using TopN pushdown with expression-based sort items (Inferred)

During TopN pushdown, expression-based sort items could be generated incorrectly; this is now fixed.

> - Fix the issue that an expression-based TopN sort item is incorrectly generated during TopN pushdown #60655 @ hawkingrei

https://docs.pingcap.com/tidb/stable/release-7.5.7

### TiDB default for `tidb_cost_model_version` fixed

- Affects: TiDB setups using the default for `tidb_cost_model_version` (Inferred)

The variable’s default value was set incorrectly.

> - Fix the issue that the default value of the system variable tidb_cost_model_version is set incorrectly #61565 @ hawkingrei

https://docs.pingcap.com/tidb/stable/release-7.5.7

### Index-adding hang fixed

- Affects: index additions

Executing `ADMIN CANCEL DDL JOBS` while an index was being added could make the index-adding process hang.

> - Fix the issue that executing ADMIN CANCEL DDL JOBS while adding an index causes the index-adding process to hang #61087 @ tangenta

https://docs.pingcap.com/tidb/stable/release-7.5.7

### TiKV data and index consistency fixed for multi-schema changes

- Affects: multi-schema changes that add multiple indexes

After adding multiple indexes through a multi-schema change, data and indexes could become inconsistent.

> - Fix the issue that data and indexes became inconsistent after adding multiple indexes through multi-schema change #61255 @ tangenta

https://docs.pingcap.com/tidb/stable/release-7.5.7

### TiKV use of client-incompatible compression fixed

- Affects: TiKV deployments using `server.grpc-compression-type` with clients that cannot decode TiKV's compression algorithm

During an upgrade from 4.0 to 6.5.11 with `server.grpc-compression-type` configured, TiKV could use a compression algorithm the client could not decode, causing TiDB to fail to provide normal service. This incompatible algorithm use is fixed, though the note does not specify the exact algorithm-selection behavior after the fix.

> - Fix the issue that TiKV might use a compression algorithm that the client cannot decode #18079 @ ekexium

https://docs.pingcap.com/tidb/stable/release-7.5.7

### PD goroutine leaks from TiDB Dashboard connections fixed

- Affects: PD setups handling TCP connections from TiDB Dashboard

When TiDB Dashboard TCP connections were improperly closed, PD could leak goroutines; this leak is fixed.

> - Fix the issue that improperly closing TiDB Dashboard TCP connections could lead to PD goroutine leaks #9402 @ baurine

https://docs.pingcap.com/tidb/stable/release-7.5.7

### Scheduling failures for newly added TiKV nodes fixed

- Affects: PD setups scheduling newly added TiKV nodes

Before this fix, newly added TiKV nodes might fail to be scheduled.

> - Fix the issue that newly added TiKV nodes might fail to be scheduled #9145 @ bufferflies

https://docs.pingcap.com/tidb/stable/release-7.5.7

### TSO service unavailability fixed

- Affects: PD setups with `tidb_enable_tso_follower_proxy` enabled

When `tidb_enable_tso_follower_proxy` was enabled, the TSO service could become unavailable; this release fixes that issue.

> - Fix the issue that enabling tidb_enable_tso_follower_proxy might cause the TSO service to become unavailable #9188 @ Tema

https://docs.pingcap.com/tidb/stable/release-7.5.7

### TiFlash panic from SST-file deletion fixed

- Affects: TiFlash setups using `IMPORT INTO` or BR restore (Inferred)

Accidental deletion of SST files while running `IMPORT INTO` or BR restore could make TiFlash panic.

> - Fix the panic issue caused by accidental deletion of SST files during running IMPORT INTO or BR restore #10141 @ CalvinNeo

https://docs.pingcap.com/tidb/stable/release-7.5.7

### TiFlash panic on NULL expression indexes fixed

- Affects: TiFlash setups using expression indexes in the form of `((NULL))`

Before this release, creating an expression index in the form of `((NULL))` could cause TiFlash to panic. Creating this form of expression index no longer causes the panic.

