#gmetad

Ganglia is a very nice tool to help monitor several statistics in a cluster environment.  For eg., you can track the CPU times for all the clusters in a single end point.

Setting up ganglia is fairly simple.   You just have to start two daemon processes.  gmond (Ganglia monitor) and gmetad (Ganglia meta).  And you need to copy the web scripts, the php ones into proper apache directory ( assuming your web server is apache).

gmond -  This has to run on every node in the cluster.

gmetad -  This has to run where the end point is specified.

gmond collects statistics and stores in a location.  It frequently sends data to gmetad.  Usually gmond is set to support multicast (by default) the meta data.  This can be set to unicast to gmetad. Once this setup is done, you should be able to see the webpage showing graphs.

For detailed setup follow this link.  

The data available in METRICS elements from Ganglia gmetad XML output with parameter /?filter=summary is, as suggested by the attribute name SUM instead of VAL, a sum of metric values for each host in a cluster or grid. 

The number of values in the SUM is given separately in the NUM attribute of each METRICS element. This may be different from the total number of running hosts (see HOSTS element) if not all hosts supply the given metric.