Ask Your Question

Revision history [back]

Of course the compute node can join the controller. It's not magic.

Your job is to install the operating system on the compute node and give Packstack its IP address and root password. Packstack takes care of the rest: It uses Puppet to provision the node with the necessary Nova and Neutron components and set up the configuration files.

One of the configuration parameters is the message queue in nova.conf, for example:

transport_url = rabbit://stackrabbit:password@192.168.1.203:5672/

When nova-compute on the compute node starts up, it attempts to get in touch with the controller via the message queue. If that succeeds, the controller understands that a new node has joined. You can then see result in the service list:

openstack compute service list

By the way, it's not true that you need a second network interface to add compute nodes.