通过Spark Rest 服务监控Spark任务执行情况

1、Rest服务

　　Spark源为了方便用户对任务做监控，从1.4版本启用Rest服务，用户可以通过访问地址，得到application的运行状态。

　　Spark的REST API返回的信息是JSON格式的，开发者们可以很方便地通过这个API来创建可视化的Spark监控工具。目前

　　这个API支持正在运行的应用程序，也支持历史服务器。在请求URL都有/api/v1。比如，对于历史服务器来说，我们可以通过

　　http://***:18080/api/v1 来获取一些信息，端口可以改；对于正在运行的Spark应用程序，我们可以通过 https://***/api/v1 

　　来获取一些信息。

 

　　主要用途： 通过rest服务，可以轻松对任务时长、stage等做监控，同时可以配合时间序列数据库，对集群各个任务做监控。

2、实例代码（Python）

　　通过脚本指令，先得到自己程序有关的application id，根据application id 去请求rest服务，获取结果集，进而对结果进行其他分析。

#!/usr/bin/env python
# -*- coding: utf-8 -*-
'''
    Created by zhangy on Aug 25, 2017
'''
import datetime
import json, urllib2
import os
import time  


if __name__ == '__main__':
    command = "yarn application -list |grep Noce |awk -F'\t' '{print $1}'"
    val = os.popen(command).read()
    appids = val.split("\n")
    for pid in appids:
        if pid.__eq__(""):continue
        url = "http://th04-znwg-sgi620-001:18088/api/v1/applications/" + pid
        req = urllib2.Request(url)
        res_data = urllib2.urlopen(req)
        res = res_data.read()
        jo = json.loads(res)
        dict1 = jo['attempts'][0]
        st = dict1['startTime']
        GMT_FORMAT = '%Y-%m-%dT%H:%M:%S.%fGMT'
        sti = datetime.datetime.strptime(st, GMT_FORMAT)
        startTime = time.mktime(sti.timetuple()) + 8 * 60 * 60
        nowTime = long(time.time())
        sub = nowTime - startTime
        if sub > 4 * 60 * 60:
            killCommand = "yarn application -kill " + pid
            res = os.popen(killCommand ).read() 
　　　　　　　cc = time.strftime("%Y-%m-%d %H:%M:%S", time.localtime(float(nowTime)))
　　　　　　　f = open("/home/noce1/run_noce/his/monitor/" + pid + ".txt", "a") 
　　　　　　　f.write(cc + " : " + "pid ： " + "\n" + sub + " seconds") 
　　　　　　　f.write(res + "\n") 
 　　　　    f.close()

   　　测试实例中，只是对spark任务的时间做了监控，如果任务超过理想执行时长（4个小时），则终止任务，释放资源。

　结果：

例如：
http://132.12*****:18088/api/v1/applications/application_1502706935975_233268/

返回内容：json格式
{
  "id" : "application_1502706935975_233268",
  "name" : "FRT3_73",
  "attempts" : [ {
    "startTime" : "2017-09-04T01:29:53.986GMT",
    "endTime" : "2017-09-04T01:31:52.955GMT",
    "sparkUser" : "noce1",
    "completed" : true
  } ]
}

 

3、官方其他（2.1.0版本，http:**** :18080/api/v1/）

 　　　　官方给出了rest服务的其他地址接口，通过这些请求这些地址，可以获取不同的json结果集合。

Endpoint	Meaning
/applications	A list of all applications. ?status=[completed|running] list only applications in the chosen state. ?minDate=[date] earliest start date/time to list. ?maxDate=[date] latest start date/time to list. ?minEndDate=[date] earliest end date/time to list. ?maxEndDate=[date] latest end date/time to list. ?limit=[limit] limits the number of applications listed. Examples: ?minDate=2015-02-10 ?minDate=2015-02-03T16:42:40.000GMT ?maxDate=2015-02-11T20:41:30.000GMT ?minEndDate=2015-02-12 ?minEndDate=2015-02-12T09:15:10.000GMT ?maxEndDate=2015-02-14T16:30:45.000GMT ?limit=10
/applications/[app-id]/jobs	A list of all jobs for a given application. ?status=[running|succeeded|failed|unknown] list only jobs in the specific state.
/applications/[app-id]/jobs/[job-id]	Details for the given job.
/applications/[app-id]/stages	A list of all stages for a given application.
/applications/[app-id]/stages/[stage-id]	A list of all attempts for the given stage.
/applications/[app-id]/stages/[stage-id]/[stage-attempt-id]	Details for the given stage attempt.
/applications/[app-id]/stages/[stage-id]/[stage-attempt-id]/taskSummary	Summary metrics of all tasks in the given stage attempt. ?quantiles summarize the metrics with the given quantiles. Example: ?quantiles=0.01,0.5,0.99
/applications/[app-id]/stages/[stage-id]/[stage-attempt-id]/taskList	A list of all tasks for the given stage attempt. ?offset=[offset]&length=[len] list tasks in the given range. ?sortBy=[runtime|-runtime] sort the tasks. Example: ?offset=10&length=50&sortBy=runtime
/applications/[app-id]/executors	A list of all active executors for the given application.
/applications/[app-id]/allexecutors	A list of all(active and dead) executors for the given application.
/applications/[app-id]/storage/rdd	A list of stored RDDs for the given application.
/applications/[app-id]/storage/rdd/[rdd-id]	Details for the storage status of a given RDD.
/applications/[base-app-id]/logs	Download the event logs for all attempts of the given application as files within a zip file.
/applications/[base-app-id]/[attempt-id]/logs	Download the event logs for a specific application attempt as a zip file.
/applications/[app-id]/streaming/statistics	Statistics for the streaming context.
/applications/[app-id]/streaming/receivers	A list of all streaming receivers.
/applications/[app-id]/streaming/receivers/[stream-id]	Details of the given receiver.
/applications/[app-id]/streaming/batches	A list of all retained batches.
/applications/[app-id]/streaming/batches/[batch-id]	Details of the given batch.
/applications/[app-id]/streaming/batches/[batch-id]/operations	A list of all output operations of the given batch.
/applications/[app-id]/streaming/batches/[batch-id]/operations/[outputOp-id]	Details of the given operation and given batch.
/applications/[app-id]/environment	Environment details of the given application.