distinguish samples from notification and polling
Currently, samples from notification and samples from pollster are not be distinguished, this may lead to wrong alarm evaluation.
Please consider this scenario:
We have created an threshold alarm with the condition is :instance >= 5.0 during 1 x 60s, and there are 3 instances in our environment, when creating an instance the alarm may be changed to alarm state because of samples yielding from notifications of vm creation. That is unreasonable.
This blueprint aim to distinguish the samples from notifications and pollsters, the basic idea is:
* add a column to sample db table to identify the sample from notification or pollster
* add the above field as a query filter in statistic API
* ...
Whiteboard
The related discussion in mailing-list:
https://<email address hidden>
this might be connected to this: https:/
Hi gordon, after a scan of the spec, it is a little different between the goal of the spec and this, this bp want to add a flag to sample object to distinguish the samples from polling-agent or notification-agent, this want to address issue described by above mail thread (posted by ZhiQiang Fan). --liusheng (2015.6.30)
what happens when now all our samples will run through notification agent -- gordc (14.07.15)
can this be handled by this: https:/
we have deprecated the samples with volume=1 from notification, so this issue is unlikely to occur again, abandon this -- liusheng (28.11.15)
