X-MimeOLE: Produced By Microsoft Exchange V6.5
Received: by onstor-exch02.onstor.net 
	id <01C7AD58.1350C851@onstor-exch02.onstor.net>; Tue, 12 Jun 2007 17:13:39 -0800
MIME-Version: 1.0
Content-Type: text/plain;
	charset="us-ascii"
Content-Transfer-Encoding: quoted-printable
Content-class: urn:content-classes:message
Subject: Additional item for our discussion of Delorean FCS at 1:00 pm Wednesday
Date: Tue, 12 Jun 2007 17:13:39 -0800
Message-ID: <BB375AF679D4A34E9CA8DFA650E2B04E041BFB60@onstor-exch02.onstor.net>
X-MS-Has-Attach: 
X-MS-TNEF-Correlator: 
Thread-Topic: Additional item for our discussion of Delorean FCS at 1:00 pm Wednesday
Thread-Index: AcetWBMknKvrsfq2QZOgg9ScLo+I0Q==
From: "Jay Michlin" <jay.michlin@onstor.com>
To: "Jerry Lopatin" <jerry.lopatin@onstor.com>,
	"Paul Hammer" <paul.hammer@onstor.com>,
	"Sandrine Boulanger" <sandrine.boulanger@onstor.com>,
	"Brian DeForest" <brian.deforest@onstor.com>
Cc: "Andrew LeFebvre" <andrew.lefebvre@onstor.com>,
	"Caeli Collins" <caeli.collins@onstor.com>

Folks,

Today at the staff meeting we also explored the defective flash shipped
to AT&T for the Delorean beta. Sandrine has examined it and found
hundreds of files missing. Our hypothesis is that BSD encountered an
exception while trying to mount the flash, and then ran fsck in repair
mode, thus doing the additional damage we observe. If this is correct,
we do not yet have a hypothesis about what caused the exception to
begimn with.

Jerry observed that silently running fsck in repair mode is a bad idea
from the perspective of supporting customers. We are altering, and
perhaps further damaging the flash without alert or notification. He
directed that we change this to verify mode.

I subsequently discussed this in SW Development and heard a different
opinion. If BSD is seeing an exception, then the only hope of getting
the filer booted and alive is to run fsck. Otherwise, the filer will be
unresponsive. Especially without a console attached, customers, and our
support staff, will see only a dead machine with no message to explain
it.

Additionally, when an ONStor CS rep gets to the machine, his first
action will be to run fsck in repair mode anyway. So the argument is
that fsck in repair mode is probably certain, whether driven by BSD or
by our engineer. In the former case, there is some hope of getting the
filer at least partly responsive earlier.

We will explore this as a second agenda item at our Delorean FCS
discussion tomorrow. From a SW Development perspective, we will do
whatever you decide. We just thought this additional thinking ought to
be aired.

jay
