On Fri, Nov 30, 2018 at 9:19 PM 小威 <zxw110876189@qq.com> wrote:
> I'm working on 11.0, no beta version.and my db include BUG#15039 patch,but still occur this issue. actually,I don't
knownwhat happened.I only sure there still has a bug in my case.
I think that you'll probably need to provide a script to reproduce the
issue. Can you make the data anonymous before sending? Alternatively,
you can send it off-list. I promise that I will treat the application
specific parts as confidential.
Using amcheck would let us know if the index is completely corrupt
(e.g. tuples are in the wrong order), or if the duplicate is garbage
that has somehow become visible again, a symptom of bugs which is also
possible. It's easy to use amcheck. Maybe there is another bug that is
similar to #15309, but we need to establish if the index is truly
corrupt, and what it looks like. It would also be good to use
contrib/pageinspect to look at any blocks that amcheck shows in an
error message.
--
Peter Geoghegan