tests(ticdc): add lossy_ddl test (#8859)

close #8686

tests/integration_tests/lossy_ddl/data/prepare.sql

@@ -0,0 +1,171 @@
+DROP DATABASE IF EXISTS `lossy_ddl`;
+CREATE DATABASE `lossy_ddl`;
+USE `lossy_ddl`;
+
+-- int -> unsigned int
+CREATE TABLE example1
+(
+    id INT PRIMARY KEY,
+    b  INT
+);
+INSERT INTO example1 (id, b)
+VALUES (1, 1);
+ALTER TABLE example1 MODIFY COLUMN b INT UNSIGNED;
+
+-- int -> varchar
+CREATE TABLE example2
+(
+    id INT PRIMARY KEY,
+    b  INT
+);
+INSERT INTO example2 (id, b)
+VALUES (1, 1);
+ALTER TABLE example2 MODIFY COLUMN b VARCHAR (100);
+
+
+-- timestamp -> datetime
+CREATE TABLE example3
+(
+    id INT PRIMARY KEY,
+    b  TIMESTAMP
+);
+INSERT INTO example3 (id, b)
+VALUES (1, '2023-04-19 11:48:00');
+ALTER TABLE example3 MODIFY COLUMN b DATETIME;
+
+-- varchar(256) -> varchar(100)
+CREATE TABLE example4
+(
+    id INT PRIMARY KEY,
+    b  VARCHAR(256)
+);
+INSERT INTO example4 (id, b)
+VALUES (1, '2023-04-19 11:48:00');
+ALTER TABLE example4 MODIFY COLUMN b VARCHAR (100);
+
+-- Drop column
+CREATE TABLE example5
+(
+    id INT PRIMARY KEY,
+    b  INT
+);
+INSERT INTO example5 (id, b)
+VALUES (1, -1);
+ALTER TABLE example5 DROP COLUMN b;
+
+-- Add column
+CREATE TABLE example6
+(
+    id INT PRIMARY KEY,
+    b  INT
+);
+INSERT INTO example6 (id, b)
+VALUES (1, -1);
+ALTER TABLE example6
+    ADD COLUMN c INT;
+
+-- Modify collation
+CREATE TABLE example7
+(
+    id INT PRIMARY KEY,
+    b  VARCHAR(256) CHARACTER SET utf8 COLLATE utf8_general_ci
+);
+INSERT INTO example7 (id, b)
+VALUES (1, '2023-04-19 11:48:00');
+ALTER TABLE example7 MODIFY COLUMN b VARCHAR (256) CHARACTER SET utf8mb4 COLLATE utf8mb4_bin;
+
+-- Drop partition
+CREATE TABLE example8
+(
+    id INT PRIMARY KEY,
+    b  INT
+) PARTITION BY RANGE (id) (
+    PARTITION b0 VALUES LESS THAN (0),
+    PARTITION b1 VALUES LESS THAN MAXVALUE
+);
+INSERT INTO example8 (id, b)
+VALUES (-1, -1);
+INSERT INTO example8 (id, b)
+VALUES (1, 1);
+ALTER TABLE example8 DROP PARTITION b0;
+
+-- varchar(256) -> varchar(300)
+CREATE TABLE example9
+(
+    id INT PRIMARY KEY,
+    b  VARCHAR(256)
+);
+INSERT INTO example9 (id, b)
+VALUES (1, '2023-04-19 11:48:00');
+ALTER TABLE example9 MODIFY COLUMN b VARCHAR (300);
+
+
+-- double -> float
+CREATE TABLE example10
+(
+    id INT PRIMARY KEY,
+    b  DOUBLE
+);
+INSERT INTO example10 (id, b)
+VALUES (1, 1.0);
+ALTER TABLE example10 MODIFY COLUMN b FLOAT;
+
+-- bigint -> int
+CREATE TABLE example11
+(
+    id BIGINT PRIMARY KEY,
+    b  BIGINT
+);
+INSERT INTO example11 (id, b)
+VALUES (1, 1);
+ALTER TABLE example11 MODIFY COLUMN b INT;
+
+-- longtext -> varchar(100)
+CREATE TABLE example12
+(
+    id INT PRIMARY KEY,
+    b  LONGTEXT
+);
+INSERT INTO example12 (id, b)
+VALUES (1, '2023-04-19 11:48:00');
+ALTER TABLE example12 MODIFY COLUMN b VARCHAR (100);
+
+-- Enum('a', 'b', 'c') -> Enum('a', 'b')
+CREATE TABLE example13
+(
+    id INT PRIMARY KEY,
+    b  ENUM('a', 'b', 'c')
+);
+INSERT INTO example13 (id, b)
+VALUES (1, 'a');
+ALTER TABLE example13 MODIFY COLUMN b ENUM('a', 'b');
+
+-- Set No STRICT_TRANS_TABLES
+SET
+@@SESSION.sql_mode = 'NO_ENGINE_SUBSTITUTION';
+
+-- varchar(256) -> varchar(1) with a long value.
+CREATE TABLE example14
+(
+    id INT PRIMARY KEY,
+    b  VARCHAR(256)
+);
+INSERT INTO example14 (id, b)
+VALUES (1, '2023-04-19 11:48:00');
+ALTER TABLE example14 MODIFY COLUMN b VARCHAR (1);
+
+-- int -> unsigned int with a negative value.
+CREATE TABLE example15
+(
+    id INT PRIMARY KEY,
+    b  INT
+);
+INSERT INTO example15 (id, b)
+VALUES (1, -1);
+ALTER TABLE example15 MODIFY COLUMN b INT UNSIGNED;
+
+CREATE TABLE `finish_mark`
+(
+    id   INT PRIMARY KEY,
+    name VARCHAR(20)
+);

