Comment 2 for bug 1408763

John, thanks for looking into this bug. I'd like to explain a bit about the origin of 'allocated_capacity_gb'. This parameter was introduced (in commit 254e37ab3c995f6514084d38f2f797da9cf5e5a9) to scheduler so that filter scheduler could simulate the behavior of Simple scheduler (because we wanted to remove Simple/Chance scheduler). So allocated_capacity_gb was intentionally implemented to represent apparent Cinder level value (simply equals sum of all volumes hosted on the backend, even including volumes in 'error' state, because that was how Simple scheduler behave) and because of that the tracking and calculation of this parameter was put in volume manager instead of driver.

Before we dive into the discussion of fixing the bug, I'd like to understand are we going to change the semantic of 'allocated_capacity_gb' to something other than what it is now?