I'm gearing up to either upgrade or migrate to 7.6 . Aside from 6.x --> 7.x I've always upgraded. Going from 7.5 to 7.6 via upgrade I'm assuming I'll still need to manually apply HF9 post upgrade. So the way I see it I have the pain of the upgrade to 7.6 + the pain of applying the hotfix. Both of these tasks have caused me a ton of grief in production so the thought is not super appealing. The idea of building out a working 7.6 HF9 environment and migrating sounds like a cleaner way but also a lot of work building everything out from the ground up again. We have a fully distributed environment which has had a lot of tweaking done over the years and it is all really stable when not trying to get through an upgrade. Since I've not been migrating I have no experience on how clean or error prone/free the process is. I find myself stuck square in the middle of which route to take. I'm curious if others on here might share some opinions. It would be much appreciated.
A bit late in responding, however I've done several vRA 7.x to 7.6 upgrades recently covering embedded and external vRO instances. I've not seen a migration since the 6.x to 7.x days. You are correct in saying upgrade would be to 7.6 GA and then apply HF9 post upgrade.
Most of the issues I've seen during upgrades are minor like incorrect version of java for the upgrade, insufficient disk space etc which just prevent the upgrade taking place. Get the prereqs completed and clean up any stuck in progress or missing machines and you are likely to have a successful upgrade. Also check the correct permissions are in place on IaaS servers like DEM and Proxy Agents for the service account, I've seen the upgrade fail where the service account has had its rights or privileges changed after the original install and the upgrade is unable to complete the management agent upgrade for example due to insufficient access.
If something does go wrong you need to make sure you have a clean set of snapshots, and backups including your database to go back to. I've only had to resort to reverting to snapshot once, and so far luckily never needed to go back to the SQL database backup. Take snapshots frequently during the upgrade as well, so you have for example one set with all servers powered off before any work is done, one set taken after performing any pre upgrade clean up steps, one set after upgrading the appliances before upgrading IaaS, one set post upgrade pre HF install etc. Assuming you can't use Lifecycle manager to perform the upgrade, then if you do need to revert you won't always have to repeat all the steps you performed.
Timings for completing the upgrade I have seen vary massively depending on how heavily used the environment is, and how many servers are configured. How cleanly the environment was originally installed and has been maintained can also make a difference. If all components are online and working without issue before the upgrade and the environment has had regular housekeeping to remove any decommed servers etc then you increase the chances of a painless upgrade.
HF install was painless the last time I did it for an upgraded system other than having to reboot all the servers.