缘起
最近浏览<<Go微服务实战>> (刘金亮, 2021.1)
本系列笔记拟采纳golang练习之
Saga模式
- saga模式将分布式长事务切分为一系列独立短事务
- 每个短事务是可通过弥补动作进行撤销的
- 事务动作和补动作偿都是幂等的, 容许反复执行而不会有副作用
Saga由一系列的子事务“Ti”组成,每个Ti都有对应的弥补“Ci”,当Ti呈现问题时Ci用于解决Ti执行带来的问题。能够通过上面的两个公式了解Saga模式。T = T1 T2 … TnT = TCTSaga模式的核心理念是防止应用长期持有锁(如14.2.2节介绍的两阶段提交)的长事务,而应该将事务切分为一组按序顺次提交的短事务,Saga模式满足ACD(原子性、一致性、持久性)特色。摘自 <<Go微服务实战>> 刘金亮, 2021.1
指标
- 为实现saga模式的分布式事务, 先撸一个pub/sub事务音讯队列服务
事务音讯队列服务的功能性要求
- 音讯不会失落: 音讯的长久化
- 音讯的唯一性: 要求每个音讯有全局ID和子事务ID
- 确保投递胜利: 投递队列长久化, 投递状态长久化, 失败重试
子目标(Day 6)
终于能够进行根本功能测试了
- 订阅接口测试: /subscribe
- 公布接口测试: /publish
- 告诉接口测试: /notify
重构
- database.go: 重构DDL语句, 以兼容sqlx的StructScan字段映射
- 所有Queryx返回的rows, 必须加上defer rows.Close()
- 增加大量的过程诊断日志
- tLiveMsgSource: fix handleMsgPublished未接收数据就return的bug
单元测试
mqs_test.go, 顺次清空数据库, 测试订阅接口, 公布接口和告诉接口, 并诊断过程日志和数据库记录变动.
package sagaimport ( "bytes" "encoding/json" "errors" "fmt" "github.com/jmoiron/sqlx" "io/ioutil" "learning/gooop/saga/mqs/cmd" "learning/gooop/saga/mqs/database" "learning/gooop/saga/mqs/logger" "learning/gooop/saga/mqs/models" "net/http" "sync" "testing" "time")var gRunOnce sync.Oncefunc fnBootMQS() { gRunOnce.Do(func() { // boot mqs go cmd.BootMQS() // wait for mqs up time.Sleep(1 * time.Second) })}func fnAssertTrue (t *testing.T, b bool, msg string) { if !b { t.Fatal(msg) }}func Test_MQS(t *testing.T) { // prepare mqs fnClearDB(t) fnBootMQS() // subscribe fnTestSubscribe(t) t.Log("passed fnTestSubscribe") // publish and notify fnTestPublishAndNotify(t) t.Log("passed fnTestPublishAndNotify")}func fnTestPublishAndNotify(t *testing.T) { t.Log("testing fnTestPublishAndNotify") msg := &models.TxMsg{ GlobalID: "test-global-id", SubID: "test-sub-id", SenderID: "test-sender-id", Topic: "test-topic", Content: "test-content", } fnPost(t, msg, "http://localhost:3333/publish") // check log fnAssertTrue(t, logger.Count("handlers.Publish, msg=test-global-id")==1, "expecting log: handlers.Publish, msg=test-global-id") // check notify time.Sleep(100 * time.Millisecond) fnAssertTrue(t, logger.Count("tLiveMsgSource.handleMsgPublished, clientID=test-client, msg=test-global-id")==1, "expecting log: tLiveMsgSource.handleMsgPublished") fnAssertTrue(t, logger.Count("tDeliveryWorker.afterDeliverySuccess, done, id=test-client, msg=test-global-id")==1, "expecting log: tDeliveryWorker.afterDeliverySuccess") fnAssertTrue(t, logger.Count("handlers.Notify, msg=")==1, "expecting log: handlers.Notify, msg=") // check success queue fnAssertTrue(t, fnDBCount(t, "select count(1) from success_queue where ClientID='test-client'")==1, "expectiang db.success_queue.count == 1")}func fnClearDB(t *testing.T) { fnDBExec(t, "delete from subscriber") fnDBExec(t, "delete from tx_msg") fnDBExec(t, "delete from delivery_queue") fnDBExec(t, "delete from success_queue")}func fnDBCount(t *testing.T, sql string, args... interface{}) int { sum := []int{ 0 } err := database.DB(func(db *sqlx.DB) error { r,e := db.Queryx(sql, args...) if e != nil { return e } defer r.Close() if !r.Next() { return errors.New("empty rows") } e = r.Scan(&sum[0]) if e != nil { return e } return nil }) if err != nil { t.Fatal(err) } return sum[0]}func fnDBExec(t *testing.T, sql string, args... interface{}) int { rows := []int64{ 0 } err := database.DB(func(db *sqlx.DB) error { r,e := db.Exec(sql, args...) if e != nil { return e } rows[0], e = r.RowsAffected() if e != nil { return e } return nil }) if err != nil { t.Fatal(err) } return int(rows[0])}func fnTestSubscribe(t *testing.T) { t.Log("testing fnTestSubscribe") // clear subscriber