Re: Corrupt Table - Mailing list pgsql-general

From Tom Lane
Subject Re: Corrupt Table
Date
Msg-id 14855.968958571@sss.pgh.pa.us
Whole thread Raw
In response to Re: Corrupt Table  ("Bryan White" <bryan@arcamax.com>)
List pgsql-general
"Bryan White" <bryan@arcamax.com> writes:
> I found a core file.  I am not all that familiar with gdb but the backtrace
> looks useless:
> #0  0x8064fb4 in ?? ()
> #1  0x809da10 in ?? ()

Looks like you are running a stripped executable :-(.  You might want to
consider recompiling with debug symbols so we can get more info if this
happens again.

> Yes I am running 7.0.2.  The 4 pages in question have 0's in the first 16
> bytes but other data after that.  I see some text that look like real data.

Oh, that's interesting.  This isn't a previously known kind of failure.

These dropouts must have occurred since your last vacuum, since vacuum
would have thought that the pages are uninitialized and "fixed" them.
Perhaps they have the same cause as the problems in the other two pages.

I recall once having seen a similar kind of failure (aligned segments
of pages suddenly becoming zeroes) that turned out to be from a hardware
problem --- disk controller wasn't quite compatible with the
motherboard, or something like that, and would occasionally transfer
bad data to/from memory.  I'm not ready to blame the hardware yet, but
it's a possibility to keep in mind, particularly if you've changed
the hardware setup recently.

> Ok here is my plan:

Seems reasonable.  Good luck!

            regards, tom lane

pgsql-general by date:

Previous
From: Matthew
Date:
Subject: RE: Upgrading from 6.3.2 to 7.0.2
Next
From: "Adam Lang"
Date:
Subject: Starting postmaster at boot