Re: What is wrong with hashed index usage?

On Mon, 22 Apr 2002 14:15:37 -0700
"Dann Corbit" <DCorbit(at)connx(dot)com> wrote:
> From here:
> http://osdb.sourceforge.net/
> We find this quote:
> "For you long-suffering OSDB PostgreSQL users, we offer
>
> --postgresql=no_hash_index
>
> to work around the hash index problems of OSDB with PostgreSQL V7.1 and
> 7.2. As always, let us know of any problems. May the source be with
> you!"
>
> Does anyone know what the above is all about?

Yes -- search the list archives, or check the PostgreSQL docs. This problem
has been brought up several times: hash indexes deadlock under concurrent
load. A run of pgbench with a reasonably high concurrency level (10 or 15)
produces the problem consistently.

Previously, I had volunteered to fix this, but

(a) I'm busy with the PREPARE/EXECUTE stuff at the moment.

(b) I'm not sure it's worth the investment of time: AFAIK,
hash indexes don't have many advantages over btrees for
scalar data.

On the other hand, if someone steps forward with some data on a
specific advantage that hash indexes have over btrees, I don't
expect that the concurrency problems should be too difficult to
solve.

Cheers,

Neil

--
Neil Conway <neilconway(at)rogers(dot)com>
PGP Key ID: DB3C29FC