ganglia3.6 监控 hadoop2.2.0

既然hadoop都用了最新版本，那ganglia干脆也用最新的，下面记录一下使用ganglia3.6监控hadoop2.2.0遇到问题：

ganglia3.6安装：

1、编译安装ganglia
tar xf ganglia-3.6.0.tar.gz
cd ganglia-3.6.0
./configure --prefix=/usr/local/ganglia --with-gmetad
make && make install

2、拷贝启动文件和命令
    cp ganglia-3.6.0/gmetad/gmetad.init /etc/init.d/gmetad
    cp ganglia-3.6.0/gmond/gmond.init /etc/init.d/gmond
    cp /usr/local/ganglia/sbin/* /usr/sbin/

3、编辑/usr/local/ganglia/etc/gmetad.conf，修改RRDS的默认路径
大概129行添加：rrd_rootdir "/usr/local/ganglia/rrds"

4、生成gmond配置文件
gmond --default_config > /usr/local/ganglia/etc/gmond.conf

5、创建RRDS目录并修改权限
mkdir -p /usr/locaol/ganglia/rrds/
chown nobody:nobody /usr/locaol/ganglia/rrds/

6、解压ganglia-web-3.5.10.tar.gz
tar xf ganglia-web-3.5.10.tar.gz
mv ganglia-web /var/www/html/ganglia-web

7、COPY默认配置文件
cp -rp /var/www/html/ganglia-web/conf_default.php /var/www/html/ganglia-web/conf.php

8、修改/var/www/html/ganglia-web/conf.php几个默认项
修改gweb_confdir            $conf['gweb_confdir'] = "/var/www/html/ganglia-web";
修改gmetad_root             $conf['gmetad_root'] = "/usr/local/ganglia";               
创建dwoo_compiled_dir    mkdir -p /var/www/html/ganglia-web/dwoo/compiled
创建dwoo_cache_dir        mkdir -p /var/www/html/ganglia-web/dwoo/cache
修改compiled和cache的权限
chmod 777 /var/www/html/ganglia-web/dwoo/compiled
chmod 777 /var/www/html/ganglia-web/dwoo/cache

9、启动gmond和gmetad
/etc/init.d/gmetad start
/etc/init.d/gmond start 

遇到的问题：

libpcre not found, specify --with-libpcre=no to build without PCRE support。 解决：

1. wget ftp://ftp.csx.cam.ac.uk/pub/software/programming/pcre/pcre-8.12.tar.bz2 
2. tar -jxvf pcre-8.12.tar.bz2 
3. cd pcre-8.12 
4. ./configure && make && make install 

 zlib library not configured properly

解决：

sudo yum install -y zlib-devel 

 

/usr/local/lib/libpython2.7.a: could not read symbols: Bad value

./configure --with-gmetad --sysconfdir=/etc/ganglia --enable-gexec --with-python=/usr/local/bin/python2.7
make && make install

*****************

这里我在make的时候，发生错误“/usr/local/lib/libpython2.7.a: could not read symbols: Bad value”

解决办法：需要重装python 带上 --enable-shared 参数，如下：

./configure --enable-shared && make && make install

或者 使用机器自带的python

sudo ./configure --prefix=/usr/local/ganglia --with-python=/usr/bin/python2.4 

ganglia有两种方式收集数据，单播和组播，为了减少网络流量，我们采用单播方式。

gmond.conf 配置文件需要修改以下个点点：

globals {
  daemonize = yes
  setuid = yes
  user = root
  debug_level = 0
  max_udp_msg_len = 1472
  mute = no
  deaf = yes          /*聋子，只发送数据，不接受数据*/
  allow_extra_data = yes
  host_dmax = 86400 /*secs. Expires (removes from web interface) hosts in 1 day */
  host_tmax = 20 /*secs */
  cleanup_threshold = 300 /*secs */
  gexec = no
  # By default gmond will use reverse DNS resolution when displaying your hostname
  # Uncommeting following value will override that value.
  # override_hostname = "mywebserver.domain.com"
  # If you are not using multicast this value should be set to something other than 0.
  # Otherwise if you restart aggregator gmond you will get empty graphs. 60 seconds is reasonable
  send_metadata_interval = 0 /*secs */

}

udp_send_channel {
  #mcast_join = 239.2.11.71
  host = l-datalogm1.data.cn1
  port = 8649
  ttl = 1
}

udp_recv_channel {
 # mcast_join = 239.2.11.71
  port = 8649
  #bind = 239.2.11.71
  #retry_bind = true
  # Size of the UDP buffer. If you are handling lots of metrics you really
  # should bump it up to e.g. 10MB or even higher.
  # buffer = 10485760
}
 

hadoop2.0 的metrics 文件已经变话成了hadoop-metrics2.properties

所以，刚开始一直捣鼓hadoop-metrics.properties，怎么都出不来数据。

hadoop-metrics2.properties：

*.sink.ganglia.class=org.apache.hadoop.metrics2.sink.ganglia.GangliaSink31
*.period=10
datanode.sink.ganglia.servers=l-datalogm1.data.cn1:8649

ganglia3.6 版本的界面很好看，have fun：