Ganglia监控Hadoop与HBase集群
以下基于上篇Hadoop2.6与HBase1.0集群架构:
http://lizhenliang.blog.51cto.com/7876557/1665130
http://lizhenliang.blog.51cto.com/7876557/1661354
Hadoop集群基本部署完成,接下来就需要有一个监控系统,能及时发现性能瓶颈,给故障排除提供有力依据。监控hadoop集群系统好用的比较少,自身感觉ambari比较好用,但不能监控已有的集群环境,挺悲催的。ganglia在网上看到原生支持Hadoop、Hbase性能监控,试用一段时间感觉还不错,监控项比较全面,配置简单,软件包在epel网络源中,使用yum安装方便快捷。
Ganglia是一个开源集群监控系统,主要用来监控系统性能,如:cpu、mem、硬盘利用率、I/O负载、网络流量等。
Ganglia涉及到三个组件:
gmetad:定期从gmond获取数据,并将数据存储到RRD存储引擎中。
gmond:被监控端代理程序,用于收集监控信息,并发送给gmetad。
ganglia-web:WEB前端,RRD绘图后通过PHP展示。
四、Ganglia集群监控系统部署
1. CentOS7 YUM源自带epel网络源,直接安装(部署到HMaster0上,尽量单独放一台机器上)
1
2
|
# yum install epel-release # yum install ganglia-web ganglia-gmetad ganglia-gmond |
此时会安装出来我们要安装的ganglia之外,还要安装httpd、php、rrdtool、gd、png、jpeg等相关软件包
2. 配置监控端(在HMaster0上配置)
2.1 配置监控程序
1
2
3
|
# vi /etc/ganglia/gmetad.conf #修改以下两项
data_source "Hadoop" 192.168.18.215 192.168.18.216 192.168.18.217 192.168.18.218192.168.18.219
case_sensitive_hostnames 1
|
参数说明:
第一个是数据源,Hadoop是集群名字,后跟IP是集群中被监控节点地址。
第二个意思是是否区分主机名大小写,默认是0,修改为1,否则节点中主机名有大写的将保存为小写,写到/var/lib/ganglia/rrds/Hadoop中,导致图形无数据。
2.2 关联Apache,因为Ganglia自创建的配置ganglia.conf有问题,所以先删除,再创建个软连接到Apache根目录下。
1
2
|
# rm /etc/httpd/conf.d/ganglia.conf
# ln -s /usr/share/ganglia /var/www/html/ganglia
|
2.3启动Apache和Ganglia,并设置开机启动
1
2
3
4
|
# systemctl start httpd
# systemctl start gmetad
# systemctl enable httpd
# systemctl enable gmetad
|
3. 安装与配置被监控端(每台同样配置)
3.1 安装与配置代理程序
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
|
# yum install ganglia-gmond
# vi /etc/ganglia/gmond.conf
……
cluster{
name = "Hadoop" #集群名,和上面那个一样
owner = "unspecified"
latlong = "unspecified"
url = "unspecified"
}
/* Thehost section describes attributes of the host, like the location */
host {
location = "unspecified"
}
/*Feel free to specify as many udp_send_channels as you like. Gmond
used to only support having a single channel*/
udp_send_channel{
#bind_hostname = yes # Highly recommended,soon to be default.
# This option tells gmond to use asource address
# that resolves to themachine's hostname. Without
# this, the metrics mayappear to come from any
# interface and the DNSnames associated with
# those IPs will be usedto create the RRDs.
#mcast_join = 239.2.11.71 #关闭多播
host = 192.168.18.215 #添加发送IP/主机名
port = 8649 #默认端口
ttl = 1
}
/* Youcan specify as many udp_recv_channels as you like as well. */
udp_recv_channel{
#mcast_join = 239.2.11.71
port = 8649
bind = 192.168.18.215 #接收地址
retry_bind = true
# Size of the UDP buffer. If you are handlinglots of metrics you really
# should bump it up to e.g. 10MB or evenhigher.
# buffer = 10485760
}
……
|
3.2 将修改好的gmond.conf配置scp到其他节点
1
|
# scp /etc/ganglia/gmond.conf [email protected]:/etc/ganglia/gmond.conf #分别拷贝HSlave0、1、2 |
3.3 启动代理程序,并设置开机启动
1
2
|
# systemctl start gmond # systemctl enable gmond |
4. 添加Hadoop被Ganglia监控,去掉文件中以***释并修改(每台同样配置)
1
2
3
4
5
6
7
8
9
10
|
# vi /opt/hadoop-2.6.0/etc/hadoop/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=192.168.18.215:8649 #当有多个ganglia监控系统时,以逗号分隔
datanode.sink.ganglia.servers=192.168.18.215:8649 #都指定ganglia服务器
resourcemanager.sink.ganglia.servers=192.168.18.215:8649 nodemanager.sink.ganglia.servers=192.168.18.215:8649 |
5. 添加HBase被Ganglia监控,添加如下(每台同样配置)
1
2
3
4
5
|
# vi /opt/hbase-1.0.1.1/conf/hadoop-metrics2-hbase.properties *.sink.ganglia.class=org.apache.hadoop.metrics2.sink.ganglia.GangliaSink31 *.sink.ganglia.period=10 hbase.sink.ganglia.period=10 hbase.sink.ganglia.servers=192.168.18.215:8649 |
6. 重启Hadoop和HBase
1
2
3
4
5
6
|
# stop-dfs.sh # stop-yarn.sh # stop-hbase.sh # start-dfs.sh # start-yarn.sh # start-hbase.sh #HRegionServer节点需要手动启动(hbase-daemon.sh start regionserver) |
7. 通过WEB查看性能监控数据
访问地址:http://192.168.18.215/ganglia