Besides the basic Ganglia configuration for monitoring OpenNebula hosts and resource usage of VMs, we can configure the OpenNebula frontend to use information about hosts and VMs collected by Ganglia, rather than the default Information Manager drivers.
The Ganglia drivers for OpenNebula should reduce the monitoring overhead in a large deployment as they do not rely on SSH connections to the nodes to get all the information. On the other side, cron jobs must be installed on the nodes to provide VM information to the Ganglia system in a not-so-fashionable way, as we will see later.
To enable the Ganglia Information Manager driver defined in oned.conf
:
IM_MAD = [ name = "im_ganglia", executable = "one_im_sh", arguments = "ganglia" ]
To enable the Ganglia information polling, append -l poll=poll_ganglia
to the existing parameters listed in the arguments
attribute of the VM manager driver in oned.conf
:
VM_MAD = [ name = "vmm_kvm", ...