test(dm): add MariaDB source smoke test and next-gen integration test

Makefile

@@ -508,7 +508,8 @@ install_test_python_dep:
 	@echo "install python requirments for test"
 	pip install --user -q -r ./dm/tests/requirements.txt
 
-check_third_party_binary_for_dm : sync-diff-inspector
+check_third_party_binary_for_dm:
+	@which bin/sync_diff_inspector
 	@which bin/tidb-server
 	@which mysql
 	@which bin/minio

dm/tests/_utils/cluster_lib.sh

@@ -0,0 +1,87 @@
+#!/bin/bash
+# Cluster lifecycle operations for DM integration tests.
+#
+# Sourced by test_prepare. Provides functions to start, stop, and restart
+# downstream TiDB clusters in both classic and next-gen modes.
+#
+# Startup delegates to standalone scripts (which manage their own processes).
+# Cleanup runs in the test's shell (needs access to pgrep/kill).
+
+CUR_CLUSTER_LIB=$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)
+
+# ---------------------------------------------------------------------------
+# Cleanup
+# ---------------------------------------------------------------------------
+
+# Kill the port-4000 TiDB (SYSTEM TiDB on next-gen, unistore on classic).
+cleanup_tidb_server() {
+	local pattern='tidb-server.*-P 4000'
+	local pids
+	pids=$(pgrep -f "$pattern" || true)
+	echo "tidb-server on port 4000 pids=${pids:-none}"
+	if [ -n "$pids" ]; then
+		kill -HUP $pids 2>/dev/null || true
+	fi
+	for _ in $(seq 1 120); do
+		if ! pgrep -f "$pattern" >/dev/null 2>&1; then
+			echo "tidb-server on port 4000 already exit"
+			rm -f /tmp/*_tidb/*/tmp-storage/_dir.lock 2>/dev/null || true
+			return 0
+		fi
+		sleep 1
+	done
+	echo "tidb-server on port 4000 didn't exit in 120s"
+	pgrep -af "$pattern" || true
+	return 1
+}
+
+# Tear down the full downstream cluster.
+# Next-gen: only TiDB (preserve PD + TiKV + MinIO + tikv-worker).
+# Classic: kill everything + clean unistore data.
+cleanup_downstream_cluster() {
+	if [ "${NEXT_GEN:-}" = "1" ]; then
+		cleanup_tidb_server
+	else
+		killall -9 tidb-server 2>/dev/null || true
+		killall -9 tikv-server 2>/dev/null || true
+		killall -9 pd-server 2>/dev/null || true
+		wait_process_exit tidb-server
+		wait_process_exit tikv-server
+		wait_process_exit pd-server
+		rm -rf /tmp/tidb
+	fi
+}
+
+# ---------------------------------------------------------------------------
+# Startup
+# ---------------------------------------------------------------------------
+
+# Start or restart a single downstream TiDB.
+# Args: port password [config_file]
+run_tidb_server() {
+	"$CUR_CLUSTER_LIB/run_tidb_server" "$@"
+}
+
+# Start a full downstream cluster (PD + TiKV + TiDB).
+# Classic: single PD + TiKV + TiDB.
+# Next-gen: MinIO + PD + TiKV + tikv-worker + SYSTEM TiDB + user TiDB.
+# Args: work_dir
+run_downstream_cluster() {
+	if [ "${NEXT_GEN:-}" = "1" ]; then
+		"$CUR_CLUSTER_LIB/run_downstream_cluster_nextgen" "$@"
+	else
+		"$CUR_CLUSTER_LIB/run_downstream_cluster_classic" "$@"
+	fi
+}
+
+# Start a TLS-enabled downstream cluster.
+# Classic: full PD + TiKV + TiDB with TLS on separate ports.
+# Next-gen: restart only user TiDB with client-facing TLS.
+# Args: work_dir conf_dir cluster_ca cluster_cert cluster_key db_ca db_cert db_key
+run_downstream_cluster_with_tls() {
+	if [ "${NEXT_GEN:-}" = "1" ]; then
+		"$CUR_CLUSTER_LIB/run_downstream_cluster_with_tls_nextgen" "$@"
+	else
+		"$CUR_CLUSTER_LIB/run_downstream_cluster_with_tls_classic" "$@"
+	fi
+}

