On Wed, 2005-06-08 at 13:39, Bruno G. Albuquerque wrote:
> I am having a weird problem here. I have the automated process to
> install PostgreSQL (8.0.1) on Windows 2000 machines. Besides installing
> the database server itself, my process does the following:
>
> 1 - Runs initdb to created the database I will be using.
> 2 - Runs a SQL script that creates the entire database and populates
> some of the tables.
> 3 - Runs another script that populates other tables (using COPY).
>
> What happens is that, sometimes, after the entire process run, I end up
> with a database with virtually all indexes corrupt! Try to use the
> indexes for anything results in "index is not a btree" error. This
> happens randomly and I can get it fixed sometimes by simply erasin
> everything and starting over.
Note that such behaviour is normally associated with bad hardware (RAM,
hard drive, controller, CPU, you name it...) But since the windows port
is pretty new, you may have found some corner case.
Do you have a "pre-packaged" test case that induces failures most of the
time or something like that?
Are these machines all basically make/model/brand/chipset etc... of
machines? Then it might be a buggy driver or something.