Fred Habash <fmhabash@xxxxxxxxx> writes: > According to pgstattuple, dead_tuple_count = 0. If this is the case, then > what other explanations do we have? Could be that old tuple(s) now require freezing. > I mean, how can I find out what blocker > session is holding the bufferpin to terminate it? I don't think there's any ready way to discover that from SQL level. regards, tom lane