chaos: integrate mysql-8 and mariadb (pingcap#1367) (pingcap#1368)

.github/workflows/chaos-mesh.yml

@@ -106,33 +106,35 @@ jobs:
     #   run: |
     #     kind load docker-image dm:chaos --name dm-chaos
 
-    # Set up upstream MySQL instances
-    - name: Set up MySQL
+    # Set up upstream instances
+    - name: Set up sources
       run: |
-         kubectl apply -f $GITHUB_WORKSPACE/chaos/manifests/mysql.yaml
-         kubectl get -f $GITHUB_WORKSPACE/chaos/manifests/mysql.yaml
-         kubectl describe -f $GITHUB_WORKSPACE/chaos/manifests/mysql.yaml
-    - name: Wait for MySQL ready # kubectl wait --all not working
+         kubectl apply -f $GITHUB_WORKSPACE/chaos/manifests/sources.yaml
+         kubectl get -f $GITHUB_WORKSPACE/chaos/manifests/sources.yaml
+         kubectl describe -f $GITHUB_WORKSPACE/chaos/manifests/sources.yaml
+    - name: Wait for sources ready # kubectl wait --all not working
       run: |
-         kubectl wait --for=condition=Ready pod/mysql-0 --timeout=300s || true
+         kubectl wait --for=condition=Ready pod/mysql57-0 --timeout=300s || true
+         kubectl wait --for=condition=Ready pod/mysql8-0 --timeout=300s || true
+         kubectl wait --for=condition=Ready pod/mariadb-0 --timeout=300s || true
          sleep 10
-         kubectl wait --for=condition=Ready pod/mysql-1 --timeout=300s || true
          echo show pvc
-         kubectl get pvc -l app=mysql -o wide
+         kubectl get pvc -l app=sources -o wide
          echo show pv
          kubectl get pv -o wide
          echo show svc
-         kubectl get svc -l app=mysql -o wide
+         kubectl get svc -l app=sources -o wide
          echo show sts
-         kubectl get sts -l app=mysql -o wide
+         kubectl get sts -l app=sources -o wide
          echo show po
-         kubectl get po -l app=mysql -o wide
+         kubectl get po -l app=sources -o wide
          echo describe po
-         kubectl describe po -l app=mysql
+         kubectl describe po -l app=sources
          echo describe pvc
-         kubectl describe pvc -l app=mysql
-         kubectl wait --for=condition=Ready pod/mysql-0 --timeout=0s
-         kubectl wait --for=condition=Ready pod/mysql-1 --timeout=0s
+         kubectl describe pvc -l app=sources
+         kubectl wait --for=condition=Ready pod/mysql57-0 --timeout=0s
+         kubectl wait --for=condition=Ready pod/mysql8-0 --timeout=0s
+         kubectl wait --for=condition=Ready pod/mariadb-0 --timeout=0s
 
     # Set up downstream TiDB instance (deploy a TiDB with mockTiKV, not a TidbCluster managed by TiDB-operator)
     - name: Set up TiDB

chaos/cases/conf/source1.yaml

@@ -1,9 +1,9 @@
-source-id: "mysql-replica-01"
-enable-gtid: false
-enable-relay: false
+source-id: "replica-01"
+enable-gtid: true
+enable-relay: true
 
 from:
-  host: "mysql-0.mysql" # same namespace with MySQL
+  host: "mysql57-0.sources" # same namespace with MySQL
   user: "root"
   password: ""
   port: 3306

chaos/cases/conf/source2.yaml

@@ -1,9 +1,9 @@
-source-id: "mysql-replica-02"
-enable-gtid: true
-enable-relay: true
+source-id: "replica-02"
+enable-gtid: false
+enable-relay: false
 
 from:
-  host: "mysql-1.mysql" # same namespace with MySQL
+  host: "mysql8-0.sources" # same namespace with MySQL
   user: "root"
   password: ""
   port: 3306

