The official online guide of openstack forces us to have two network interfaces on all the compute nodes; one to connect to the public network and another for the private network. This is the defined architecture:
Why is it so? Why can't we connect all the compute nodes and the controller through one public switch?
This is reference architecture and good practice in production environment. It is suggested to segregate the internal traffic like vm to vm and other process communication. If you have only NIC on compute nodes you can still make it work the way you are describing. I have done it in lab setup. I have done all this in single box with single NIC as well. You can play around all this in Lab. For production I suggest you follow the official doc. Hope it is clear.
I tried making it work using one NIC but got stuck in enabling the networking. I have described my problem with that here: https://ask.openstack.org/en/question/11725/how-to-enable-networking-when-both-compute-and-controller-have-only-one-network-interface-which-is-public/
Asked: 2014-02-13 23:16:21 -0500
Seen: 1,074 times
Last updated: Feb 13 '14
Do we need to assign IP to both the interfaces or is it that the second network interface acts only as a bridge with no IP allotted to it....???