Thread: reproducing this issue on PG 8.0.0 ERROR: index "patient_pkey" is not a btree

surabhi.ahuja wrote:
>  hi
> I am using Postgres 8.0.0 and we found this issue "ERROR:  index
> "patient_pkey" is not a btree" I have been informed that we should
> shift to Postgres 8.0.9
>
> I discussed this with my team member and they are asking if we can
> upgrade to the latest Postgres version i.e. 8.1.5

The difference is that updating to 8.0.9 would be dead simple (just
upgrade the binaries), but upgrading to 8.1.5 is more difficult because
it requires you to dump and restore the database, and there may be minor
incompatibilities for your apps.  So an idea might be to have a
medium-term plan to migrate to a new major series (say 8.1 or 8.2), and
a very short term plan to migrate to 8.0.9.

> I have some questions regarding this:
> 1.will this vesion solve the problem that I have mentioned?

If it is solved by 8.0.9 then it will be solved by 8.1.5 as well.  It
may be something independent of Postgres, for example hardware failure,
in which case it probably won't be solved by either.

--
Alvaro Herrera                                http://www.CommandPrompt.com/
PostgreSQL Replication, Consulting, Custom Development, 24x7 support

On Tue, Nov 14, 2006 at 01:58:49PM +0530, surabhi.ahuja wrote:
> hi
> I am using Postgres 8.0.0 and we found this issue "ERROR:  index "patient_pkey" is not a btree"
> I have been informed that we should shift to Postgres 8.0.9

I beleive you've asked this a few times already. As a rule it always
best to run the latest minor release, because bugs do get fixed, that's
why there are minor releases.

Do you get this often? Because a reindex fixes it, obviously.

> I have some questions regarding this:
> 1.will this vesion solve the problem that I have mentioned?

No idea.

> 2. If we install postgres 8.1.5 instead of Postgres 8.0.0 I ll have to build my c++ application again right?

I don't think so. I can't imagine why. The client libraries work across
several server versions.

> 3. I am currently using postgresql-8.0-310.jdbc3.jar, for java
> applications. Would I have to change this jar as well? and if yes
> where can I find it?

That's for the backend, right? Or not? If it's for the client end you
shouldn't need to change that either.

> 4. the most important question is : Is there any test case that you ran in order to confirm that the above issue will
notoccur with PG 8.1.5. I need this test case that I can run in order to propose that we should upgrade to 8.1.5 

Do you have a test-case that causes it to happen? That's the only test
there is.

> 5. Can you please provide a link to this Bug. I want to just see the proble, resolution, verification of this bug.

I don't think there's been a bug relating to this. In just about every
case confirmed, it's been memory corruption or disk corruption. As
someone put it:

  This [error] says that one of two fields that should never change, in
  fixed positions in the first block of a btree index, didn't have the
  right values.

No-one has come across an bug that would cause that. Basically, for
more info we need to see the hexdump of the first 8k of the index file
to see why the system thinks it's not a btree. Can you do that?

Have a nice day,
--
Martijn van Oosterhout   <kleptog@svana.org>   http://svana.org/kleptog/
> From each according to his ability. To each according to his ability to litigate.