dm/tests/_utils/env_variables

@@ -1,10 +1,13 @@
 MYSQL_HOST1=${MYSQL_HOST1:-127.0.0.1}
 MYSQL_HOST2=${MYSQL_HOST2:-127.0.0.1}
+MARIADB_HOST1=${MARIADB_HOST1:-127.0.0.1}
 TIDB_HOST=${TIDB_HOST:-127.0.0.1}
 MYSQL_PORT1=${MYSQL_PORT1:-3306}
 MYSQL_PORT2=${MYSQL_PORT2:-3307}
+MARIADB_PORT1=${MARIADB_PORT1:-3308}
 MYSQL_PASSWORD1=${MYSQL_PASSWORD1:-123456}
 MYSQL_PASSWORD2=${MYSQL_PASSWORD2:-123456}
+MARIADB_PASSWORD1=${MARIADB_PASSWORD1:-123456}
 TIDB_PASSWORD=${TIDB_PASSWORD:-123456}
 
 TIDB_PORT=${TIDB_PORT:-4000}
@@ -34,3 +37,23 @@ SOURCE_ID2="mysql-replica-02"
 RESET_MASTER=${RESET_MASTER:-true}
 
 VERBOSE=${VERBOSE:-false}
+
+# Cluster endpoints. On next-gen, PD_ADDR etc. are always set because a real
+# TiKV cluster is required. On classic, they're only set by the individual
+# cluster scripts (run_downstream_cluster_classic) when a real cluster is
+# needed — most classic tests use unistore (no PD/TiKV).
+if [ "${NEXT_GEN:-}" = "1" ]; then
+	export PD_PEER_ADDR=${PD_PEER_ADDR:-"127.0.0.1:2380"}
+	export PD_ADDR=${PD_ADDR:-"127.0.0.1:2379"}
+	export TIKV_ADDR=${TIKV_ADDR:-"127.0.0.1:2016"}
+	export TIKV_STATUS_ADDR=${TIKV_STATUS_ADDR:-"127.0.0.1:2018"}
+	export MINIO_ADDR=${MINIO_ADDR:-"127.0.0.1:9000"}
+	export MINIO_ACCESS_KEY=${MINIO_ACCESS_KEY:-"minioadmin"}
+	export MINIO_SECRET_KEY=${MINIO_SECRET_KEY:-"minioadmin"}
+	export MINIO_BUCKET=${MINIO_BUCKET:-"next-gen-test"}
+	export TIKV_WORKER_ADDR=${TIKV_WORKER_ADDR:-"127.0.0.1:19000"}
+	export KEYSPACE_NAME=${KEYSPACE_NAME:-"SYSTEM"}
+
+	# Extra CLI flags for user TiDB.
+	export TIDB_EXTRA_ARGS="-keyspace-name ${KEYSPACE_NAME}"
+fi

dm/tests/_utils/ha_cases_lib.sh

@@ -7,6 +7,17 @@ ha_test2="ha_test2"
 master_ports=($MASTER_PORT1 $MASTER_PORT2 $MASTER_PORT3)
 worker_ports=($WORKER1_PORT $WORKER2_PORT $WORKER3_PORT $WORKER4_PORT $WORKER5_PORT)
 
+# print_debug_status dumps query-status for both tasks when check_sync_diff
+# fails, so the CI log shows what went wrong before the test exits.
+function print_debug_status() {
+	run_dm_ctl $WORK_DIR "127.0.0.1:$MASTER_PORT3" \
+		"query-status test" \
+		"fail me!" 1 &&
+		run_dm_ctl $WORK_DIR "127.0.0.1:$MASTER_PORT3" \
+			"query-status test2" \
+			"fail me!" 1 && exit 1
+}
+
 function load_data() {
 	port=$1
 	pswd=$2
@@ -135,13 +146,11 @@ function start_multi_tasks_cluster() {
 	check_rpc_alive $cur/../bin/check_worker_online 127.0.0.1:$WORKER5_PORT
 
 	echo "start DM task"
-
-	dmctl_start_task &
-	pid1=$!
-	dmctl_start_task "$cur/conf/dm-task2.yaml" &
-	pid2=$!
-
-	wait "$pid1" "$pid2"
+	# Run sequentially — parallel dmctl calls write the same log file
+	# ($workdir/dmctl.$ts.log) when executed in the same second, causing
+	# output corruption and false "result count mismatch" failures.
+	dmctl_start_task
+	dmctl_start_task "$cur/conf/dm-task2.yaml"
 }
 
 function cleanup() {

dm/tests/_utils/run_downstream_cluster → dm/tests/_utils/run_downstream_cluster_classic

