http://wiki.openstack.org/QuantumV2APIIntro describes that a subnet can have a set of IP ranges called "reserved_ranges" which indicate ranges of IPs that should not be automatically allocated by Quantum (note: if an API user explicitly requests that IP, the assumption is that they know what they are doing and we will grant the IP even if it is reserved. Keystone roles should be used to limit anyone who should not have this capability).
Use Case:
- It may be that the quantum network is directly mapped at l2 to an external network not under the management of this quantum service (could be a legacy physical hosting env, or another quantum network in another data center joined by DC interconnect). You would want to be able to tell the VMs that they are on a larger subnet, but make sure that the IP allocations that quantum hands out do not conflict.
- It may be that the tenant wants to reserve a range of IP addresses for future use by a special set of servers that have not yet been deployed.
Note: lately I've been thinking that instead of specifying "reserved ranges" (i.e., the negative), perhaps we should specify "allocation_pools" (i.e., the positive). The default allocation pool would be ".2-.254", but the user could always specify alternative.
I reckon Gary's patch goes into this direction.
I'd link this bug to the bug 1008029, unless you reckon they should be addressed separately.