Open Stack

Hi Kurt,

Your understanding is correct about the the 'cpu time used' meter. It is a cumulative meter, ever increasing since boot-up and reset to zero on restart. Vsphere does not support this meter. Details about meters supported by different hypervisors can be found at: http://docs.openstack.org/developer/ceilometer/measurements.html#compute-nova.
You can use 'cpu_util' meter instead which is of type 'gauge', if it satisfies your requirement.

Personally, I think a 'delta' counter will make more sense. However, you can derive the delta value by comparing two consecutive values for the cumulative meter and figuring out the difference.

Hope that helps.

Thanks,
Piyush

From: 饶定远 [mailto:raodingyuan at chinacloud.com.cn]
Sent: Thursday, July 03, 2014 3:15 PM
To: openstack at lists.openstack.org
Subject: [Openstack] [Ceilometer] Question regarding the meter "cpu time used"

Hi folks,

I found that the “cpu time used” meter actually means “cpu time used *since last boot-up*”, at least in libvirt environments. It will be reset to ZERO after VM restart. As far as I understand, a cumulative type meter should be strictly increasing, so it looks very strange to me.

My questions are:
1, Since I have libvirt environment only, I haven’t tested on vsphere or any others. Does it behave the same in other environments?
2, Is it intended to be like this?
2, What should a billing system do to correctly use this meter?

Please help :)

Thanks a lot!


Kurt Rao
ChinaCloud Information Technology Co., Ltd.
E_mail: raodingyuan at chinacloud.com.cn<mailto:raodingyuan at chinacloud.com.cn>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack/attachments/20140704/e2dc19b6/attachment.html>