RE: issues with LSI 9260

Postgres calculates a checksum on each 16kb page of data and records it in
the page header. When it reads the page back, it calculates the checksum
against the data it's retrieved and compares it to the one that was
recorded in the header. When the checksums don't match, it indicates
corruption in the page. This error is showing that this problem has
occurred.

>From observation, when we've had this problem, in some cases the page
header appears to have been corrupted and in other cases there's evident
corruption elsewhere in the page. In one case, it appears a copy of a page
header from some other page had been written at the 8k point.




Derrick Lobo
VP, Network & Integrations


NA:  1-416-350-9660 ext 254
UK : +44 (0) 20 7107 0036 ext 254
Web: http://www.givex.com 

Experience. Excellence. 

Givex, an award winning provider of client acquisition and retention
programs. 

Before printing this email please consider the impact on the environment
and help protect it. 

CONFIDENTIALITY NOTE: This email contains information which may be
privileged, confidential or exempt or prohibited from disclosure under
applicable law. If the reader of this message is not the intended
recipient, or the employee or agent responsible for delivering the message
to the intended recipient, you are hereby NOTIFIED that any dissemination,
distribution, retention, archiving or copying of this message and/or the
contents thereof is strictly prohibited. If you have received this message
in error or in contravention of the above, please notify the sender
immediately.

-----Original Message-----
From: Manuel Bouyer [mailto:[email protected]] 
Sent: Thursday, May 28, 2015 11:57 AM
To: Derrick Lobo
Cc: [email protected]; [email protected];
[email protected]
Subject: Re: issues with LSI 9260

On Thu, May 28, 2015 at 11:53:26AM -0400, Derrick Lobo wrote:
> The config we have 4 x 2tb disk with raid10 setup on the lsi .. the 
> disk is configured using gpt
> 
> Im not seeing any errors in normal operation or on the kernel however 
> when we build a database we get these errors on the postgress logs
> 
> Block 1077542 of base/16391/24778
> Block 848733 of base/16391/24777

I have no idea what this means. Are you sure this is errors ?

> 
> These are certrans and certificate, respectively.
> 
> dd if=/data_local/cert63/base/16391/24778.16 of=badpage.certrans.63
> bs=16384 count=1 skip=28966
> dd if=/data_local/cert63/base/16391/24777.12 of=badpage.certificate.63
> bs=16384 count=1 skip=62301

Does this report errors ?

> 
> 
> when we take off the lsi and connect a sata disk and use gpt to 
> configure the drive we are able to build the database with no page
errors.

It may also be timing-related, I would expect the LSI to be much faster,
especially for writes

--
Manuel Bouyer <[email protected]>
     NetBSD: 26 ans d'experience feront toujours la difference
--