This does not make much sense to me to be honest...
Please let me know what to check on OS level.
-- Vladimir
вт, 24 сент. 2019 г. в 17:35, Peter Geoghegan <pg@bowt.ie>: > > > That's slightly inconvenient, because it's an internal page -- there > is no heap TID that you can use to look at the row in the table > directly. You're going to have to look at the index page directly, and > try to make sense of pageinspect's hex representation of the data. > > The following will show details of the pair of tuples that are in the > wrong order here: > > CREATE EXTENSION pageinspect; > SELECT * FROM bt_page_items('vocabulary_phrase_key', 29915) WHERE > itemoffset in (116, 117); > > What does that look like? > > I think that the problem here is likely to have something to do with > OS collations -- this isn't a Postgres bug in all likelihood.