On Mon, Oct 18, 2021 at 7:18 AM Michael Paquier <michael@paquier.xyz> wrote:
On Sun, Oct 17, 2021 at 10:13:48PM +0300, Florents Tselai wrote: > I did look into VACUUM(full) for it’s PROCESS_TOAST option which > makes sense, but the thing is I already had a cron-ed VACUUM (full) > which I ended up disabling a while back; exactly because of the > double-space requirement.
Please note that VACUUM FULL does not enforce a recompression on existing values. See commit dbab0c0, that disabled this choice as it introduced a noticeable performance penality in some cases when looking at the compression type of the vacuumed table attributes:
Oh dang, I missed that this was reverted. Thanks for pointing that out!
