Description of problem:

The Compute (Nova) meters list is generally incomplete in the Metric dropdown on the Stats tab of the admin/Resource Usage panel.

The content of this dropdown list also changes from time to time.

This is because the list is populated from the meter links in the *first* resource reported by the ceilometer API with availability zone metadata (i.e. indicating a nova instance).

However, the set of meters associated with instances is not uniform in ceilometer (i.e. a certain meter gathered for one resource may not necessarily be gathered for another).

Also the content of this list changes from time to time, depending on the order of sample acquisition (as a different instance, with a different set of associated meters, may be at the head of the resources list reported by the ceilometer API).

For example, once instances of different flavors have been spun up in an openstack deployment, the Metric list will not contain all the possible compute meters, nor will it even always be incorrect in the same way. Only the 'instance:<flavor>' meter relating to the first instance appears in the dropdown, and the identity of this first instance may change over time.

So at any given point in time, only a subset of the compute meters are accessible via the dashboard.


Version-Release number of selected component (if applicable):

RHOS 4.0 beta


How reproducible:

100%


Steps to Reproduce:
1. Spin two instances with different flavor (say m1.tiny and m1.micro)

2. Open the admin/Resource Usage panel

3. Note that only one 'instance:<flavor>' meter is reported in the Metric downdrop box


Actual results:

Only one 'instance:<flavor>' meter is reported in the Metric downdrop box.


Expected results:

A 'instance:<flavor>' meter for each existing instance type should be reported in the Metric downdrop box.