On Wed, Mar 19, 2025 at 4:54 AM Christophe Pettus <x...@thebuild.com> wrote: > > Everything seems to point to the vacuum free space map operation, since it > would have a lot of work to do in that particular situation, it happens at > just the right place in the vacuum cycle, and its resource consumption is not > throttled the way the regular vacuum operation is.
Do you know how big the FSM was? As others have said, it could be worth adding a phase to pg_stat_progress_vacuum. I know you said you saw this all the way back to 15, but it made me wonder if this would actually get worse after 17 and the tidstore optimization -- since a large table with indexes requiring multiple vacuum passes would actually end up having its FSM vacuumed sooner. - Melanie