fnDBExec(t, "delete from subscriber") msg := &models.SubscribeMsg{ ClientID: "test-client", Topic: "test-topic", NotifyUrl: "http://localhost:3333/notify", ExpireTime: time.Now().UnixNano() + int64(30*time.Second), } fnPost(t, msg, "http://localhost:3333/subscribe") // check log fnAssertTrue(t, logger.Count("handlers.Subscribe, event=subscriber.registered") == 1, "expecting event=subscriber.registered") // check db count := fnDBCount(t, "select count(1) as n from subscriber where ClientID=? and topic=?", msg.ClientID, msg.Topic) fnAssertTrue(t, count == 1, "expecting subscriber.count == 1")}func fnPost(t *testing.T, msg interface{}, url string) { body,_ := json.Marshal(msg) rsp, err := http.Post(url, "application/json;charset=utf-8", bytes.NewReader(body)) if err != nil { t.Fatal(err) } defer rsp.Body.Close() j, err := ioutil.ReadAll(rsp.Body) if err != nil { t.Fatal(err) } ok := &models.OkMsg{} err = json.Unmarshal(j, ok) if err != nil { t.Fatal(err) } fnAssertTrue(t, ok.OK, fmt.Sprintf("expecting replying ok from %s", url))}
测试输入
依据assert失败提醒, 逐渐排查诊断日志, 次要是sqlx字段映射谬误, 和未及时调用rows.Close()谬误.
$ go test -v mqs_test.go === RUN Test_MQSeventbus.Pub, event=system.boot, handler=gDeliveryService.handleBootEvent[GIN-debug] [WARNING] Creating an Engine instance with the Logger and Recovery middleware already attached.[GIN-debug] [WARNING] Running in "debug" mode. Switch to "release" mode in production. - using env: export GIN_MODE=release - using code: gin.SetMode(gin.ReleaseMode)[GIN-debug] GET /ping --> learning/gooop/saga/mqs/handlers.Ping (4 handlers)[GIN-debug] POST /subscribe --> learning/gooop/saga/mqs/handlers.Subscribe (4 handlers)[GIN-debug] POST /publish --> learning/gooop/saga/mqs/handlers.Publish (4 handlers)[GIN-debug] POST /notify --> learning/gooop/saga/mqs/handlers.Notify (4 handlers)[GIN-debug] Listening and serving HTTP on :3333tDeliveryService.beginCreatingWorkerstDeliveryService.beginCleanExpiredWorkers mqs_test.go:139: testing fnTestSubscribehandlers.Subscribe, event=subscriber.registered, msg=&{test-client test-topic http://localhost:3333/notify 1615868155394998875}eventbus.Pub, event=subscriber.registered, handler=gDeliveryService.handleSubscriberRegistered[GIN] 2021/03/16 - 12:15:25 | 200 | 8.118873ms | ::1 | POST "/subscribe"[GIN] 2021/03/16 - 12:15:25 | 200 | 8.214968ms | ::1 | POST "/subscribe"tDeliveryWorker.afterInitialLoad, clientID=test-client, rows=0database.DB, err=empty rows mqs_test.go:45: passed fnTestSubscribe mqs_test.go:54: testing fnTestPublishAndNotifyhandlers.Publish, msg=test-global-id/test-sub-id/test-topic, msgId=15[GIN] 2021/03/16 - 12:15:25 | 200 | 15.200109ms | ::1 | POST "/publish"[GIN] 2021/03/16 - 12:15:25 | 200 | 15.216578ms | ::1 | POST "/publish"handlers.Publish, pubLiveMsg 15handlers.Publish, pubLiveMsg, msgId=15, rows=1handlers.Publish, event=msg.published, clientID=test-client, msg=test-global-id/test-sub-id/http://localhost:3333/notifyeventbus.Pub, event=msg.published, handler=tLiveMsgSource.test-clienttLiveMsgSource.handleMsgPublished, clientID=test-client, msg=test-global-id/test-sub-id/test-topichandlers.Notify, msg=&{test-global-id test-sub-id 0 test-topic test-content}[GIN] 2021/03/16 - 12:15:25 | 200 | 48.38µs | ::1 | POST "/notify"[GIN] 2021/03/16 - 12:15:25 | 200 | 110.659µs | ::1 | POST "/notify"tDeliveryWorker.afterDeliverySuccess, done, id=test-client, msg=test-global-id/test-sub-id mqs_test.go:49: passed fnTestPublishAndNotify--- PASS: Test_MQS (1.18s)PASSok command-line-arguments 1.187s
(未完待续)