Comment 25 for bug 1311470

Configuration options is a sore point in OpenStack in general and I personally do not think we should promote an uncontrolled sprawl of static config options, especially to solve this class of problems. As a matter of fact, if the desire is to make type driver a dynamic property of Neutron deployments, then this should be driven via an admin API, rather than a config option which is supposed to be empty only until the state of the deployment change: this is very awkward to say the least.

To your point about DB triggers not being enough, there's a plethora of integration mechanisms to let 3rd parties know about state changes; direct method calls is a tightly coupled mechanism that should be used as last resort. As for the specific issue with type drivers, I believe that if an admin wants to decommission type drivers, he/she should not do so without contemplating the state of his/her cloud. If there are segments in use, then he/she must understand the consequences. As for the situation reported in this bug, the consequences can be dealt with with a simple DB cleanup, because the type driver interface is so trivial that nothing esoteric can happen during the segment release operation, even if a vendor type driver wanted to.