Re: pg recovery

[Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>]

Bernhard D Rohrer <graylion(at)sm-wg(dot)net> writes:
> Tom Lane wrote:
>> I think you've got a cross-version problem, as in the database is really
>> PG 8.0 or earlier but you're trying to run 8.1 against it.  What is in
>> the PG_VERSION file?  Have you done "pg_resetxlog -f", and if so do you
>> have the original pg_control file to put back?

> as for the versions see for yourself:
> root(at)collab:/home/adminlion# cat /var/lib/postgresql/8.1/main/PG_VERSION
> 8.1
> root(at)collab:/home/adminlion# cat 
> /olddrive/var/lib/postgresql/8.1/main/PG_VERSION
> 8.1

Hmmm ... but it sure looks like the values are offset a few fields from
where they belong ... [ meditates awhile... ]  Ah, I've sussed it: the
pg_controldata output you showed can be explained exactly by the
assumption that this copy of pg_controldata thinks time_t is 64 bits
wide, where the pg_control file actually has 32-bit-wide time_t fields.
That explains both the ridiculously large dates (quite impossible for
32-bit time_t's) and the offsetting of the following fields.

So the short answer is probably that you're trying to use a 64-bit build
of Postgres against a 32-bit database.  You need to get a matching build.

(We really need to stop using time_t in pg_control.h ...)

			regards, tom lane