@@ -1,19 +1,16 @@
 #!/usr/bin/env bash
-# tools to run a TiDB cluster
+# Start a classic (single-PD + single-TiKV + single-TiDB) downstream cluster.
 # parameter 1: work directory
 set -eux
+CUR=$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)
+source "$CUR/env_variables"
 WORK_DIR=$1
 
-export PD_PEER_ADDR="127.0.0.1:2380"
-export PD_ADDR="127.0.0.1:2379"
-
-export TIDB_IP="127.0.0.1"
-export TIDB_PORT="4000"
-export TIDB_ADDR="127.0.0.1:4000"
-
-export TIDB_STATUS_ADDR="127.0.0.1:10080"
-export TIKV_ADDR="127.0.0.1:2016"
-export TIKV_STATUS_ADDR="127.0.0.1:2018"
+# Classic cluster needs these — on next-gen they come from env_variables.
+export PD_PEER_ADDR=${PD_PEER_ADDR:-"127.0.0.1:2380"}
+export PD_ADDR=${PD_ADDR:-"127.0.0.1:2379"}
+export TIKV_ADDR=${TIKV_ADDR:-"127.0.0.1:2016"}
+export TIKV_STATUS_ADDR=${TIKV_STATUS_ADDR:-"127.0.0.1:2018"}
 
 start_pd() {
 	echo "Starting PD..."
@@ -44,7 +41,7 @@ EOF
 		i=$((i + 1))
 		if [ "$i" -gt 20 ]; then
 			echo 'Failed to start PD'
-			return 1
+			exit 1
 		fi
 		echo 'Waiting for PD ready...'
 		sleep 3
@@ -80,47 +77,18 @@ start_tikv() {
 		i=$((i + 1))
 		if [ "$i" -gt 20 ]; then
 			echo 'Failed to initialize TiKV cluster after 20 attempts'
-			return 1
+			exit 1
 		fi
 
 		echo 'Waiting for TiKV ready...'
 		sleep 5
 	done
 }
 
-start_tidb() {
-	echo "Starting TiDB..."
-	bin/tidb-server -V
-	bin/tidb-server \
-		-P 4000 \
-		--status 10080 \
-		--advertise-address="127.0.0.1" \
-		--store tikv \
-		--path "$PD_ADDR" \
-		--log-file "$WORK_DIR/tidb.log" &
-	sleep 5
-	# wait until TiDB is online...
-	i=0
-	while true; do
-		response=$(curl -s -o /dev/null -w "%{http_code}" "http://$TIDB_IP:10080/status" || echo "")
-		echo "curl response: $response"
-		if [ "$response" -eq 200 ]; then
-			echo 'Start TiDB success'
-			break
-		fi
-		i=$((i + 1))
-		if [ "$i" -gt 50 ]; then
-			echo 'Failed to start TiDB'
-			return 1
-		fi
-		echo 'Waiting for TiDB ready...'
-		sleep 3
-	done
-}
-
 start_pd
 start_tikv
-start_tidb
 
-mysql -uroot -h127.0.0.1 -P4000 --default-character-set utf8 -e "CREATE USER 'test'@'%' IDENTIFIED BY '123456';" || true
-mysql -uroot -h127.0.0.1 -P4000 --default-character-set utf8 -e "GRANT ALL PRIVILEGES ON *.* TO 'test'@'%' WITH GRANT OPTION;" || true
+# PD_ADDR is already exported above; run_tidb_server derives --store tikv
+# from it automatically.
+CUR_DIR=$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)
+"$CUR_DIR/run_tidb_server" 4000 123456