chaos/cases/conf/source3.yaml

@@ -0,0 +1,9 @@
+source-id: "replica-03"
+enable-gtid: false
+enable-relay: true
+
+from:
+  host: "mariadb-0.sources" # same namespace with MySQL
+  user: "root"
+  password: ""
+  port: 3306

chaos/cases/conf/task-optimistic.yaml

@@ -2,7 +2,12 @@
 name: "task_optimistic"
 task-mode: all
 shard-mode: optimistic
-ignore-checking-items: ["auto_increment_ID"] # tables generated by go-sqlsmith may have auto increment ID
+# Display width specification for integer data types was deprecated in MySQL 8.0.17
+# create table tb(a int(12)) will be `int` in mysql:8.0.17+ but `int(12)` in mysql:5.7
+# so ignore sharding table check, seems has no effect on sync-diff-inspector
+# see https://dev.mysql.com/doc/relnotes/mysql/8.0/en/news-8-0-19.html
+ignore-checking-items: ["schema_of_shard_tables","auto_increment_ID"] # tables generated by go-sqlsmith may have auto increment ID
+timezone: "UTC"
 
 target-database:
   host: "tidb-0.tidb"
@@ -12,13 +17,19 @@ target-database:
 
 mysql-instances:
   -
-    source-id: "mysql-replica-01"
+    source-id: "replica-01"
     black-white-list:  "instance"
     mydumper-thread: 4
     loader-thread: 16
     syncer-thread: 16
   -
-    source-id: "mysql-replica-02"
+    source-id: "replica-02"
+    black-white-list:  "instance"
+    mydumper-thread: 4
+    loader-thread: 16
+    syncer-thread: 16
+  -
+    source-id: "replica-03"
     black-white-list:  "instance"
     mydumper-thread: 4
     loader-thread: 16

chaos/cases/conf/task-pessimistic.yaml

@@ -2,7 +2,12 @@
 name: "task_pessimistic"
 task-mode: all
 shard-mode: pessimistic
-ignore-checking-items: ["auto_increment_ID"] # tables generated by go-sqlsmith may have auto increment ID
+# Display width specification for integer data types was deprecated in MySQL 8.0.17
+# create table tb(a int(12)) will be `int` in mysql:8.0.17+ but `int(12)` in mysql:5.7
+# so ignore sharding table check, seems has no effect on sync-diff-inspector
+# see https://dev.mysql.com/doc/relnotes/mysql/8.0/en/news-8-0-19.html
+ignore-checking-items: ["schema_of_shard_tables","auto_increment_ID"] # tables generated by go-sqlsmith may have auto increment ID
+timezone: "UTC"
 
 target-database:
   host: "tidb-0.tidb"
@@ -12,13 +17,19 @@ target-database:
 
 mysql-instances:
   -
-    source-id: "mysql-replica-01"
+    source-id: "replica-01"
     black-white-list:  "instance"
     mydumper-thread: 4
     loader-thread: 16
     syncer-thread: 16
   -
-    source-id: "mysql-replica-02"
+    source-id: "replica-02"
+    black-white-list:  "instance"
+    mydumper-thread: 4
+    loader-thread: 16
+    syncer-thread: 16
+  -
+    source-id: "replica-03"
     black-white-list:  "instance"
     mydumper-thread: 4
     loader-thread: 16

chaos/cases/conf/task-single.yaml

@@ -1,6 +1,7 @@
 ---
 name: "task_single"
 task-mode: all
+timezone: "UTC"
 
 target-database:
   host: "tidb-0.tidb"
@@ -10,7 +11,7 @@ target-database:
 
 mysql-instances:
   -
-    source-id: "mysql-replica-01"
+    source-id: "replica-01"
     black-white-list:  "instance"
     mydumper-thread: 4
     loader-thread: 16

chaos/cases/config.go

@@ -33,6 +33,7 @@ type config struct {
 
 	Source1 config2.DBConfig `toml:"source-1" yaml:"source-1" json:"source-1"`
 	Source2 config2.DBConfig `toml:"source-2" yaml:"source-2" json:"source-2"`
+	Source3 config2.DBConfig `toml:"source-3" yaml:"source-3" json:"source-3"`
 	Target  config2.DBConfig `toml:"target" yaml:"target" json:"target"`
 }
 
