diff --git a/import-example-data.md b/import-example-data.md
index 49136bdb2a883..2ecadba9e2816 100644
--- a/import-example-data.md
+++ b/import-example-data.md
@@ -44,6 +44,7 @@ CREATE TABLE trips (
 You can import files individually using the example `LOAD DATA` command here, or import all files using the bash loop below:
 
 ```sql
+SET tidb_dml_batch_size = 20000;
 LOAD DATA LOCAL INFILE '2017Q1-capitalbikeshare-tripdata.csv' INTO TABLE trips
   FIELDS TERMINATED BY ',' ENCLOSED BY '"'
   LINES TERMINATED BY '\r\n'
@@ -61,8 +62,8 @@ end_station_number, end_station, bike_number, member_type);
 To import all `*.csv` files into TiDB in a bash loop:
 
 ```bash
-for FILE in `ls *.csv`; do
+for FILE in *.csv; do
  echo "== $FILE =="
- mysql bikeshare --local-infile=1 -e "LOAD DATA LOCAL INFILE '${FILE}' INTO TABLE trips FIELDS TERMINATED BY ',' ENCLOSED BY '\"' LINES TERMINATED BY '\r\n' IGNORE 1 LINES (duration, start_date, end_date, start_station_number, start_station, end_station_number, end_station, bike_number, member_type);"
+ mysql bikeshare --local-infile=1 -e "SET tidb_dml_batch_size = 20000; LOAD DATA LOCAL INFILE '${FILE}' INTO TABLE trips FIELDS TERMINATED BY ',' ENCLOSED BY '\"' LINES TERMINATED BY '\r\n' IGNORE 1 LINES (duration, start_date, end_date, start_station_number, start_station, end_station_number, end_station, bike_number, member_type);"
 done;
 ```
diff --git a/sql-statements/sql-statement-load-data.md b/sql-statements/sql-statement-load-data.md
index 6b0094ef865f8..73659d1218f8d 100644
--- a/sql-statements/sql-statement-load-data.md
+++ b/sql-statements/sql-statement-load-data.md
@@ -104,13 +104,16 @@ In the above example, `x'2c'` is the hexadecimal representation of the `,` chara
 
 ## MySQL compatibility
 
-* TiDB will by default commit every 20 000 rows. This behavior is similar to MySQL NDB Cluster, but not the default configuration with the InnoDB storage engine.
+This statement is understood to be fully compatible with MySQL. Any compatibility differences should be [reported via an issue](https://github.com/pingcap/tidb/issues/new/choose) on GitHub.
 
 > **Note:**
 >
-> Committing through splitting a transaction is at the expense of breaking the atomicity and isolation of the transaction. When performing this operation, you must ensure that there are **no other** ongoing operations on the table. When an error occurs, **manual intervention is required to check the consistency and integrity of the data**. Therefore, it is not recommended to use `LOAD DATA` on any tables which are actively being read from or written to.
+> In earlier releases of TiDB, `LOAD DATA` committed every 20000 rows. By default, TiDB now commits all rows in one transaction. This can result in the error `ERROR 8004 (HY000) at line 1: Transaction is too large, size: 100000058` after upgrading from TiDB 4.0 or earlier versions.
+>
+> The recommended way to resolve this error is to increase the `txn-total-size-limit` value in your `tidb.toml` file. If you are unable to increase this limit, you can also restore the previous behavior by setting [`tidb_dml_batch_size`](/system-variables.md#tidb_dml_batch_size) to `20000`.
 
 ## See also
 
 * [INSERT](/sql-statements/sql-statement-insert.md)
 * [Import Example Database](/import-example-data.md)
+* [TiDB Lightning](/tidb-lightning/tidb-lightning-overview.md)
diff --git a/system-variables.md b/system-variables.md
index 20da1a97c3a61..0bf0c66b417cf 100644
--- a/system-variables.md
+++ b/system-variables.md
@@ -271,6 +271,14 @@ Constraint checking is always performed in place for pessimistic transactions (d
 - Use a bigger value in OLAP scenarios, and a smaller value in OLTP scenarios.
 - For OLAP scenarios, the maximum value cannot exceed the number of CPU cores of all the TiKV nodes.
 
+### tidb_dml_batch_size
+
+- Scope: SESSION
+- Default value: 0
+- Example value: 20000
+- When this value is greater than `0`, TiDB will batch commit statements such as `INSERT` or `LOAD DATA` into smaller transactions. This reduces memory usage and helps ensure that the `txn-total-size-limit` is not reached by bulk modifications.
+- Only the value `0` provides ACID compliance. Setting this to any other value will break the atomicity and isolation guarantees of TiDB.
+
 ### tidb_enable_cascades_planner
 
 - Scope: SESSION | GLOBAL
@@ -819,7 +827,7 @@ SET tidb_slow_log_threshold = 200;
 
 - Scope: SESSION | GLOBAL
 - Default value: SYSTEM
-- This variable sets the sytem time zone. Values can be specified as either an offset such as '-8:00' or a named zone 'America/Los_Angeles'.
+- This variable sets the system time zone. Values can be specified as either an offset such as '-8:00' or a named zone 'America/Los_Angeles'.
 
 ### transaction_isolation