Cacti collects values over SNMP and draws graphs from them (RRDtool). For a small virtualized infrastructure, a standard installation on one dedicated machine is enough to watch both the virtualization hosts and the virtual machines on them.
possibly outdated More than two years have passed since publication. The content may no longer hold.
Deployment
- A dedicated host for monitoring only, separate from the monitored infrastructure, so its failure is unrelated to the failure of what it is supposed to watch.
- A standard Cacti installation (web interface, poller, RRDtool).
- SNMP on the monitored machines – each virtualization host and each virtual machine runs an SNMP agent that Cacti polls for values.
What is monitored
The same basic metrics on both hosts and virtual machines:
- CPU and memory load;
- network traffic;
- file systems (usage);
- uptime and service availability.
Guests: paravirtualized and fully virtualized
Both kinds of XEN guest are among the monitored virtual machines:
- paravirtualized (e.g. Debian) – the guests only need an SNMP agent, as on an ordinary machine;
- fully virtualized (e.g. CentOS) – from the monitoring point of view they behave the same; the SNMP agent returns the same metrics.
Cacti thus shows the whole cluster – from the physical hosts’ load down to individual virtual machines – in one place.
Why a dedicated host
Monitoring should give you a picture precisely when something is wrong. If it runs on the same machine as the monitored service, it goes down with it. A dedicated host removes that problem and keeps the history of the metrics even after the monitored machines reboot.