Wondering if it might be worth trying a restart of the computer since sometimes an uninstall isn't really complete until then. Also, have different versions of Tableau Server been installed on there before?
After the restart I would check the directories once again just to be sure there wasn't something left over. Though not 100% sure if we're looking for regkeys during install to see if we think Tableau Server is still installed.
Thanks for the reply Mark. Yes, I did try that but no such luck. I've deleted all traces of Tableau directories I could find. Do I need to delete the postgres database separately?
I'll probably just shut down the workers altogether when I'm able to have our VM admins start them back up again (I don't have that access) so that the backup primary can't push anything out to them again.
My goal is to do a clean single node install on the backup primary and then save that tabsvc.yml file separately for future upgrades. We wrote an automatic failover script (because c'mon...manual failover is no failover at all IMO) and we do regular failover tests which replaces that tabsvc.yml file with the one from the primary. Hopefully that will force the backup primary to leave the worker nodes out of it's upgrade since they were already done with the primary.
My hunch is that the upgrade article was written with the assumption that the backup primary is still configured as a single node because in theory it has never had a failover. In any case, the upgrade process didn't work for me as advertised. Just wanted to let folks know and see if anyone had any hints to share.