I have previously tried to import a Confluence back up that was linked to a Crowd instance directory, but it would not import to OnDemand saying that it didn't know anything about the Crowd directory the backup was linked to. Is there a way to overcome this issue? Unfortunately, I cannot get into this Confluence instance to change settings as it was hosted on the Rackspace Cloud and I have long since shutdown the server, so that IP address has been assigned to a different Rackspace Cloud customer and Confluence won't migrate to another machine with a different IP address.
Confluence imports to OnDemand have to be done space-by-space. Please check this document about Wiki importing inside OnDemand: https://confluence.atlassian.com/display/AOD/Importing+wiki+data