@@ -49,16 +50,21 @@ func newConfig() *config {
 	fs.IntVar(&cfg.MasterCount, "master-count", 3, "expect count of dm-master")
 	fs.IntVar(&cfg.WorkerCount, "worker-count", 3, "expect count of dm-worker")
 
-	fs.StringVar(&cfg.Source1.Host, "source1.host", "mysql-0.mysql", "host of the first upstream source")
+	fs.StringVar(&cfg.Source1.Host, "source1.host", "mysql57-0.sources", "host of the first upstream source")
 	fs.IntVar(&cfg.Source1.Port, "source1.port", 3306, "port of the first upstream source")
 	fs.StringVar(&cfg.Source1.User, "source1.user", "root", "user of the first upstream source")
 	fs.StringVar(&cfg.Source1.Password, "source1.password", "", "password of the first upstream source")
 
-	fs.StringVar(&cfg.Source2.Host, "source2.host", "mysql-1.mysql", "host of the second upstream source")
+	fs.StringVar(&cfg.Source2.Host, "source2.host", "mysql8-0.sources", "host of the second upstream source")
 	fs.IntVar(&cfg.Source2.Port, "source2.port", 3306, "port of the second upstream source")
 	fs.StringVar(&cfg.Source2.User, "source2.user", "root", "user of the second upstream source")
 	fs.StringVar(&cfg.Source2.Password, "source2.password", "", "password of the second upstream source")
 
+	fs.StringVar(&cfg.Source3.Host, "source3.host", "mariadb-0.sources", "host of the third upstream source")
+	fs.IntVar(&cfg.Source3.Port, "source3.port", 3306, "port of the third upstream source")
+	fs.StringVar(&cfg.Source3.User, "source3.user", "root", "user of the third upstream source")
+	fs.StringVar(&cfg.Source3.Password, "source3.password", "", "password of the third upstream source")
+
 	fs.StringVar(&cfg.Target.Host, "target.host", "tidb-0.tidb", "host of the downstream target")
 	fs.IntVar(&cfg.Target.Port, "target.port", 4000, "port of the downstream target")
 	fs.StringVar(&cfg.Target.User, "target.user", "root", "user of the downstream target")
@@ -85,8 +91,10 @@ func (c *config) adjust() {
 	// `ALTER TABLE .* ADD COLUMN (.* TIMESTAMP)` will have different default value
 	c.Source1.Session = map[string]string{"sql_mode": "", "explicit_defaults_for_timestamp": "on"}
 	c.Source2.Session = map[string]string{"sql_mode": "", "explicit_defaults_for_timestamp": "on"}
+	c.Source3.Session = map[string]string{"sql_mode": ""} // explicit_defaults_for_timestamp enabled when deploy it
 
 	c.Source1.Adjust()
 	c.Source2.Adjust()
+	c.Source3.Adjust()
 	c.Target.Adjust()
 }

chaos/cases/main.go

@@ -90,15 +90,15 @@ func main() {
 		log.L().Error("fail to check members ready", zap.Error(err)) // only log a error, still try to do other things.
 	}
 
-	// create two sources.
+	// create sources.
 	err = createSources(ctx, masterCli, cfg)
 	if err != nil {
 		log.L().Error("fail to create source", zap.Error(err))
 		os.Exit(2)
 	}
 
 	// set upstream and downstream instances state.
-	err = setInstancesState(ctx, cfg.Target, cfg.Source1, cfg.Source2)
+	err = setInstancesState(ctx, cfg.Target, cfg.Source1, cfg.Source2, cfg.Source3)
 	if err != nil {
 		log.L().Error("fail to set instances state", zap.Error(err))
 		os.Exit(2)
@@ -109,7 +109,7 @@ func main() {
 	defer cancel3()
 
