I found this link, related to 4.3 to 4.4 for RHV, that seems to somehow confirm the need of a "spare" host....

I don't know if the author reads the ML.

But this goes completely against what seems to be present inside the RHV docs, where, as described in my previous post:

"

If you decide to use a new host, you must assign a unique name to the new host and then add it to the existing cluster before you begin the upgrade procedure.

"

And also the oVirt docs that are substantially based on RHV ones:

"

When upgrading oVirt Engine, it is recommended that you use one of the existing hosts. If you decide to use a new host, you must assign a unique name to the new host and then add it to the existing cluster before you begin the upgrade procedure.

. . .

"

So I would have in this case the same problem related to wrong fingerprint, or I should be forced to copy fingerprint before scratching the host and reuse it (if supported passing from node in version 7 to node in version 8 of the OS).....

It seems strange this problem didn't arise before...

the phrase:

"

The upgraded host with the 4.4 self-hosted engine reports that HA mode is active,...

"

lets think that the host name remains consistent with a pre-existing one and as a reinstall is mandatory (7-->8) I don't see how it could work...