Re: Index question

From: <stevedhoward_at_gmail.com>
Date: Tue, 25 Aug 2009 18:09:09 -0700 (PDT)
Message-ID: <c0d7d88b-2269-490c-98e9-550c01b82a39_at_v36g2000yqv.googlegroups.com>

On Aug 25, 5:32 pm, "Jonathan Lewis" <jonat..._at_jlcomp.demon.co.uk> wrote:
> <stevedhow..._at_gmail.com> wrote in message

> I'd favour an optimizer error over corruption - the plan that
> got the right answer had to find the bottom left corner of each
> index and scan from there - so there wasn't any corruption in
> that area of the index partitions.
>
> If you have time, I'd suggest coalesce rather than rebuild.
>
> On a related topic - if you've got an index that's inconsistent
> with the table, then a coalesce (or rebuild that scans the index)
> could still leave you with an inconsistent index.
>

Thanks for your thoughts, Jonathan.

This is interesting on a couple of fronts:

1. I would define anything that returns incorrect data as "corruption" (or at least the people would that pay Oracle, and more importantly me ;)). I guess if the data itself is structurally correct, you're saying that if a different plan were selected, the correct data *would* be returned? That sounds like a nasty bug.
2. I never knew an index rebuild could just scan the existing index and rebuild from that. That sounds like the definition of "garbage in garbage out". Why would Oracle do it this way?

Thanks,

Steve Received on Tue Aug 25 2009 - 20:09:09 CDT