Updating (patching) vCenter HA environment within the same "level", like from 6.5.0a to 6.5.0d, or 6.5.0 U1 to 6.5.0 U1e is well documented in VMware Docs and various blogs. In short: enter HA maintenance mode, update witness, update passive, switch-over, update new-passive (all via appliancesh with mounted update iso on the nodes). So far so good.
But I haven't found much info specificly about updating/upgrading between "levels", like from 6.5.0 (GA) to 6.5.0 Update 1. And suddenly I'm in the possition where I need to do this update (upgrade).
I'm bringing this up, since I talked to one of the speakers (VMware dude, can't remember the name) at a vCenter HA session at VMworld 2017 in Barcelona, and he told me that for patching within the same "level" (GA, U1) the update method mentioned above is supported/fine, but upgrade between levels, like GA to Update 1 for instance (or later U1 to U2), you must kill the HA cluster, update the single node and then recreate the HA cluster. The reason was that the database schema most likely would change and a sync could be messing things up.
Any thoughs? Anyone here have any other information about upgrading from 6.5.0 GA (or any of the "letter patches" from 65GA) to any of the 6.5.0 Update 1 versions?
I have not seen any issues in minor updates (i.e., 6.5U1a to 6.5U1b) and so wouldn't destroy vCHA for those; but I would for major updates (I don't consider this an upgrade), for example 6.5 to 6.5U1d.
I would agree with that statement to destroy the vCHA configuration and redeploy after the update is finished. No doubt later versions of vSphere will refine this process to make such swings and roundabouts superfluous, but in its 1.0 version with vSphere 6.5, vCHA definitely has some quirks and bugs that need to be ironed out. I've found it necessary in several circumstances to delete and redeploy the entire vCHA configuration, and I certainly would in this case to ward off any potential issues which would make that required at a later point in time.
daphnissov - So you (would) do both patching ("same levels") and upgrade ("between levels") by the means of destroying the HA-cluster, patch single node, and redeploy HA-cluster? Or only the last part, between levels?
I have not seen any issues in minor updates (i.e., 6.5U1a to 6.5U1b) and so wouldn't destroy vCHA for those; but I would for major updates (I don't consider this an upgrade), for example 6.5 to 6.5U1d.
Then we see it the same way. That was how I saw me handling it, and good to see I wasn't off track.