Ganglia, it is a distributed monitoring system for High Perfomance Computing Clusters and Grid Architecture. But this could be used to monitor two individual systems as well.
In this tutorial I will explain, how to configure it on a High Performance Computing Cluster ( with one master node and couple of compute node). By default this tool is integrated with most of the cluster building toolkits ( like Rocks, InClus) but if you are building manual cluster then you have install it manually on every machine.
Lets begin then,
First task is to download ganglia tar,
Website:- Official Ganglia Website
Installation procedure is really simple you just need to do configure, make and make install.
If you don’t want to install using tar then rpms are also available,
Website: Ganglia RPMS
It might possible that using the procedure of rpms/tar you might face error of dependencies. If you have opted for rpms procedure then it is always a good idea to make a repository of all the ganglia packages and install it using yum( which will make use of default distribution packages to resolve dependencies).
Difference between normal folder with all rpms and repository is that, repository directory will have a repo directory inside it. This repo directory will hold all the information about repository like available rpms, dependencies structure etc. This information is vital for package installer like yum, they always search for this folder inside any repository. To make repo directory you need to run createrepo command inside that rpms directory.
Next thing is to make this repository available to yum, yum checks .repo extensions files /etc/yum.repos.d/. You need to create your repo file(ganglia.repo) inside this directory, content of file will be:
[ganglia]
name= ganglia
baseurl = http://localhost/ganglia
enabled = 1
gpgcheck = 0
We are sharing our repository using http, so your ganglia folder should be available at location /var/www/html/ganglia.
Once this is done, yum could use our repository to install ganglia. As you have both the ganglia and OS distribution repository with you.
Run the following command:
yum clean all
then
yum -y install ganglia-gmetad ganglia-gmond
And on the compute node run the command(after configuring yum on that)
yum -y install ganglia-gmond
Download ganglia-web/gweb tar from the Ganglia website and extract it inside /var/www/html/.
This is required for web interface of ganglia, extract it and run “make install”
Installation finished, Now the important task is configuration,
There are two services which needed to be configured for ganglia, gmond and gmetad.
Gmetad: Its configuration file is at location /etc/ganglia/gmetad.conf. And it could be start, stop as service from /etc/init.d/gmetad.
This is the server of ganglia, it will only run on master node. Inside its configuration file you need to make only one change, that is to provide data source name as per your wish(In most cases, people assign this name as my cluster or with name of organization).
Search for data_source in /etc/ganglia.conf and provide some name.
Gmond: This is the resouce collector for ganglia, it will be installed on master node as well as on compute node. Its configuration file is at location, /etc/ganglia/gmond.conf and its service is at /etc/init.d/gmond.
In this file you need to edit the following section:
cluster {
name = "my cluster"
owner = "unspecified"
latlong = "unspecified"
url = "unspecified"
}
udp_send_channel {
host = mycluster.com
port = 8649
}
udp_recv_channel {
port = 8649
}
tcp_accept_channel {
port = 8649
}
This mycluster.com should be the name of server which is running gmetad deamon, in our case master node. Instead of providing name you can also provide IP Address, choice is yours, if you are providing name then it should get resolved either by /etc/hosts file or through DNS.
Restart the Collector Daemon (gmond): service gmond restart
Restart the Server Daemon (gmetad): service gmetad restart
Gmetad will run only on master and gmond will also run on master node as well as on all compute node.
Now, to verify your installation is successful or not open web browser on master node and type
if you have any query regarding installation procedure, please comment….
Follow us on Facebook.
Thanks
Abhishek.
Networls 