hadoop的ganglia数据监控

如果我们想知道当前运行的hadoop集群的状态，可以通过hadoop的客户端和web页面来获得，但是如果我们想知道当前集群的繁忙程度，如读写次数，这些工具就办不到了。幸运的是hadoop提供了一种ganglia的数据采集方式。在这篇文章里，将介绍一下hadoop与ganglia的配置方式。

Hadoop 版本:1.2.1

OS 版本: Centos6.4

Jdk 版本: jdk1.6.0_32

Ganglia 版本:3.1.7

环境配置

机器名	Ip地址	功能
Hadoop1	192.168.124.135	namenode, datanode, secondNameNode jobtracker, tasktracer
Hadoop2	192.168.124.136	Datanode, tasktracker
Hadoop3	192.168.124.137	Datanode, tasktracker
ganglia	192.168.124.140	Gmetad,gmond   ganglia-web

基本架构

hadoop1, hadoop2, hadoop将数据发送给ganglia节点上的gmond, gmetad定期向gmond获取数据，最后通过httpd显示出来。

安装ganglia

Yum仓库中没有ganglia，需要安装一个epel仓库

rpm -Uvh http://dl.Fedoraproject.org/pub/epel/6/x86_64/epel-release-6-8.noarch.rpm

在ganglia依次运行

Yum install ganglia-gmetad

Yum install ganglia-gmond

Yum install ganglia-web

运行完这三条命令后,整个ganglia环境就准备好了,包括httpd,php

配置ganglia

vi /etc/ganglia/gmetad.conf 修改data_source

data_source "my_cluster" ganglia

vi /etc/ganglia/gmond.conf

单播模式

cluster {

name = "my_cluster"

owner = "unspecified"

latlong = "unspecified"

url = "unspecified"

}

udp_send_channel {

#bind_hostname = yes # Highly recommended, soon to be default.

# This option tells gmond to use a source address

# that resolves to the machine's hostname.  Without

# this, the metrics may appear to come from any

# interface and the DNS names associated with

# those IPs will be used to create the RRDs.

#mcast_join = 239.2.11.71

host = 192.168.124.140

port = 8649

ttl = 1

}

/* You can specify as many udp_recv_channels as you like as well. */

udp_recv_channel {

#mcast_join = 239.2.11.71

port = 8649

#bind = 239.2.11.71

}

vi conf/hadoop-metrics2.properties

*.sink.ganglia.class=org.apache.hadoop.metrics2.sink.ganglia.GangliaSink31

*.sink.ganglia.period=10

*.sink.ganglia.supportsparse=true

*.sink.ganglia.slope=jvm.metrics.gcCount=zero,jvm.metrics.memHeapUsedM=both

*.sink.ganglia.dmax=jvm.metrics.threadsBlocked=70,jvm.metrics.memHeapUsedM=40

namenode.sink.ganglia.servers=192.168.124.140:8649

datanode.sink.ganglia.servers=192.168.124.140:8649

jobtracker.sink.ganglia.servers=192.168.124.140:8649

tasktracker.sink.ganglia.servers=192.168.124.140:8649

maptask.sink.ganglia.servers=192.168.124.140:8649

reducetask.sink.ganglia.servers=192.168.124.140:8649

启动

先关闭防火墙: service iptables stop

启动httpd: service httpd start

启动gmetad: service gmetad start

启动gmond: service gmond start

启动 hadoop集群：bin/start-all.sh

结果

从图上可以看出，我们已经成功的显出ganglia, hadoop1, hadoop2, hadoop3的信息

Hadoop2和hadoop3都监控datanode,tasktracker，他们显示的metric是一样的

Hadoop1比hadoop2,hadoop3多运行三个组件：namenode, secondnamenode, jobtracker，所以会多出dfs.FSNameSystem metrics，dfs.namenode metrics，mapred.Queue metrics，mapred.jobtracker metrics

下面我们将列出hadoop1节点上所有metric的图，有兴趣的可以看一看。

结论

1. 此hadoop集群是没有启动security，因为ugi没有数据
2. 可以看出hadoop的一些参数信息
3. 可以看出目前hadoop的一些系统信息，是否繁忙