Thread: Moving filesystems

Moving filesystems

From
Michael Lush
Date:
I'm trying to set up a warm backup via log shipping, however when I
transfer the database filesystem(1) to the backup machine and start
postgres on it I get FATAL: incorrect checksum in control file.

Both machines are running postgres 8.2.5

The production server is running CentOS 5 (kernel 2.6.18.el5)
on a 4 CPU system (Dual Core AMD Opteron(tm) Processor 885)

The would be backup server is running CentOS 5.2 (kernel 2.6.18-92.el5PAE)
on a 2 CPU system (Intel(R) Xeon(TM) CPU 2.80GHz)

Should I expect this to work?  or am I doing something dumb shunting the
files over (I tar'ed them and had to chown'ed them as the backup machines
user and group ID differs)

(1) I have tried this the production and a minimal test database (one
schema one table)

--
Michael
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Michael John Lush PhD            Tel:44-1223 492626
Bioinformatician
HUGO Gene Nomenclature Committee    Email: hgnc@genenames.org
European Bioinformatics Institute
Hinxton, Cambridge
URL: http://www.genenames.org
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Re: Moving filesystems

From
Tom Lane
Date:
Michael Lush <mjlush@ebi.ac.uk> writes:
> I'm trying to set up a warm backup via log shipping, however when I
> transfer the database filesystem(1) to the backup machine and start
> postgres on it I get FATAL: incorrect checksum in control file.

> Both machines are running postgres 8.2.5

> The production server is running CentOS 5 (kernel 2.6.18.el5)
> on a 4 CPU system (Dual Core AMD Opteron(tm) Processor 885)

> The would be backup server is running CentOS 5.2 (kernel 2.6.18-92.el5PAE)
> on a 2 CPU system (Intel(R) Xeon(TM) CPU 2.80GHz)

> Should I expect this to work?

No, not if one machine is 64-bit and the other is 32-bit, which is what
I suspect given your reference to Opteron.  In general log shipping has
to have the same CPU architecture as well as the same Postgres version
on both ends.

You could possibly make it work by running a 32-bit build of Postgres
on both machines.

            regards, tom lane