[openstack-dev] [Ceilometer] Issues with transformers

Adrian Turjak adriant at catalyst.net.nz
Mon Jan 27 21:36:38 UTC 2014

On 27/01/14 23:41, Julien Danjou wrote:
> On Mon, Jan 27 2014, Adrian Turjak wrote:
>> I created a gauge metric that is updated via notifications, and a pollster.
>> The data from both of those needs to be transformed in to a cumulative
>> metric. The transformer object works as intended, but the issue is that
>> while my pollster/notifications combo does create samples for the same named
>> metric, two different transformer objects are being created for each data
>> input for the metric.
> If I understand correctly what you're trying to do, it won't work
> unfortunately. Transformers are run _before_ publishing to a collector,
> and are local to a pipeline inside a Ceilometer deamon. The deamons
> handling the pollsters and the notifications are two different programs,
> and therefore don't communicate.

Is there a way around this or some other approach because unless I can 
get both sets of data the final calculations are not useful. Unless I 
can somehow force a notification on an interval... but that is likely 
impossible, or requires work in Nova.

What I've made is a gauge metric that represents the state of a VM at a 
given time so that I can then do accurate uptime calculations taking 
into account suspended states and shutdowns. The state metric I've made 
is in a blueprint here (and there is a link to my code):

Since I need consistent state data as well as accurate transitional data 
it needs to be both notification and pollster. What my transformer then 
does is build a cumulative value of uptime based on the state 
information, and allows a user to define what states they care about for 
billing via the pipeline parameters.

Is there a way to transform data once it reaches the collector? Or would 
an approach be to build a separate agent to transform this, and likely 
other data, into usable billing data to post to the collector?

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack-dev/attachments/20140128/f198ee1b/attachment.html>

More information about the OpenStack-dev mailing list