Ask Your Question

Revision history [back]

We also had that issue and it came down to a permission issue on the compute nodes.

In the case you use a NFS export to store your instances, you need to make sure that your nova users have the same UID and GID and that the permissions for /var/lib/nova are correct on all(!) compute nodes.

In our case the initial scheduling worked, then nova-compute wasn't able to store the instance under /var/lib/nova/instances due to that permission problem and rescheduling started. But Neutron already assigned an IP address for the first deploy try, so our instances became ACTIVE with just 2 IPs ;)