openstack usage is reporting inaccurate metrics
Issue
- We use OpenStack 16 in Training classes. Students can launch their lab environments to take classes on specific software and we use a Heat stack to define this environment. When these environments are not in use we do one of two things:
- Delete all resources except the VM volumes (so we have a persistent state). When students want to start their lab environments, we do a stack update with the previously-saved volumes and it will create the VMs, networks, and other resources.
- Shelve the individual VMs.
-
In case 1, we get an accurate measurement of the CPU time used using the openstack usage show --project PROJECT command. We use this information for charge backs.
-
In case 2, we get inaccurate measurement of CPU time used. When we shelve (or stop) the individual VMs, the CPU utilization metric continues to increase for a number of days, then it mysteriously stops.
-
We would like to get an accurate measurement of CPU time used when the VMs are shelved or stopped so we can accurate do a charge back.
Environment
- Red Hat OpenStack Platform 16.1 (RHOSP)
Subscriber exclusive content
A Red Hat subscription provides unlimited access to our knowledgebase, tools, and much more.