On Thu, Mar 09, 2006 at 12:37:52PM +1100, Noel Faux wrote:
> I've been watching the post: Re: [GENERAL] Fixing up a corrupted toast table
> In there they mention deletion of the bad rows from the table based on
> the citid. If I could come up with a def of a back row, would this
> work, or are there other issues?
If you have a corrupt tuple within an otherwise good block then you
can try deleting that tuple, but if the block header is corrupt
then you have no way of addressing any of that block's tuples.
Errors implying a bad tuple include "missing chunk number" and
"invalid memory alloc request size"; but "invalid page header in
block" means the block itself is bad.
--
Michael Fuhr