This sounds pretty good to me, but I don't like the idea of containers on a VM.  One or the other.

 

I think that traffic management  (round robin) is a bit ambitious.  We really don't need it for a site with a few dozen users at best.  It gives us fallback if a node goes down, but it also increases maintenance and points-of-failure.  I have seen distributed/fallback systems fail more often than  they work.

 

If some kind of Linux bomb drops and suddenly end users are overflowing the system we can easily expand.