zabbix监控kafka集群消息队列情况

通过zabbix监控kafka运行时的主题、分区、偏移量以及消息的堆积消费等。

 

一、配置zabbix

配置定义zabbix的监控项，其中192.168.2.3:8088是kafkaOffsetMonitor服务的IP和端口

[root@prod-datakafka zabbix_agentd.d]# cat userparameter_kafkaconsumer.conf

UserParameter=kafka.discover,/usr/bin/python /etc/zabbix/externalscripts/discover_kafka.py 192.168.2.3:8088 logstash

UserParameter=kafka.info[*],/usr/bin/python /etc/zabbix/externalscripts/kafka.py 192.168.2.3:8088 logstash $1 $2

 

# discover_kafka.py

#!/usr/bin/env python
#-*- coding:utf-8 -*-

import json
import sys
import os

def kafkaConsumerDiscovery():
    # kafka消费者自动发现
    import threading
    zabbixDiscovery = {"data": []}
    tp = []
    url = sys.argv[1]
   # try:
    group = sys.argv[2]
  #  except:
     #   ifInfo = False
    # 处理函数
   # def action(url, group):
    try:
        info = json.loads(os.popen("/usr/bin/curl -s '%s/group/%s'" % (url, group)).read().strip())['offsets']
       # print(info)
        for partition in info:
            # 没有订阅则跳过
            if len(partition) <= 7:
                continue
            if True:
                zabbixDiscovery['data'].append(
                    #{"{#URL}": url, "{#GROUP}": group, "{#TOPIC}": str(partition['topic']), "{#OWNER}": str(partition['owner'])})
                    {"{#TOPIC}": str(partition['topic']), "{#OWNER}": str(partition['owner'])})
            else:
                if partition['partition'] == 0:
                    zabbixDiscovery['data'].append(
                        {"{#URL}": url, "{#GROUP}": group, "{#TOPIC}": str(partition['topic'])})
    except:
        pass
    #action(url,ifInfo)
     #print(zabbix)
    # 创建线程池
#    for group1 in json.loads(os.popen("/usr/bin/curl -s '%s/group'" % url).read().strip()):
#        group = str(group1)
 #       t = threading.Thread(target=action, args=(url, group))
#        t.setDaemon(True)
#        tp.append(t)
    # 控制并发数量
#    while tp:
#        tp.pop().start()
#        if threading.activeCount() > 10:
#            while threading.activeCount() > 10:
#                time.sleep(1)
#        elif threading.activeCount() == 0:
#            break
#        else:
#            pass
    
    print(json.dumps(zabbixDiscovery))
    return json.dumps(zabbixDiscovery)



kafkaConsumerDiscovery()

 

# kafka.py

#!/usr/bin/env python
#-*- coding:utf-8 -*-
import sys
import os
import json
import time

def kafkaLag():
    # 收集kafka消息信息
    url = sys.argv[1]
    group = sys.argv[2]
    topic = sys.argv[3]
    type = sys.argv[4]
    count = 0
    info = json.loads(os.popen("/usr/bin/curl -s '%s/group/%s'" % (url, group)).read().strip())['offsets']
    for partition in info:
        if partition['topic'] == topic:
            if type == 'lag':
                count = partition['logSize'] - partition['offset'] + count
            else:
                count += partition[type]
    print count


def kafkaLastSeen():
    # 收集kafka消息最后被消费至今的时间差
    url = sys.argv[1]
    group = sys.argv[2]
    topic = sys.argv[3]
   # owner = sys.argv[4]
    difftime = 0
    info = json.loads(os.popen("/usr/bin/curl -s '%s/group/%s'" % (url, group)).read().strip())['offsets']
    for partition in info:
        #if partition['topic'] == topic and partition['owner'] == owner:
        if partition['topic'] == topic:
            if partition['logSize'] - partition['offset'] > 0:
                difftime = time.time() - float(partition['modified']) / 1000
                break
    print float(difftime)


def kafkaConsumerDiscovery():
    # kafka消费者自动发现
    import threading
    zabbixDiscovery = {"data": []}
    tp = []
    url = sys.argv[1]
    try:
        ifInfo = sys.argv[2]
    except:
        ifInfo = False
    # 处理函数
    def action(url, group):
        try:
            info = json.loads(os.popen("/usr/bin/curl -s '%s/group/%s'" % (url, group)).read().strip())['offsets']
           # print(info)
            for partition in info:
                # 没有订阅则跳过
                if len(partition) <= 7:
                    continue
                if ifInfo:
                    zabbixDiscovery['data'].append(
                        {"{#URL}": url, "{#GROUP}": group, "{#TOPIC}": str(partition['topic']), "{#OWNER}": str(partition['owner'])})
                else:
                    if partition['partition'] == 0:
                        zabbixDiscovery['data'].append(
                            {"{#URL}": url, "{#GROUP}": group, "{#TOPIC}": str(partition['topic'])})
        except:
            pass
    action(url,ifInfo) 

    # 创建线程池
#    for group1 in json.loads(os.popen("/usr/bin/curl -s '%s/group'" % url).read().strip()):
#        group = str(group1)
 #       t = threading.Thread(target=action, args=(url, group))
#        t.setDaemon(True)
#        tp.append(t)
    # 控制并发数量
#    while tp:
#        tp.pop().start()
#        if threading.activeCount() > 10:
#            while threading.activeCount() > 10:
#                time.sleep(1)
#        elif threading.activeCount() == 0:
#            break
#        else:
#            pass
    print json.dumps(zabbixDiscovery)


def kafkaStatus():
    # kafka集群主机的健康状态
    ip_list = []
    url = sys.argv[1]
    group = sys.argv[2]
    ip = sys.argv[3]
    info = json.loads(os.popen("/usr/bin/curl -s '%s/group/%s'" % (url, group)).read().strip())['brokers']
    for i in info:
       ip_list.append(i['host'])
    if ip in  ip_list:
        print 1
    else:
        print 0
    #count = len(info)
   # print count


def kafkaCount():
    url = sys.argv[1]
    group = sys.argv[2]
    info = json.loads(os.popen("/usr/bin/curl -s '%s/group/%s'" % (url,group)).read().strip())['brokers']
   #print(info)
    count = len(info)
    print (count)


if __name__ == '__main__':
    if sys.argv[3] == "count":
        kafkaCount()
    elif sys.argv[4] == "number":
        kafkaLastSeen()
    elif sys.argv[4] == "status":
        kafkaStatus()
    else:
        kafkaLag()
   # kafkaLastSeen()

 

二、安装kafkaOffsetMonitor监控服务

下载地址：链接:https://pan.baidu.com/s/1C-cpX8K8ZN-WLp48kwMBgw 提取码:8l0e

把下载的jar包放到指定目录下，使用如下脚本启动即可

kafka_monitor_start.sh

#!/bin/bash

java -cp KafkaOffsetMonitor-assembly-0.2.0.jar com.quantifind.kafka.offsetapp.OffsetGetterWeb --zk 172.16.16.12:2181,172.16.16.13:2181,172.16.16.14:2181 --port 8088 --refresh 5.seconds --retain 1.days &

上面三个zk地址是kafka的三台集群节点，即指定你需要监控的kafka节点的ip

 

kafka到上面就都配置完成了，接下来只要在zabbix上配置好自动发现就行，这部分比较基本，自行配置即可。