Re: BUG #17245: Index corruption involving deduplicated entries

Hi,

On 2021-10-29 22:52:39 +0300, Kamigishi Rei wrote:
> On 29.10.2021 21:36, Kamigishi Rei wrote:
> > btree index "azurlane_wiki.mediawiki.page_main_title":
> >     ERROR:  item order invariant violated for index "page_main_title"
> >     DETAIL:  Lower index tid=(17,157) (points to heap tid=(540,5))
> > higher index tid=(17,158) (points to heap tid=(540,5)) page
> > lsn=2/A019DD78.
> 
> I forgot to include WAL archive grepped for the "page" table's heap page
> 540. Note that 540,5 is mentioned below just before VACUUM. `SELECT h.* FROM
> heap_page_items(get_raw_page('page',540)) h;` returns lp_flags=3 for lp=5
> for the snapshot (though I am not sure if lp=5 here is the same 5 as in the
> tid=(540,5) tuple).

Could you search for btree WAL records before the following records? Most
importantly for the corrupted page in the corrupted index, but other ones
might be interesting as well

> rmgr: Heap2       len (rec/tot):     53/  7937, tx:          0, lsn:
> 2/90CEF528, prev 2/90CEF4E8, desc: VACUUM nunused 3, blkref #0: rel
> 1663/19243/19560 blk 540 FPW
> rmgr: Heap2       len (rec/tot):     53/  8109, tx:          0, lsn:
> 2/97ED2598, prev 2/97ED2558, desc: VACUUM nunused 1, blkref #0: rel
> 1663/19243/19560 blk 540 FPW

The currently most likely seeming explanation is that somehow the index
entries pointing to the heap items removed here weren't removed
themselves. Subsequently leading to visible corruption once those tids are
reused.

Greetings,

Andres Freund