On Tue, Jan 14, 2020 at 7:18 AM Michael Paquier <michael@paquier.xyz> wrote:
>
> On Mon, Jan 13, 2020 at 03:39:32PM +0100, Julien Rouhaud wrote:
> > Actually, can't it create a security hazard, for instance if you call
> > pg_file_sync() on a heap file and the calls errors out, since it's
> > bypassing data_sync_retry?
>
> Are you mistaking security with durability here?
Yes, data durability sorry.
> By default, the
> function proposed is only executable by a superuser, so that's not
> really a security concern.. But I agree that failing to detect a
> PANIC on a fsync for a sensitive Postgres file could lead to
> corruptions. That's why we PANIC these days.
Exactly. My concern is that some superuser may not be aware that
pg_file_sync could actually corrupt data, so there should be a big red
warning explaining that.