On Fri, Jan 24, 2025 at 09:53:09AM -0500, Andrew Dunstan wrote: > On 2025-01-23 Th 4:06 PM, Robert Haas wrote: >> On Thu, Jan 23, 2025 at 3:51 PM Andres Freund <and...@anarazel.de> wrote: >> > I wonder if it's a mistake that a role membership that has WITH ADMIN on >> > another role is silently removed if the member role is removed. We e.g. do >> > *not* do that for pg_auth_members.grantor: >> > >> > ERROR: 2BP01: role "r1" cannot be dropped because some objects depend on >> > it >> > DETAIL: privileges for membership of role r2 in role r3 >> Yeah, I'm not sure about this either, but this is the kind of thing I >> was thinking about when I replied before, saying that maybe dropping >> role B shouldn't just succeed. Maybe dropping a role that doesn't have >> privileges to administer any other role should be different than >> dropping one that does. >> > > That seems reasonable and consistent with what we do elsewhere, as Andres > noted.
+1, if this is doable, I would prefer that over a new predefined role. A pg_admin_all role might still be useful, but IMHO it's a rather big hammer for this particular problem. -- nathan