Suppose you have n processes, n > 2. You want to have agreement amongst them that one is to be active. So they need to vote amonst each other to determine which one is active.
All processes may fail at any time, we want to have one process active if possible, but ...
We must never have two active at the same time, so if they can't be sure it is better to have no-one active. (Ie. we want to avoid split brain)
The only available communication mechanism between them is pub-sub messaging (not point to point).
One or more databases are available, but no one database should be a single point of failure. Ie. it would be very undesirabloe if all the processes were available to work, and the were prevented from doing so by the loss of a single database.
Design? What messages need to be published?