Rolf =?iso-8859-1?q?=D8stvik?= <rolfostvik@xxxxxxxx> writes: > I have a simple query which uses 32ms on 7.4.14 and 1015ms on 8.2.0. There's something awfully strange about that 8.2 plan --- if it knew that it'd have to scan all of uut_result_subset (which it should have known, if the stats were up-to-date), why did it use an indexscan rather than a seqscan? Are you sure you haven't tweaked any parameters you didn't tell us about, such as setting enable_seqscan = off? regards, tom lane
