"Tom Lane" <tgl@sss.pgh.pa.us> writes:
> Oooh ... I can't be sure that this is what's biting you, but I
> definitely see a bug that seems to match the symptoms. As the comments
> in index.c point out, CREATE INDEX CONCURRENTLY works like this:
>
> * validate_index() works by first gathering all the TIDs currently in the
> * index, using a bulkdelete callback that just stores the TIDs and doesn't
> * ever say "delete it". (This should be faster than a plain indexscan;
> * also, not all index AMs support full-index indexscan.) Then we sort the
> * TIDs, and finally scan the table doing a "merge join" against the TID list
> * to see which tuples are missing from the index.
>
> The scan is done using the regular heapscan code, which in 8.3 has been
> modified to enable "synchronized scanning", which means it might start
> from the middle of the table and wrap around.
Wow, I'm glad we caught this in beta. Thanks a lot to Hannes Dorbath for
testing and reporting it.
-- Gregory Stark EnterpriseDB http://www.enterprisedb.com Ask me about EnterpriseDB's RemoteDBA services!