tests/integration_tests/lossy_ddl/run.sh

@@ -0,0 +1,57 @@
+#!/bin/bash
+
+set -eu
+
+CUR=$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)
+source $CUR/../_utils/test_prepare
+WORK_DIR=$OUT_DIR/$TEST_NAME
+CDC_BINARY=cdc.test
+SINK_TYPE=$1
+
+MAX_RETRIES=10
+
+# Because we want the lossy DDL to not cause any data updates, so we can check the
+# data in the blackhole sink to see if any row is updated.
+function check_lossy_ddl() {
+	# Check finish_mark is written to the log.
+	is_finish_mark_exist=$(grep "BlackHoleSink: DDL Event" "$1/cdc.log" | grep -c "finish_mark")
+	if [[ "$is_finish_mark_exist" -ne 1 ]]; then
+		echo "can't found finish mark"
+		exit 1
+	fi
+
+	row_logs=$(grep "BlackHoleSink: WriteEvents" "$1/cdc.log")
+	echo $row_logs
+	row_logs_count=$(grep "BlackHoleSink: WriteEvents" -c "$1/cdc.log")
+	if [[ "$row_logs_count" -ne 16 ]]; then
+		echo "can't found 16 row logs, got $row_logs_count"
+		exit 1
+	fi
+}
+
+function run() {
+	# Use blackhole sink to check if the DDL is lossy.
+	# So no need to run this test for other sinks.
+	if [ "$SINK_TYPE" != "storage" ]; then
+		return
+	fi
+
+	rm -rf $WORK_DIR && mkdir -p $WORK_DIR
+	start_tidb_cluster --workdir $WORK_DIR
+	cd $WORK_DIR
+
+	pd_addr="http://$UP_PD_HOST_1:$UP_PD_PORT_1"
+	SINK_URI="blackhole://"
+
+	run_cdc_server --workdir $WORK_DIR --binary $CDC_BINARY --addr "127.0.0.1:8300" --pd $pd_addr
+	cdc cli changefeed create --pd=$pd_addr --sink-uri="$SINK_URI"
+
+	run_sql_file $CUR/data/prepare.sql ${UP_TIDB_HOST} ${UP_TIDB_PORT}
+
+	ensure $MAX_RETRIES check_lossy_ddl $WORK_DIR
+	cleanup_process $CDC_BINARY
+}
+
+trap stop_tidb_cluster EXIT
+run $*
+echo "[$(date)] <<<<<< run test case $TEST_NAME success! >>>>>>"

tests/integration_tests/run_group.sh

@@ -47,7 +47,7 @@ groups=(
 	["G12"]='changefeed_fast_fail tidb_mysql_test server_config_compatibility'
 	["G13"]='resourcecontrol processor_etcd_worker_delay'
 	["G14"]='batch_update_to_no_batch gc_safepoint changefeed_pause_resume'
-	["G15"]='cli simple cdc_server_tips changefeed_resume_with_checkpoint_ts'
+	["G15"]='cli simple cdc_server_tips changefeed_resume_with_checkpoint_ts lossy_ddl'
 	["G16"]='processor_err_chan resolve_lock move_table autorandom'
 	["G17"]='ddl_attributes many_pk_or_uk capture_session_done_during_task'
 	["G18"]='tiflash new_ci_collation_without_old_value region_merge common_1'