Peter Eisentraut <peter.eisentr...@2ndquadrant.com> writes: > I think the real fix is to have pg_upgrade copy over the statistics. > They are right there after all, just take them.
Well, it's not "just take them", we need to develop some infrastructure that will permit coping with cross-version differences in the stats storage. This isn't insoluble, but it will take some work. We had a prior discussion that trailed off without much result: https://www.postgresql.org/message-id/flat/724322880.K8vzik8zPz%40abook regards, tom lane
