Re: again trouble - Mailing list pgsql-general

From Tom Lane
Subject Re: again trouble
Date
Msg-id 9150.1026480578@sss.pgh.pa.us
Whole thread Raw
In response to again trouble  ("Henrik Steffen" <steffen@city-map.de>)
List pgsql-general
"Henrik Steffen" <steffen@city-map.de> writes:
> vacuum verbose analyze zugriffsrechte
> ERROR:  IndexSupportInitialize: no pg_index entry for index 14958692

> tom, could this again be due to a broken SD-RAM chip?

Perhaps.  It's certainly looking more and more like you've got some
kind of hardware problem ... Postgres is just not this flaky for most
people ;-).

If you wanted to dig into it, you could pg_filedump
pg_index_indrelid_index and look to see why it doesn't have an entry for
14958692.  My bet is that the entry is indeed there, but is not being
found --- either its own key is corrupted, or some nearby keys are
corrupted in a way that renders them out-of-order, causing binary search
to conclude the desired key is not present.  A single-bit error in the
key field would certainly suffice to cause such a problem.

            regards, tom lane

pgsql-general by date:

Previous
From: Tom Lane
Date:
Subject: Re: 7.2.1 optimises very badly against 7.2
Next
From: Tom Lane
Date:
Subject: Re: plpgsql Cursor Mismatched Parentheses