Open Stack

Hi Julien

grouping statistics by resource_id would not help, I have the same problem even if I ask statistics for a single resource_id, what I need to 
do is bill customers on the number of instances, therefore I need to know how many instances existed during a certain period of time,  
I can count the number of instances per minute counting the samples (via ceilometer statistics) but I will always get a wrong count since 
ceilometer collects one sample per minute via the pollster plus the notification samples. Any suggestions on an alternative way of doing it?

Thanks
Alex

-----Original Message-----
From: Julien Danjou [mailto:julien at danjou.info] 
Sent: mercoledì 7 agosto 2013 08:59
To: Alessandro Barabesi
Cc: openstack at lists.openstack.org
Subject: Re: [Openstack] [Ceilometer] filter event samples

On Tue, Aug 06 2013, Alessandro Barabesi wrote:

> The problem is that statistics count all samples therefore all event 
> samples contribute to the values returned, which is not good for what 
> I need to do (there is at least one 'compute.instance.exists' sample 
> per instance per hour plus other possible events). So it would be 
> great if I could avoid event samples.

No there's no way, and that wouldn't be the best possible solution out there since you could still have other events you wouldn't know.
IIUC, grouping statistics by resource_id would solve your problem, and we're working on that for Havana.

--
Julien Danjou
-- Free Software hacker - freelance consultant
-- http://julien.danjou.info