unix - Ganglia without multicast - Server Fault

Was facing the same problem with multicast mode while configuring Ganglia on Amazon EC2 cloud which prevents the use of multicast in its network. The possible solution is to switch to unicast mode which works fortunately.

To be very concise, below given simple steps are there to get rid of multicast mode.

1. Make one of your nodes master running gmond (ganglia data collector) daemon.

Example: 10 Nodes are there which are running gmond daemon. Pick a node any one from 10 and make that Master which will be getting all the data from 10 Nodes even should also be the slave of itself.

# Define the cluster.  cluster {    name = "Yellow"    owner = "Your Company"    latlong = "N34.02 W118.45"    url = "http://yourcompany.com/"  }    # Disable multicast and define the host, the yellow master, where nodes in the cluster send data.    udp_send_channel {    # mcast_join = 239.2.11.71  (No need to join as mcast is not being used)    host = master.among10node.com  (put the IP/Hostname of server from any 10 nodes to ack as                  master)    port = 8649    ttl = 1  }    udp_recv_channel {    # mcast_join = 239.2.11.71   (Disabled mcast as it is not being used)    port = 8649    # bind = 239.2.11.71  (No need to bind as mcast is not being used)  }   

Note: Copy the same configuration on all 10 nodes running gmond daemon. Restart Master first then all others. Hope it will work and the Master Node will be having all the data from other nodes.

Now configure Ganglia data consolidator (gmetad) daemon to use your Master Node as a primary data source.

Example:

data_source "Yellow" master.among10node.com     # default port is 8649, define here if you are using non default  

Now restart the gmetad daemon and let the magic begin.

Cheers Mohd Mozammil Khan

Read full article from unix - Ganglia without multicast - Server Fault