On Fri, Dec 20, 2013 at 5:28 AM, Anand Kumar, Karthik
<Karthik.AnandKumar@classmates.com> wrote:
> HI,
>
> We have an issue with possibly corrupt data in our postgresql server. Errors
> like:
>
> ERROR: index "photos_p00_n2" contains unexpected zero page at block 0
> ERROR: invalid page header in block 12707 of relation
> pg_tblspc/5020557/PG_9.1_201105231/16393/9014673
>
> Thanks to all the suggestions from this list. We are in the process of
> moving our database out to a different server, and we'll then set
> zero_dameged_pages to on, run a full vacuum and reindex.
>
> The question I have is – what is the best method to transfer the data over
> to ensure we don't copy over bad/corrupt data? I would think a filesystem
> based copy (rsync, etc) should be avoided, and a pg_dump with a new initdb
> is best?
You should go with pg_dump if you are able to get a clean dump. Such
block errors happen because of hardware issues, so you are not safe
from additional failures that might happen while you do a copy of the
existing data folder to a new system.
Regards,
--
Michael
