I did eventually figure out the issue: I mis-understood the question about "cloud-init" as "engine-init", once I answered yes to cloud-init, I was allowed to set a root password and run engine-setup.
I am curious about some of your questions too, actually.
I was following the instructions from here:
In that, he said that in order to enable the gluster support in the engine, one has to do it manually. It seemed odd, or perhaps a misunderstanding of the procedure.
My understanding was that after deploying my three hosts and gluster, I had to deploy the hosted-appliance manually on host 1 so that after the initial deploy, but before it was taken over by HA, I could go in and set the gluster service to on, then finish the setup with the reboot, etc.
The other thing I was wondering about was setting up the additional hosts. The instructions say to ssh to each host and deploy the engine there through the ssh command line. The tool itself says that I shouldn't be doing it, I should be running through the website. However, when I did that, I had a number of issues that I wasn't able to correct, and when I did it through ssh, I had the best functioning ovirt build yet.
One of the issues I had was when I tried to create a private network for gluster sync'ing, the web interface saw the gluster IPs for hosts 2 and 3 (although host 1 was added with its ovirt/real management IP, and the two networks were not routed). The web itnerface host add ended up adding hosts 2 and 3 with the gluster IP, and thus things broke a lot. I had no means of overriding those settings that I saw through the web interface. When adding through SSH, I had a lot more control over adding the hosts, and was able to add them "correctly".
I suspect this means the overall procedure was in error, and I'd like to learn the "better" way to do it.
--Jim