Per this article VMWare claims that you can upgrade a UAG with zero downtime.
I have a problem with this claim. They forgot the * in the claim.
*You can upgrade a UAG with zero downtime if you wait until everyone disconnects and then reconnects through the non Quiesce Mode UAG.
Quiesce mode doesn't magically move users to the other UAG. It simply stops users from being able to connect through that UAG. What's the difference between Quiesce mode and simply disabling the UAG in the Load Balancer and waiting it out for no more users to be connected to that UAG?
So in a 24/7 environment there would be down time because the existing users that are connected to the UAG to be upgrade must log out and log back in. That sounds like an inconvenience and it definately sounds like down time (although short) for the user.
Someone please enlighten me or prove me wrong. Am missing something or is the claim of zero downtime just not true or a terrible marketing claim?
The idea is you run the old versions of UAG side by side with the new ones until all users have logged out from the old versions. This can be after several days. Thereby from a user point of view there is no need to force them to logout and back in. It will happen by it self over a couple days.
The zero down time claim is not due to anything magical on the UAG's part though. It's simply doing what a load balancer would do. It's draining the UAG of connections over time.
So in a DR scenario where one of the UAG's suddenly stops accepting connections or there is a sudden need to move all users connected to one UAG to the other, one would not have the luxury of waiting around. Everyone connected to a specific UAG would drop. Cold turkey. Disconnected. Business down. Money lost.
This is also one of my points. So everyone knows what DRS is correct? If the UAG could do with existing connections as what DRS does with Virtual Machines than we would truly have magic on our hands or maybe to truly kick ass, all of the UAG's would need to act as a single unit (Perhaps a VIP). Then if one of the UAG's went down it wouldn't matter because the connections would simply stay up and just be routed via the UAG that is up. I would brag about this feature to my wife, my kids, my parents, my grandma, and even my pet Budgie, "Pepper".
That is basically true.. If a UAG with active sessions goes down the user would have to reconnect.. Not sure how much money lost that means but I guess that is depending heavily on the business. If authentication is needed depends on implementation.
Having sessions synced between UAGs would be cool but is a completely different architecture. That said it is quite interesting idea.. Not sure it is in scope for UAG though. At least not at the present time..
But that said, your initial statement regarding zero downtime upgrade from a user perspective is possible. Yes, we do rely on the load balancer for much of that. But we do support different versions of UAG working side by side.. So we have done some work
Well you guys have done well so far and come very far. And it is nice to be able to run different versions of the UAG at the same time. Thank you for your response
We have found Quiesce Mode very helpful when working in a multi-vendor environment.
Some of our Managed Services clients use a 3rd-party to provide network services, including load balancing, while we manage their Horizon environment. In this model, it can be very difficult to work with the 3rd-party to disable load balancing members for patching or upgrades.
Quiesce Mode allows us to take control of the draining of a UAG, which means we can upgrade that appliance with zero down time for end users without involving the network provider.