Re: Volatile write caches on macOS and Windows, redux

On Tue, 18 Jul 2023 at 05:29, Thomas Munro <thomas.munro@gmail.com> wrote:
> It's possible that fcntl(F_FULLFSYNC) might fail with ENOSUPP or other
> errors in obscure cases (eg unusual file systems).  In that case, you
> could manually lower fsync to just "on" and do your own research on
> whether power loss can toast your database, but that doesn't seem like
> a reason for us not to ship good solid defaults for typical users.

Is this the only reason why you're suggesting adding fsync=full,
instead of simply always setting F_FULLFSYNC when fsync=true on MacOS.
If so, I'm not sure we really gain anything by this tri-state. I think
people either care about data loss on power loss, or they don't. I
doubt many people want his third intermediate option, which afaict
basically means lose data on powerloss less often than fsync=false but
still lose data most of the time.

If you're going to keep this tri-state for MacOS, then it still seems
nicer to me to "fix" fsync=true on MacOS and introduce a fsync=partial
or something. Then defaults are the same across platforms and anyone
setting fsync=yes currently in their postgresql.conf would get the
fixed behaviour on upgrade.