On Wed, Jan 18, 2023 at 1:08 PM Andres Freund <and...@anarazel.de> wrote: > I suggested nearby to only have ANALYZE dead_tuples it if there's been no > [auto]vacuum since the stats entry was created. That allows recovering from > stats resets, be it via crashes or explicitly. What do you think?
I like that idea. It's far simpler than the kind of stuff I was thinking about, and probably just as effective. Even if it introduces some unforeseen problem (which seems unlikely), we can still rely on pgstat_report_vacuum() to set things straight before too long. Are you planning on writing a patch for this? I'd be very interested in seeing this through. Could definitely review it. -- Peter Geoghegan