ganglia的部署步骤

1 简介ganglia

这块就长话短说，百度很多，ganglia分为三个组件，分别是gmetad，gmond，gweb。
gmetad：简单来说就是为了整合所有信息
gmond：类似与子节点，需要部署在每台主机上
gweb：进行可视化界面

在配置的时候请注意选择单播模式还是组播模式，不然很容易出错，这里部署采用单播模式。简单介绍一下单播和组播。

单播：代表主机之间一对一的通讯模式，也就说gmond的收集信息会直接反馈给gmetad。
组播：代表主机之间一对一组的通讯模式，一个gmetad可以接受多个gmod组成一组的信息

2 环境部署

在安装ganglia时，前提必要条件就是搭建所需的依赖，这块只需要复制粘贴即可。
yum进行安装

1 首先看一下自己主机上的EPEL源

rpm -qa | grep epel

如果没有出现依赖，则需要自己下载一个。
还需要下载一个httpd，用来进行web访问的

yum -y install httpd

安装依赖源

rpm -Uvh http://dl.fedoraproject.org/pub/epel/6/i386/epel-release-6-8.noarch.rpm

2 安装完依赖之后再进行查询，看看有没有ganglia包

必须要有，没有就不能进行安装！！！
如果安装了依赖也找不到的话，尝试下面的语句再进行查询

yum -y install epel-release

继续安装相应依赖包

yum -y install httpd-devel automake autoconf libtool ncurses-devel libxslt groff pcre-devel pkgconfig php

3 第一个节点的部署

3 安装节点
我有三台服务器，分别是hadoop101 hadoop102 hadoo103，我将在hadoop101上进行安装gmetad，gmond，剩下两个节点则安装gmond即可。

①进行hadoop101上gmetad的安装

yum search ganglia
yum install -y ganglia-gmetad.x86_64 ganglia-web.x86_64
这个版本号就是search出来的，要复制对了，你的可能和我不一样
rpm -qa | grep ganglia

9.2日-补充：我今天又重新做了一遍，发现CentOS7在yum安装时会出现错误。

这是因为CentOS7默认安装的python版本是2.7，CentOS6则是2.6版本，2.6版本符合最新的ganglia，yum安装不会出错。
那这块只能采取编译安装了，编译安装我重新写一个文档，做个链接。
编译安装请看这里
编译安装请看这里
编译安装请看这里
ps：最近闹肚子，把公司的编译安装写了，不写这上面的了，直不起腰了。

进行配置，默认安装位置为/etc/ganglia/gmetad.conf

双引号里的是你监控集群的名字，集群名字后面的gmond都要用到，后面则是跟着你的IP地址，我写的是hostsname

②进行hadoop101上gmond的安装

yum search ganglia
yum install -y ganglia-gmond.x86_64

安装完成后进行gmond.conf的配置

注意红框的配置，文本如下：

globals {
  4   daemonize = yes
  5   setuid = yes
  6   user = nobody
  7   debug_level = 0
  8   max_udp_msg_len = 1472
  9   mute = no
 10   deaf = no
 11   allow_extra_data = yes
 12   host_dmax = 86400 /*secs. Expires (removes from web interface) hosts in 1 day */
 13   host_tmax = 20 /*secs */
 14   cleanup_threshold = 300 /*secs */
 15   gexec = no
 16   # By default gmond will use reverse DNS resolution when displaying your hostname
 17   # Uncommeting following value will override that value.
 18   # override_hostname = "mywebserver.domain.com"
 19   # If you are not using multicast this value should be set to something other than 0.
 20   # Otherwise if you restart aggregator gmond you will get empty graphs. 60 seconds is reasonable
 21   send_metadata_interval = 30 /*secs */
 22 
 23 }
 24 
 25 /*
 26  * The cluster attributes specified will be used as part of the <CLUSTER>
 27  * tag that will wrap all hosts collected by this instance.
 28  */
 29 cluster {
 30   name = "hadoop101"
 31   owner = "unspecified"
 32   latlong = "unspecified"
 33   url = "unspecified"
 34 }
 35 
 36 #Hadoop集群监控，你们不需要配置这个
 37 cluster{
 38   name = "Hadoop"
 39   owner = "umspecified"
 40   latlong = "unspecified"
 41   url = "unspecified"
 42 }
 43 
 44 
 45 /* The host section describes attributes of the host, like the location */
 46 host {
 47   location = "unspecified"
 48 }
 49 
 50 /* Feel free to specify as many udp_send_channels as you like.  Gmond
 51    used to only support having a single channel */
 52 udp_send_channel {   ----发送端
 53   #bind_hostname = yes # Highly recommended, soon to be default.
 54                        # This option tells gmond to use a source address
 55                        # that resolves to the machine's hostname.  Without
 56                        # this, the metrics may appear to come from any
 57                        # interface and the DNS names associated with
 58                        # those IPs will be used to create the RRDs.
 59 #  mcast_join = 239.2.11.71 删除组播
 60   host=hadoop101
 61   port = 8649
 62   ttl = 1
 65 /* You can specify as many udp_recv_channels as you like as well. */
 66 udp_recv_channel {---接受端
 67 #  mcast_join = 239.2.11.71  把组播删除了
 68   port = 8649
 69   bind = hadoop101
 70   retry_bind = true
 71   # Size of the UDP buffer. If you are handling lots of metrics you really
 72   # should bump it up to e.g. 10MB or even higher.
 73   # buffer = 10485760
 74 }
 75 
 76 /* You can specify as many tcp_accept_channels as you like to share
 77    an xml description of the state of the cluster */
 78 tcp_accept_channel {---请求
 79   port = 8649
 80   # If you want to gzip XML output
 81 #  gzip_output = no
 82 }
 83 
 84 /* Channel to receive sFlow datagrams */
 85 #udp_recv_channel {
 86 #  port = 6343
 87 #}

