X-MimeOLE: Produced By Microsoft Exchange V6.5
Received: by onstor-exch02.onstor.net 
	id <01C8650E.9C759929@onstor-exch02.onstor.net>; Fri, 1 Feb 2008 13:11:20 -0700
MIME-Version: 1.0
Content-Type: text/plain;
	charset="us-ascii"
Content-Transfer-Encoding: quoted-printable
Content-class: urn:content-classes:message
Subject: RE: data corruption problem in submittal 5
Date: Fri, 1 Feb 2008 13:11:20 -0700
Message-ID: <BB375AF679D4A34E9CA8DFA650E2B04E0812AB83@onstor-exch02.onstor.net>
In-Reply-To: <BB375AF679D4A34E9CA8DFA650E2B04E0812A97D@onstor-exch02.onstor.net>
X-MS-Has-Attach: 
X-MS-TNEF-Correlator: 
Thread-Topic: data corruption problem in submittal 5
Thread-Index: Achk75KPUo/xNMLbSjyeVikik3qHnQACoZTwAAR9wCA=
From: "Jobi Ariyamannil" <jobi.ariyamannil@onstor.com>
To: "Raj Kumar" <raj.kumar@onstor.com>,
	"Jeff Miller" <jeff.miller@onstor.com>,
	"dl-Cougar" <dl-Cougar@onstor.com>

Most of the runtime consistency checks in filesystem code (checksum,
read-verify etc) are for checking metadata consistency only.

We need to develop some test cases which write data in various ways
(sequential/random/sparse, aligned/unaligned, append/overwrite/truncate
etc) and verify the writes by reading back, in parallel with various
other optional tests (snapshots, icache/bcache stress, dump, quota etc).
Some of these requirements are documented at
http://wiki/wiki/FS/Testing_Requirements.

Regards,
Jobi

-----Original Message-----
From: Raj Kumar=20
Sent: Friday, February 01, 2008 9:55 AM
To: Jeff Miller; dl-Cougar
Subject: RE: data corruption problem in submittal 5

We EEK the volume regularly, so that should detect if there are any meta
data corruptions.

As for as the user data, tests based on treecreate/treeverify,
tar/untar, several compress/uncompress routines, Winsmash, etc are being
used.

We are also setting up filer to enable FS runtime consistency check
(fscmd rcc) and see if we can detect any issues with data integrity.

-----Original Message-----
From: Jeff Miller=20
Sent: Friday, February 01, 2008 8:29 AM
To: dl-Cougar
Subject: data corruption problem in submittal 5

Change 27512, which went in after submittal 5, solves a serious data
corruption problem.  The corruption would occur during write to disk and
a few bytes (probably 16) would be incorrect in the data received by the
disk.  This problem showed up in Max's spec testing.

How much data integrity testing are we doing?  Can this problem be
easily reproduced somehow using sub5, and the fix then verified with
sub6?

Another thing to keep in mind is that an data written before submittal 6
may be
corrupted.

Jeff

