On Thu, Apr 11, 2019 at 11:13:17AM -0600, Michael Lewis wrote:
> Wouldn't "dead but not yet removable" be high if there were long running
> transactions holding onto old row versions?
You got it right. You need to look at the number behind the tuples
dead, but not removable which is usually a pattern caused by long
running transactions holding snapshot references which prevents
autovacuum and/or vacuum to do their homework.
--
Michael
