On Tue, 2010-02-09 at 17:14 -0800, Erik Jones wrote:
> Anyways, I realized that the dump run with zero_damaged_pages does
> actually finish.
Yeah, it should finish, it's just a question of whether the warnings
continue, and if you need to keep zero_damaged_pages on to keep reading.
> Also, I found that I can actually select all of the data by doing
> per-day queries to cause data access to be done via index scans since
> there is a date column indexed; I'm guessing that's because that
> avoids having to read the data pages' headers?
Hmm... I don't think that will actually avoid the issue. My guess is
that those pages happened to be cached from an earlier read with
zero_damaged_pages on. An index scan does not use the ring buffer I was
talking about, so the pages are more likely to stay in cache much
longer. I believe that's what's happening, and the issue is just more
hidden than before.
Regards,
Jeff Davis
--
Sent via pgsql-general mailing list (pgsql-general@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-general
