The instance meter doesn't work as described here:
http://docs.openstack.org/developer/ceilometer/measurements.html#compute-nova
Currently the volume value is always "1.0" and doesn't change. If this was meant to represent state (1 = active, 0 = suspended, etc) then this might work, but that isn't the case either.
The documentation says it tracks "Duration of instance" which sounds like a cumulative meter to me rather than a gauge, but the meter is explicitly a gauge. Nor are the logs consistent enough to use by adding up all the volume values. Mainly because nova (or the compute agent) forces new ceilometer logs on suspensions and reactivation. These forced logs would be wonderful if this was a meter representing state changes, and would mean things like uptime could be easily computed locally based on the forced logs and changes in state.
As it stands, this meter doesn't seem to do what it is meant to. Or there are differing opinions on what it should do.
A gauge meter that represents states would be very useful, especially since there isn't anything like it in ceilometer, and it seems this meter isn't too far from being that as it is.
An uptime (or duration) meter would also be useful, but if we had a state change meter and it was well documented then uptime could be computed from it.
Currently the instance meter is neither, but seems like it tries to be both. :(
The documentation is wrong, this meter meters the instance itself, not its duration.
Things like uptime, duration, states etc should be provided by the API via the metadata of Ceilometer.