This procedure worked for our HE, which is on Skylake.
I think I have a process that should work for moving our EPYC clusters to
4.3. If it works this weekend I will post it for others.
Ryan
On Thu, Feb 7, 2019 at 12:06 PM Simone Tiraboschi <stirabos(a)redhat.com>
wrote:
On Thu, Feb 7, 2019 at 7:15 PM Juhani Rautiainen <
juhani.rautiainen(a)gmail.com> wrote:
>
>
> On Thu, Feb 7, 2019 at 6:52 PM Simone Tiraboschi <stirabos(a)redhat.com>
> wrote:
>
>>
>>
>> For an hosted-engine cluster we have a manual workaround procedure
>> documented here:
>>
https://bugzilla.redhat.com/show_bug.cgi?id=1672859#c1
>>
>>
> I managed to upgrade my Epyc cluster with those steps. I made new cluster
> with Epyc CPU Type and cluster already in 4.3 level. Starting engine in new
> cluster complained something about not finding vm with that uuid but it
> still started engine fine. When all nodes were in new cluster I still
> couldn't upgrade old cluster because engine was complaining that couple of
> VM's couldn't be upgraded (something to do with custom level). I moved them
> to new cluster too. Had to just change networks to management for the move.
> After that I could upgrade old cluster to Epyc and 4.3 level. Then I just
> moved VM's and nodes back (same steps but backwards). After that you can
> remove the extra cluster and raise datacenter to 4.3 level.
>
> -Juhani
>
Thanks for the report!
We definitively have to figure out a better upgrade flow when a cluster
CPU change is required/advised.