We have a cluster multiple clusters of 2 x M-8000 NSP with Active-Active High Availability configured, when we had to do the software upgrade which is usually released twice yearly. We faced major downtime as our assymtrical cluster was out of synch, and the whole data center off line.
We had to make an emergency change in our network to upgrade one at a time and lose the connectivity. When we contacted McAfee they said that we need to bring a third M-8000 just for the sake of the upgrade!!!!!!
I did not buy it, it is either that something is wrong or we bought the wrong product?
Juniper and other company have much neater way of doing it.
Were your 8000's in an XC cluster? Were you not able to remove 1 M-8000 at a time from the load balancer, perform upgrade, reinstate to the load balancer, then perform the upgrade to the other?