On Mon, Jan 31, 2022 at 04:57:13PM +0900, Michael Paquier wrote: > On Tue, Jan 25, 2022 at 01:12:32PM +0900, Michael Paquier wrote: >> Once you remove the requirement of a running server, we have basically >> what has been recently implemented with postgres -C for >> runtime-computed GUCs, because we already go through a read of the >> control file to be able to print those GUCs with their correct >> values. This also means that it is already possible to check if a >> data folder is compatible with a set of binaries with this facility, >> as any postgres -C command with a runtime GUC would trigger this >> check. Using any of the existing runtime GUCs may be confusing, but >> that would work. And I am not really convinced that we have any need >> to add a specific GUC for this purpose, be it a sort of >> is_controlfile_valid or controlfile_checksum (CRC32 of the control >> file). > > Thinking more about this one, we can already do that, so I have > marked the patch as RwF. Perhaps we could just add a GUC, but that > feels a bit dummy.
Sorry, I missed this thread earlier. You're right, we can just do something like the following to achieve basically the same result: postgres -D . -C data_checksums Unless Vik has any objections, this can probably be marked as Withdrawn. Perhaps we can look into providing a new option for "postgres" at some point in the future, but I don't sense a ton of demand at the moment. -- Nathan Bossart Amazon Web Services: https://aws.amazon.com