Re: Unfamous 'could not read block ... in file "...": read only 0 of 8192 bytes' again

On Tue, Feb 21, 2012 at 1:37 PM, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:

> Maxim Boguk <maxim(dot)boguk(at)gmail(dot)com> writes:
> >> Do you know why the mod date on the file is 2012-02-20 12:04?
>
> > Cron was attempt to populate the table once per hour after that problem
> > happened.
> > And each time it was produced the same error.
>
> That's interesting ... is there any possibility that the insertions were
> attempting to insert values that matched a previously-existing primary
> key value? I'm thinking there's no reason for the INSERT per se to be
> touching nonexistent blocks, but if for some reason the pkey index still
> had entries pointing at vanished rows (as it seems to) then the errors
> could be coming from uniqueness checks attempting to fetch those rows to
> see if they're live.
>
> regards, tom lane
>

Hi,

There isn't possibility but close to 100% new inserted values were matched
a previously-existing primary
key value.
The table is hand-made 'materialyzed view'-type statistic table which is
getting recalculated via cron.

--
Maxim Boguk
Senior Postgresql DBA.