Re: "invalid page header in block 597621 of relation..."error

On 26/11/05 4:48 pm, "Tom Lane" <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:

> Adam Witney <awitney(at)sgul(dot)ac(dot)uk> writes:
>> I deleted the two datasets in mba_data_base that were affected by the empty
>> pages, I also deleted the relevant two rows in measured_bioassay_base... But
>> maybe it didn't do the right thing with the toast table for these two rows?
>
> Evidently the missing data in the toast table is associated with yet a
> different dataset.
>
> I'd suggest first looking into the toast table to see if you can confirm
> that the missing data corresponds to a swath of zeroed-out pages. If
> that's the case then it gives even more urgency to the need to find out
> what's going wrong with your filesystem (or possibly your disk drive,
> but my gut feel is that this is a kernel filesystem problem).
>
> The other thing you'd need to do is figure out which dataset you have to
> reload. A tedious way to do this is something like
> select sum(length(bigfield)) from maintable where dataset = 'xxx';
> for various values of xxx until you see the error.

Well I tracked down which row went wrong and deleted that dataset also, the
backups worked fine and it seems to be ok now.

Not really sure what caused all this, all these datasets (190 in total) went
in in one batch the other day, so for some reason 3 of them got screwed up.

Anyway I have upgraded the box to linux 2.6.14.3, so I will keep an eye on
it and see how things go.

Thanks again for your help Tom

Adam

--
This message has been scanned for viruses and
dangerous content by MailScanner, and is
believed to be clean.