Hi Denis,
That sound really strange. I would suspect some storage problems or something. As i told you earlier, output of --vm-status may shed light on that issue.
Unfortunately, I can't replicate it at the moment due to the need to keep the VMs up.
Did you tried to migrate form bare metal engine to the hosted engine?
Yes, I used this procedure: http://www.ovirt.org/documentation/self-hosted/chap-Migrating_from_Bare_Meta... Essentially, I used a brand new host not joined to the cluster to deploy the Hosted Engine VM.
Engine is responsible for starting those VMs. As you had no engine, there was no one to start them. Hosted Engine tools are only responsible for the engine VM, not other VMs.
I could not find out why the engine would not start from the logs I looked at. I didn't have the time to spend on it as I had to get the VMs up and running
I know, there exists 'bare metal - to - hosted engine' migration procedure, but i doubt i knew it good enough. If i remember correctly, you need to take a backup of your bare metal engine database, run migration preparation script, that will handle spm_id duplications, deploy your first HE host, restore database from the backup, deploy more HE hosts. I'm not sure if those steps are correct and would better ask Martin about migration process.
I did all these steps as per the URL above, and it did not report any errors during the process. The Hosted Engine VM started fine, but it did not appear in the list of VMs. I think the problem here was that the list of display types was incorrectly written in the hosted engine properties file. I was still left with the issue that the Hosted Engine could not be migrated to any other host. It was suggested to re-install the other hosts with the 'deploy hosted engine' option (which was missing in the official documentation). This didn't fix the issue so it was suggested that the host_id was incorrect (as it did not reflect the SPM ID of the host. I fixed this, then restarted the cluster...with the result that the engine would not start, and no VMs started. I could not see any storage errors in any of the logs I looked at, but it had not been a problem previously when rebooting hosts (though I'd never restarted the whole cluster before). When I used the old bare metal engine, I could get into the GUI to start the VMs, not sure why they didn't come up automatically. I'd like to get it working and will work with the person who takes it over to do this. I'd like to see it succeed so eventually we could use oVirt as a proof of concept to replace VMWare with RHEV. Everyone's help has been great, but unfortunately it hasn't been entirely smooth sailing (for this migration) so far. Thanks again, Cam