Hi Bruce, hi Daniel, > On Oct 27, 2023, at 23:21, Bruce Momjian <bruce@xxxxxxxxxx> wrote: > > On Fri, Oct 27, 2023 at 10:45:16PM +0200, Daniel Gustafsson wrote: >>> On 27 Oct 2023, at 20:34, Paul Förster <paul.foerster@xxxxxxxxx> wrote: >> >>> a) why isn't it possible to enable checksumming while a database cluster is up? >> >> It is surprisingly complicated to enable checksums on a live cluster, a patch >> was submitted a while back but ultimately never made it into postgres. The >> below threads may shine some light on the problem: >> >> https://www.postgresql.org/message-id/flat/CABUevEx8KWhZE_XkZQpzEkZypZmBp3GbM9W90JLp%3D-7OJWBbcg%40mail.gmail.com >> https://www.postgresql.org/message-id/flat/560A2239-5DE2-4B9C-92BC-878C6822F47C%40yesql.se#10b665b2193445a3d7f24be7917a952c > > Yeah, it was a big surprise that this feature was so hard to implement > because we have _no_ infrastructure for having multiple data layouts > active in a live system. The discussion eventually made that clear. > > If we have more features that need this kind of dynamic ability, we > might revisit this feature too. Ok, I see. But unfortunately still, my questions c) and d) are unanswered. I'd especially be interested in an answer to c), i.e. is it *safe* to "pg_checksum -e" the replica instance in a patroni cluster, switch over, and then do the other one? Cheers Paul