> - Fix the issue that creating an expression index in the form of ((NULL)) causes TiFlash to panic #9891 @ JaySon-Huang

https://docs.pingcap.com/tidb/stable/release-7.5.7

### TiFlash snapshot handling panic fixed

- Affects: TiFlash setups handling snapshots with irregular Region key-ranges

Before this release, TiFlash might panic when handling snapshots with irregular Region key-ranges; that panic is now fixed.

> - Fix the issue that TiFlash might panic when handling snapshots with irregular Region key-ranges #10147 @ JaySon-Huang

https://docs.pingcap.com/tidb/stable/release-7.5.7

### TiFlash excessive memory use fixed

- Affects: TiFlash cluster tables

When cluster tables contain many `ENUM` columns, TiFlash could consume a large amount of memory.

> - Fix the issue that TiFlash might consume a large amount of memory when tables in a cluster contain a large number of ENUM type columns #9947 @ JaySon-Huang

https://docs.pingcap.com/tidb/stable/release-7.5.7

### TiFlash restart failure after large-row insertion fixed

- Affects: TiFlash workloads that insert a single row larger than 16 MiB

After a single row larger than 16 MiB was inserted, TiFlash could fail to restart. That restart failure is fixed.

> - Fix the issue that TiFlash might fail to restart after inserting a single row of data larger than 16 MiB #10052 @ JaySon-Huang

https://docs.pingcap.com/tidb/stable/release-7.5.7

### PITR restore failure for large indexes fixed

- Affects: setups using PITR to restore indexes larger than 3072 bytes

PITR failed to restore indexes larger than 3072 bytes.

> - Fix the issue that PITR fails to restore indexes larger than 3072 bytes #58430 @ YuJuncen

https://docs.pingcap.com/tidb/stable/release-7.5.7

### TiCDC changefeed stuck-state issue with external storage fixed

- Affects: TiCDC changefeeds using external storage as the downstream

> - Fix the issue that the changefeed might get stuck when using external storage as the downstream #9162 @ asddongmen

https://docs.pingcap.com/tidb/stable/release-7.5.7

### TiCDC Kafka changefeed stuck-state issue fixed

- Affects: TiCDC changefeeds with Kafka downstreams

When replication traffic exceeded the downstream Kafka traffic threshold, a changefeed could get stuck. This issue is fixed.

> - Fix the issue that the changefeed might get stuck after the replication traffic exceeds the traffic threshold of the downstream Kafka #12110 @ 3AceShowHand

https://docs.pingcap.com/tidb/stable/release-7.5.7

### TiCDC changefeed pause stuck-state issue fixed

- Affects: TiCDC changefeed pauses using `--overwrite-checkpoint-ts`

Before this fix, the pause command could leave the changefeed stuck. That behavior is fixed.

> - Fix the issue that using the --overwrite-checkpoint-ts parameter in the changefeed pause command might cause the changefeed to be stuck #12055 @ hongyunyan

https://docs.pingcap.com/tidb/stable/release-7.5.7

### TiCDC event filter panic fixed

- Affects: TiCDC event filters on tables with virtual columns

Before this release, evaluating an event filter expression on a table with virtual columns could cause a panic.

> - Fix the issue that evaluating event filter expressions on tables containing virtual columns might cause a panic #12206 @ lidezhu

https://docs.pingcap.com/tidb/stable/release-7.5.7

### Case-sensitive name matching fixed in TiCDC dispatcher configuration

- Affects: TiCDC dispatcher configurations with column or index names affected by case-sensitive matching (Inferred)

The issue occurred during case-sensitive matching of column and index names; the resulting matching rules are not specified.

> - Fix the issue of case-sensitive matching for column and index names in the dispatcher configuration #12103 @ wk989898

https://docs.pingcap.com/tidb/stable/release-7.5.7

### TiDB Lightning stuck-import issue fixed

- Affects: TiDB Lightning imports of Parquet files from cloud storage into TiDB

When importing Parquet files from cloud storage into TiDB, TiDB Lightning could get stuck for several hours.

