Our Ops Manager documentation includes an example topology for production that is comprised of a Primary, Secondary, Arbiter setup. It can be found here.

This is the very first example listed with some huge caveats in the note box right above it, namely the fact that the application database for Ops Manager uses a write concern of 2, yet the diagram advocates arbiters which do not count towards this requirement. As the note suggests, losing one node means losing access to Ops Manager. To make things worse, the arbiters are on the same servers are the primaries, so if that single server goes down, even the blockstore would lose a majority.

This setup is extremely unreliable and should be removed in favor of a simplistic but redundant 3 server approach. At the very least, it should not be visible to the public and should only be provided by TSEs when specifically asked due to disk space concerns.

In general, I think we should try to avoid any prominent PSA examples for any part of the documentation.

Thanks!