序

本文主要研究一下xxl-job-executor-go

Executor

//执行器
type Executor interface {
	//初始化
	Init(...Option)
	//日志查询
	LogHandler(handler LogHandler)
	//注册任务
	RegTask(pattern string, task TaskFunc)
	//运行任务
	RunTask(writer http.ResponseWriter, request *http.Request)
	//杀死任务
	KillTask(writer http.ResponseWriter, request *http.Request)
	//任务日志
	TaskLog(writer http.ResponseWriter, request *http.Request)
	//运行服务
	Run() error
}

Executor定义了Init、LogHandler、RegTask、RunTask、KillTask、TaskLog、Run方法

executor

type executor struct {
	opts    Options
	address string
	regList *taskList //注册任务列表
	runList *taskList //正在执行任务列表
	mu      sync.RWMutex
	log     Logger

	logHandler LogHandler //日志查询handler
}

executor定义了opts、address、regList、runList、mu、log、logHandler属性

Init

func (e *executor) Init(opts ...Option) {
	for _, o := range opts {
		o(&e.opts)
	}
	e.log = e.opts.l
	e.regList = &taskList{
		data: make(map[string]*Task),
	}
	e.runList = &taskList{
		data: make(map[string]*Task),
	}
	e.address = e.opts.ExecutorIp + ":" + e.opts.ExecutorPort
	go e.registry()
}

Init方法遍历opts应用opt，然后初始化regList、runList、address，最后异步e.registry()

RegTask

//注册任务
func (e *executor) RegTask(pattern string, task TaskFunc) {
	var t = &Task{}
	t.fn = task
	e.regList.Set(pattern, t)
	return
}

RegTask方法往regList添加指定pattern的task

runTask

func (e *executor) runTask(writer http.ResponseWriter, request *http.Request) {
	e.mu.Lock()
	defer e.mu.Unlock()
	req, _ := ioutil.ReadAll(request.Body)
	param := &RunReq{}
	err := json.Unmarshal(req, &param)
	if err != nil {
		_, _ = writer.Write(returnCall(param, 500, "params err"))
		e.log.Error("参数解析错误:" + string(req))
		return
	}
	e.log.Info("任务参数:%v", param)
	if !e.regList.Exists(param.ExecutorHandler) {
		_, _ = writer.Write(returnCall(param, 500, "Task not registered"))
		e.log.Error("任务[" + Int64ToStr(param.JobID) + "]没有注册:" + param.ExecutorHandler)
		return
	}

	//阻塞策略处理
	if e.runList.Exists(Int64ToStr(param.JobID)) {
		if param.ExecutorBlockStrategy == coverEarly { //覆盖之前调度
			oldTask := e.runList.Get(Int64ToStr(param.JobID))
			if oldTask != nil {
				oldTask.Cancel()
				e.runList.Del(Int64ToStr(oldTask.Id))
			}
		} else { //单机串行,丢弃后续调度 都进行阻塞
			_, _ = writer.Write(returnCall(param, 500, "There are tasks running"))
			e.log.Error("任务[" + Int64ToStr(param.JobID) + "]已经在运行了:" + param.ExecutorHandler)
			return
		}
	}

	cxt := context.Background()
	task := e.regList.Get(param.ExecutorHandler)
	if param.ExecutorTimeout > 0 {
		task.Ext, task.Cancel = context.WithTimeout(cxt, time.Duration(param.ExecutorTimeout)*time.Second)
	} else {
		task.Ext, task.Cancel = context.WithCancel(cxt)
	}
	task.Id = param.JobID
	task.Name = param.ExecutorHandler
	task.Param = param
	task.log = e.log

	e.runList.Set(Int64ToStr(task.Id), task)
	go task.Run(func(code int64, msg string) {
		e.callback(task, code, msg)
	})
	e.log.Info("任务[" + Int64ToStr(param.JobID) + "]开始执行:" + param.ExecutorHandler)
	_, _ = writer.Write(returnGeneral())
}

runTask方法先判断task是否已经注册了，则根据ExecutorBlockStrategy做不同处理，若是coverEarly则cancel掉已有的task；最后通过task.Run来异步执行任务

killTask

func (e *executor) killTask(writer http.ResponseWriter, request *http.Request) {
	e.mu.Lock()
	defer e.mu.Unlock()
	req, _ := ioutil.ReadAll(request.Body)
	param := &killReq{}
	_ = json.Unmarshal(req, &param)
	if !e.runList.Exists(Int64ToStr(param.JobID)) {
		_, _ = writer.Write(returnKill(param, 500))
		e.log.Error("任务[" + Int64ToStr(param.JobID) + "]没有运行")
		return
	}
	task := e.runList.Get(Int64ToStr(param.JobID))
	task.Cancel()
	e.runList.Del(Int64ToStr(param.JobID))
	_, _ = writer.Write(returnGeneral())
}

killTask方法则执行task.Cancel()，同时将其从runList移除

taskLog

func (e *executor) taskLog(writer http.ResponseWriter, request *http.Request) {
	var res *LogRes
	data, err := ioutil.ReadAll(request.Body)
	req := &LogReq{}
	if err != nil {
		e.log.Error("日志请求失败:" + err.Error())
		reqErrLogHandler(writer, req, err)
		return
	}
	err = json.Unmarshal(data, &req)
	if err != nil {
		e.log.Error("日志请求解析失败:" + err.Error())
		reqErrLogHandler(writer, req, err)
		return
	}
	e.log.Info("日志请求参数:%+v", req)
	if e.logHandler != nil {
		res = e.logHandler(req)
	} else {
		res = defaultLogHandler(req)
	}
	str, _ := json.Marshal(res)
	_, _ = writer.Write(str)
}

taskLog方法通过e.logHandler(req)或者defaultLogHandler(req)来获取日志

小结

xxl-job-executor-go的Executor定义了Init、LogHandler、RegTask、RunTask、KillTask、TaskLog、Run方法；executor实现了Executor接口，并提供了http的api接口。

doc

• xxl-job-executor-go