 	// run tests cases
-	err = runCases(ctx3, masterCli, cfg.ConfigDir, cfg.Target, cfg.Source1, cfg.Source2)
+	err = runCases(ctx3, masterCli, cfg.ConfigDir, cfg.Target, cfg.Source1, cfg.Source2, cfg.Source3)
 	if err != nil {
 		log.L().Error("run cases failed", zap.Error(err))
 		os.Exit(2)

chaos/cases/source.go

@@ -25,10 +25,11 @@ import (
 )
 
 // createSources does `operate-source create` operation for two sources.
-// NOTE: we put two source config files (`source1.yaml` and `source2.yaml`) in `conf` directory.
+// NOTE: we put source config files in `conf` directory.
 func createSources(ctx context.Context, cli pb.MasterClient, cfg *config) error {
 	s1Path := filepath.Join(cfg.ConfigDir, "source1.yaml")
 	s2Path := filepath.Join(cfg.ConfigDir, "source2.yaml")
+	s3Path := filepath.Join(cfg.ConfigDir, "source3.yaml")
 
 	s1Content, err := ioutil.ReadFile(s1Path)
 	if err != nil {
@@ -38,19 +39,28 @@ func createSources(ctx context.Context, cli pb.MasterClient, cfg *config) error
 	if err != nil {
 		return err
 	}
+	s3Content, err := ioutil.ReadFile(s3Path)
+	if err != nil {
+		return err
+	}
 
 	cfg1 := config2.NewSourceConfig()
 	cfg2 := config2.NewSourceConfig()
+	cfg3 := config2.NewSourceConfig()
 	if err = cfg1.ParseYaml(string(s1Content)); err != nil {
 		return err
 	}
 	if err = cfg2.ParseYaml(string(s2Content)); err != nil {
 		return err
 	}
+	if err = cfg3.ParseYaml(string(s3Content)); err != nil {
+		return err
+	}
 
 	// replace DB config.
 	cfg1.From = cfg.Source1
 	cfg2.From = cfg.Source2
+	cfg3.From = cfg.Source3
 	s1Content2, err := cfg1.Yaml()
 	if err != nil {
 		return err
@@ -59,10 +69,14 @@ func createSources(ctx context.Context, cli pb.MasterClient, cfg *config) error
 	if err != nil {
 		return err
 	}
+	s3Content3, err := cfg3.Yaml()
+	if err != nil {
+		return err
+	}
 
 	resp, err := cli.OperateSource(ctx, &pb.OperateSourceRequest{
 		Op:     pb.SourceOp_StartSource,
-		Config: []string{s1Content2, s2Content2},
+		Config: []string{s1Content2, s2Content2, s3Content3},
 	})
 	if err != nil {
 		return err

chaos/cases/task.go

@@ -37,9 +37,9 @@ import (
 const (
 	tableCount      = 10               // tables count in schema.
 	fullInsertCount = 100              // `INSERT INTO` count (not rows count) for each table in full stage.
-	diffCount       = 30               // diff data check count
-	diffInterval    = 10 * time.Second // diff data check interval
-	incrRoundTime   = 20 * time.Second // time to generate incremental data in one round
+	diffCount       = 10               // diff data check count
+	diffInterval    = 20 * time.Second // diff data check interval
+	incrRoundTime   = 10 * time.Second // time to generate incremental data in one round
 )
 
 // task is a data migration task test case with one or more sources.
@@ -343,7 +343,7 @@ func (t *task) genIncrData(ctx context.Context) (err error) {
 		default:
 		}
 
-		if rand.Intn(1000) < 10 {
+		if rand.Intn(3000) < 10 {
 			query, err = randDDL(t.ss)
 			if err != nil {
 				return err

chaos/manifests/dm-worker.yaml

@@ -23,7 +23,7 @@ spec:
     matchLabels:
       app: dm-worker
   serviceName: dm-worker
-  replicas: 3 # 3 DM-worker instances
+  replicas: 4 # 4 DM-worker instances
   podManagementPolicy: Parallel
   template:
     metadata: