Hi,

I've got a fairly well running mitaka cloud, running with four different networks for public, internal, vxlan, and storage. I am using it for courses on openstack, running multiple instances for different student groups.

Currently, I am using a scripted installation mechanism to install the different cloud instances (with their 6 nodes for compute, ctrl, neutron, cinder+swift), on different IP networks, where each node is actually a standard(non-openstack) KVM VM.

What I would like to do, is to be able to clone an instance running on a certain range of IP prefixes, into another range of IP-prefixes, instead of having to install everything from scratch, on the new VMs configured for the different IP prefixes.

So effectively: - Shutdown services. - Run script on all nodes(compute, neutron, ctrl, etc) to change ip-addresses - reboot. - Done.

What I DONT KNOW, is how often the databases contain "static" references to the nodes actual IP addresses? Of course the endpoints/service catalog is different, but is there a lot more that is actually "fixed info"?

I can of course dump the database, but since it will contain a lot of addresses that are volatile and depends on what has been currently registered by the components, its not easy to detect what I would actually have to change.

Any ideas, help, comment, on wether this is something I should keep trying to accomplish?

Br, Taisto