The quota usages appear to be inconsistent when an instance is created as a user then deleted as an admin (via the command line). We had a 6-core instance running and booting from a volume. After detaching the volume to do some work on the partition, we were unable to reattach it so we need to delete the instance and create a new one. I tried suspending the instance to reattach the volume, but when that didn't work we were now unable to terminate the instance via horizon. I then deleted the instance using the nova CLI (as the admin user) and attempted the recreate the instance using the horizon dashboard. I received an error stating our quota limit had been reached when the summary page stated that it clearly was not. Checking the nova database files you can see the discrepancy (screenshots attached). The database screenshot was taken from the nova.quota_usages table.
Thanks for filing this ticket. I did some tests and couldn't reproduce above scenario with Folsom Release. Basically, I don't see issue with non-admin users, but I do see issue for admin. Seems like for admin, other tenant's vms aslo considered/counted and I hit the quota when creating vm with admin. The nova absolute-limits also shows higher amount (when running as admin tenant) and not what admin has.