Hi, On 2023-11-15 16:32:48 -0500, Robert Haas wrote: > On Mon, Nov 13, 2023 at 8:26 PM Andres Freund <and...@anarazel.de> wrote: > > I think this undersells the situation a bit. We right now do > > FreeSpaceMapVacuumRange() for 8GB of data (VACUUM_FSM_EVERY_PAGES) in the > > main > > fork, while holding an exclusive page level lock. > > That sounds fairly horrific?
It's decidedly not great, indeed. I couldn't come up with a clear risk of deadlock, but I wouldn't want to bet against there being a deadlock risk. I think the rarity of it does ameliorate the performance issues to some degree. Thoughts on whether to backpatch? It'd probably be at least a bit painful, there have been a lot of changes in the surrounding code in the last 5 years. - Andres
