Ugh. I did a fresh install of HDP 2.3 and have had even more problems than before. 
(Mostly having to do with the Hive Metadata Store.) Since I have no data to preserve, I think I’m going to do a fresh install of HDP 2.2 and stick with that until I have a better understanding of what’s going on, and until HDP 2.3 has some more miles on it.
This is all on a single CentOS 6.6 box as I’m prototyping a Hadoop solution for my employer. It looked like there were some things that weren’t supported without a multi-node cluster, such as a rolling upgrade from 2.2 to 2.3, but I don’t know how many other problems I’ve had could have been attributed to that fact.