安装gweb

yum search ganglia
yum install -y ganglia-gmetad.x86_64 ganglia-web.x86_64
rpm -qa | grep ganglia

至此hadoop101的ganglia就配置完成了

4 第二、三个节点的部署

同样的操作在hadoop102，hadoop103上，只需安装gmond即可

yum search ganglia
yum install -y ganglia-gmond.x86_64

/* This configuration is as close to 2.5.x default behavior as possible
  2    The values closely match ./gmond/metric.h definitions in 2.5.x */
  3 globals {
  4   daemonize = yes
  5   setuid = yes
  6   user = nobody
  7   debug_level = 0
  8   max_udp_msg_len = 1472
  9   mute = no
 10   deaf = yes
 11   allow_extra_data = yes
 12   host_dmax = 86400 /*secs. Expires (removes from web interface) hosts in 1 day */
 13   host_tmax = 20 /*secs */
 14   cleanup_threshold = 300 /*secs */
 15   gexec = no
 16   # By default gmond will use reverse DNS resolution when displaying your hostname
 17   # Uncommeting following value will override that value.
 18   # override_hostname = "mywebserver.domain.com"
 19   # If you are not using multicast this value should be set to something other than 0.
 20   # Otherwise if you restart aggregator gmond you will get empty graphs. 60 seconds is reasonable
 21   send_metadata_interval = 30 /*secs */
 22 
 23 }
 24 
 25 /*
 26  * The cluster attributes specified will be used as part of the <CLUSTER>
 27  * tag that will wrap all hosts collected by this instance.
 28  */
 29 cluster {
 30   name = "hadoop101"
 31   owner = "unspecified"
 32   latlong = "unspecified"
 33   url = "unspecified"
 34 }
 35 
 36 #Hadoop集群监控
 37 cluster{
 38   name = "Hadoop"
 39   owner = "umspecified"
 40   latlong = "unspecified"
 41   url = "unspecified"
 42 }
 43 
 44 /* The host section describes attributes of the host, like the location */
 45 host {
 46   location = "unspecified"
 47 }
 48 
 49 /* Feel free to specify as many udp_send_channels as you like.  Gmond
 50    used to only support having a single channel */
 51 udp_send_channel {
 52   #bind_hostname = yes # Highly recommended, soon to be default.
 53                        # This option tells gmond to use a source address
 54                        # that resolves to the machine's hostname.  Without
 55                        # this, the metrics may appear to come from any
 56                        # interface and the DNS names associated with
 57                        # those IPs will be used to create the RRDs.
 58 #  mcast_join = 239.2.11.71
 59   host = hadoop101
 60   port = 8649
 61   ttl = 1
 62 }
 63 
 64 /* You can specify as many udp_recv_channels as you like as well. */
 65 udp_recv_channel {
 66 #  mcast_join = 239.2.11.71
 67   port = 8649
 68   bind = hadoop101
 69   retry_bind = true
 70   # Size of the UDP buffer. If you are handling lots of metrics you really
 71   # should bump it up to e.g. 10MB or even higher.
 72   # buffer = 10485760
 73 }
 74 
 75 /* You can specify as many tcp_accept_channels as you like to share
 76    an xml description of the state of the cluster */
 77 tcp_accept_channel {
 78   port = 8649
 79   # If you want to gzip XML output
 80   gzip_output = no
 81 }
 82 
 83 /* Channel to receive sFlow datagrams */
 84 #udp_recv_channel {
 85 #  port = 6343
 86 #}
 87 

主要改


至此hadoop102,hadoop103的gmond节点修改完毕

5 web端的启动

在hadoop101上启动：

service gmetad start
service gmond start
service httpd start
service gmetad status    -------------------验证是否启动成功
service gmond status

在hadoop102，103上启动

service gmond start

访问web界面：(你的ip)/ganglia

如果操作正常应给就没问题，但是也会出现访问失败：

解决办法：
没有权限进行读写操作，执行一下代码

chmod 777 /var/lib/ganglia/dwod/compiled
chmod 777 /var/lib/ganglia/dwoo/cache

然后又出错了

说没有这两个文件，这说明在安装的时候出现了一点问题，不过不要紧，两个办法：

第一个办法，找一下看看这两个文件在哪，然后搜索出来后有一堆，你分别打开看看，里面只有一对{}，复制到报错的路径下就可以了。

find / -name events.json
find / -name enent_color.json   

第二个办法，直接touch events.json，touch enent_color.json，在里面写入{}即可

重新登录

如果出现没有数据，则说明节点配置问题，在好好看看配置不要错了。本次配置为单播配置，比较简单。

祝你成功！如果有帮助请点个赞让我开心下！

最后

以上就是迷路灰狼为你收集整理的【亲测】部署ganglia进行监控--部署过程及Error解决ganglia的部署步骤的全部内容，希望文章能够帮你解决【亲测】部署ganglia进行监控--部署过程及Error解决ganglia的部署步骤所遇到的程序开发问题。

如果觉得靠谱客网站的内容还不错，欢迎将靠谱客网站推荐给程序员好友。