> - Fix the issue that TiDB Lightning can get stuck for several hours when importing Parquet files from cloud storage into TiDB #60224 @ joechenrh

https://docs.pingcap.com/tidb/stable/release-7.5.7

### TiDB Lightning timeout error fixed

- Affects: TiDB Lightning making RPC requests to TiKV

When an RPC request to TiKV timed out, TiDB Lightning returned the `context deadline exceeded` error.

> - Fix the issue that TiDB Lightning returns the context deadline exceeded error when the RPC request to TiKV times out #61326 @ OliverS929

https://docs.pingcap.com/tidb/stable/release-7.5.7

### Excessive TSDB memory consumption fixed

- Affects: TSDB handling high-cardinality time-series data

When TSDB handled high-cardinality time-series data, it consumed too much memory; that excessive use is fixed in this release.

> - Fix the issue that TSDB consumes too much memory under high cardinality of time series data, and provide a memory configuration option for TSDB #295 @ mornyx

https://docs.pingcap.com/tidb/stable/release-7.5.7

### Excessive TSDB memory use fixed

- Affects: TSDB deployments

Under high cardinality of time-series data, TSDB could consume too much memory. This release also provides a memory configuration option for TSDB.

> - Fix the issue that TSDB consumes too much memory under high cardinality of time series data, and provide a memory configuration option for TSDB #295 @ mornyx

https://docs.pingcap.com/tidb/stable/release-7.5.7

## Enhancements (11)

- Online flow control configuration support
- FSM schedule-related metrics added
- Raftstore I/O duration metrics added to the TiKV dashboard
- TiKV automatic-compaction settings added
- TiDB flow control interfaces for Region splitting and data ingestion
- TiKV write-speed monitoring metric for index addition
- Constant folding of `IS NULL` checks on `NOT NULL` columns to FALSE
- Constant propagation for more JOIN types
- Per-thread memory usage metrics in TiKV
- Slow logs for peer and store messages
- Additional Go runtime metrics for PD in Prometheus

## Other changes (43)

- Slow `AwakenRegions` message handling improved
- Slow handling of `CompactedEvent(RocksCompactedEvent)` addressed
- AWS storage credential retrieval fixed for IMDSv2
- Peer clearing optimized during scaling
- Faster scattering of large numbers of empty regions
- Writes continue during SST ingestion in snapshot and region-destroy paths
- Add-index flow control corrected for concurrent SST ingests
- KVDB I/O jitter detection and recovery for separate mount paths
- GC compaction prioritizes highest-ranked regions
- Compaction-score calculation considers the GC safe point
- Invalid UTF-8 SQL encoding failure addressed
- Async snapshot duration tail-latency optimization
- GC-oriented compaction moved into `gc_worker`
- tikv/tikv#18799
- Data distribution lookup optimized for simple queries on large tables
- DML locking optimized during DDL execution
- Improved control for `tidb_opt_ordering_index_selectivity_threshold`
- Temporary index merging performance improved
- TiKV compaction processes data segments in reclamation-efficiency order
- TiKV async snapshot and write tail latency optimized
- Faster Region Merge with empty tables and small Regions
- `CompactedEvent` handling moved to the split-check worker
- TiKV skips `get_sst_key_ranges` during slow SST ingest
- TiKV restart access delay optimized
- Ingest-free stale-range cleanup during region balancing
- Offline-store data retained during scale-in
- Reduced contention in Raft Engine `fetch_entries_to`
- Reduced unnecessary error logging in PD
- Improved observability for TiFlash OOM risks in wide tables
- Increased retry limit for TiFlash storage snapshots
- MPP plan selection with visible generated-column indexes fixed
- Redundant IndexLookup log entries removed
- TiDB statistics loading fixed for `BIT` tables
- TiDB statistics loading for BIT columns fixed
- Empty parentheses in TiDB cost formulas fixed
- TiDB index range construction with NULL fixed
- TiDB fixes ANALYZE subtask concurrency overruns
- TiDB background panic logging fixed
- Plan Cache skipping fixed for predicate simplification
- `lease` default value fixed
- TiFlash query-throttling issue fixed
- TiFlash schema mismatch error in rollup queries fixed
- BR table-existence check for `-f` filtering fixed

---

Summaries are produced by AI, so they can be wrong, and they are provided without warranty.
