workflow other error can't backoff retery and RetryInterval is invalid

[nogolang]

workflow 模式下，当分支返回其他错误，不会进行指数退避重试，而是会采用固定重试

	AddCommand("http_workflow_saga_rollback", func() string {
		wfName := "wf_saga_rollback"
		err := workflow.Register(wfName, func(wf *workflow.Workflow, data []byte) error {
			req := MustUnmarshalReqHTTP(data)
			_, err := wf.NewBranch().OnRollback(func(bb *dtmcli.BranchBarrier) error {
				_, err := wf.NewRequest().SetBody(req).Post(busi.Busi + "/TransOutRevert")
				return err
			}).NewRequest().SetBody(req).Post(busi.Busi + "/TransOut")
			if err != nil {
				return err
			}
			_, err = wf.NewBranch().OnRollback(func(bb *dtmcli.BranchBarrier) error {
				_, err := wf.NewRequest().SetBody(req).Post(busi.Busi + "/TransInRevert")
				return err
			}).NewRequest().SetBody(req).Post(busi.Busi + "/TransIn")
			if err != nil {
				return err
			}
			return nil
		})
		logger.FatalIfError(err)

                //只会固定时间重试，固定为10s
		req := &busi.ReqHTTP{Amount: 30, TransOutResult: "ERROR"}
		//req := &busi.ReqHTTP{Amount: 30, TransInResult: dtmcli.ResultFailure}

		gid := shortuuid.New()
		err = workflow.Execute(wfName, gid, dtmimp.MustMarshal(req))
		logger.Infof("result is: %v", err)
		return gid
	})

因为上一个问题，指数退避变为了固定重试，所以我想调整RetryInterval，但是发现 虽然在ui 里改变了间隔时间
但是实际重试间隔依旧为10s左右

	err := workflow.Register2(wfName, func(wf *workflow.Workflow, data []byte) ([]byte, error) {
		var paramData Param
		err := json.Unmarshal(data, &paramData)
		if err != nil {
			return nil, err
		}
		log.Println("进入了Register")

		wf.NewBranch().OnRollback(func(bb *dtmcli.BranchBarrier) error {
			_, err := wf.NewRequest().SetBody(model.TransRequestArgsHTTP{
				Uid:    paramData.AUid,
				Amount: paramData.Amount,
			}).Post(baseUrl + "/api/SagaTransOutRollBack")
			if err != nil {
				return err
			}
			return nil
		})
		res, err := wf.NewRequest().SetBody(&model.TransRequestArgsHTTP{
			Uid:    paramData.AUid,
			Amount: -paramData.Amount,
		}).Post(baseUrl + "/api/SagaTransOut")
		if err != nil {
			return nil, err
		}
		wf.NewBranch().OnRollback(func(bb *dtmcli.BranchBarrier) error {
			_, err := wf.NewRequest().SetBody(&model.TransRequestArgsHTTP{
				Uid:    paramData.BUid,
				Amount: -paramData.Amount,
			}).Post(baseUrl + "/api/SagaTransInRollBack")
			if err != nil {
				return err
			}
			return nil
		})
		res, err = wf.NewRequest().SetBody(&model.TransRequestArgsHTTP{Uid: paramData.BUid, Amount: paramData.Amount}).Post(baseUrl + "/api/SagaTransIn")
		if err != nil {
			return nil, err
		}
		return res.Body(), nil
	}, func(wf *workflow.Workflow) {

		//因为上一个问题，指数退避变为了固定重试，所以我想调整RetryInterval
		wf.RetryInterval = 5
	})

[zyf873753]

最近也发现了这个问题：

workflow的处理逻辑里并不会处理报错，而是直接返回上层，这里连成功的情况都不会处理

dtm/dtmsvr/trans_type_workflow.go

Line 41 in 4a663a1

return t.getURLResult(ctx, t.QueryPrepared, "00", cmc.Name, data)

它完全依靠服务完成workflow事务后，发来的/api/dtmsvr/submit来设置成功状态，这个感觉是没问题的

dtm/dtmsvr/api.go

Line 25 in 4a663a1

t.changeStatus(t.ReqExtra["status"], withRollbackReason(t.ReqExtra["rollback_reason"]), withResult(t.ReqExtra["result"]))

saga的逻辑会根据结果来处理超时时间

dtm/dtmsvr/trans_type_saga.go

Line 136 in 4a663a1

err = t.execBranch(ctx, &branches[i], i)

dtm/dtmsvr/trans_status.go

Line 242 in 4a663a1

t.touchCronTime(cronBackoff, 0)

希望 @yedf2 大佬可以完善一下workflow，是不是只要把saga的逻辑复制到workflow应该就可以

if err == nil && (time.Since(t.lastTouched)+NowForwardDuration >= 1500*time.Millisecond ||
	t.NextCronInterval > conf.RetryInterval && t.NextCronInterval > t.RetryInterval) {
	t.touchCronTime(cronReset, 0)
} else if err == dtmimp.ErrOngoing {
	t.touchCronTime(cronKeep, 0)
} else if err != nil {
	t.touchCronTime(cronBackoff, 0)
	v := t.NextCronInterval / t.getNextCronInterval(cronReset)
	retryCount := int64(math.Log2(float64(v)))
	logger.Debugf("origin: %d v: %d retryCount: %d", t.getNextCronInterval(cronReset), v, retryCount)
	if retryCount >= conf.AlertRetryLimit && conf.AlertWebHook != "" {
		_, err2 := dtmcli.GetRestyClient().R().SetBody(gin.H{
			"gid":         t.Gid,
			"status":      t.Status,
			"branch":      branch.URL,
			"error":       err.Error(),
			"retry_count": retryCount,
		}).Post(conf.AlertWebHook)
		if err2 != nil {
			logger.Errorf("alerting webhook error: %v", err2)
		}
	}
}