X-MimeOLE: Produced By Microsoft Exchange V6.5
Received: by onstor-exch02.onstor.net 
	id <01C790F7.F1ABA71B@onstor-exch02.onstor.net>; Mon, 7 May 2007 15:34:58 -0700
MIME-Version: 1.0
Content-Type: text/plain;
	charset="us-ascii"
Content-Transfer-Encoding: quoted-printable
Content-class: urn:content-classes:message
Subject: support for rational thinking
Date: Mon, 7 May 2007 15:34:58 -0700
Message-ID: <BB375AF679D4A34E9CA8DFA650E2B04E0399497F@onstor-exch02.onstor.net>
X-MS-Has-Attach: 
X-MS-TNEF-Correlator: 
Thread-Topic: support for rational thinking
Thread-Index: AceOnRqLEU3H4lhhSPeqgmrsEFCkZAACUCO2AATd4jAAAHcUIAABRCrQAAOHIjAAANEs1QAA+iFQAABpIYQAAGBM8AAAdtcqAABqwgAAAEAUfwAAJmllAAGD1oIAhOJyQA==
From: "Tim Gardner" <tim.gardner@onstor.com>
To: "Andy Sharp" <andy.sharp@onstor.com>



-----Original Message-----
From: Jerry Lopatin=20
Sent: Saturday, May 05, 2007 12:10 AM
To: Paul Hammer; Tim Gardner; Narayan Venkat; Raghu Murthy; Sandrine
Boulanger; Ian Brown
Cc: Andrew LeFebvre; dl-Delorean Core Team
Subject: Re: Minutes from 5/3 Delorean Team Meeting

Well, it sounds like linux solves the install time issue.  That sounds
pretty significant to me!

Jerry
--------------------------
Sent from my Blackberry; please excuse my typing!=20

-----Original Message-----
From: Paul Hammer
To: Tim Gardner; Narayan Venkat; Raghu Murthy; Sandrine Boulanger; Ian
Brown
CC: Andrew LeFebvre; dl-Delorean Core Team
Sent: Fri May 04 23:33:35 2007
Subject: RE: Minutes from 5/3 Delorean Team Meeting

Do not see where we need a Linux version of Bobcat at all (no
requirement or business opportunity that I see), don't see any issues
that this solves for our customers (no better quality, feature set or
performance), seems to create another release with no new customer
visible features/capabilities. Don't think we should venture down this
path. BTW, this would not force any of our current customers to run
this, they simply could stay on the BSD line for the next year or two if
they deem the kernel change to be something that they are freighted of..

________________________________

From: Tim Gardner
Sent: Fri 5/4/2007 11:22 PM
To: Narayan Venkat; Raghu Murthy; Sandrine Boulanger; Paul Hammer; Ian
Brown
Cc: Andrew LeFebvre; dl-Delorean Core Team
Subject: RE: Minutes from 5/3 Delorean Team Meeting


=20
We have already made many improvements and defect fixes to the upgrade
code in Delorean.
Bobcat/linux could be ready for FCS by Sept of this year.

________________________________

From: Narayan Venkat
Sent: Fri 5/4/2007 11:17 PM
To: Tim Gardner; Raghu Murthy; Sandrine Boulanger; Paul Hammer; Ian
Brown
Cc: Andrew LeFebvre; dl-Delorean Core Team
Subject: RE: Minutes from 5/3 Delorean Team Meeting



The challenge is that we are still 9 months away from releasing a
bobcat/linux or cougar (optimistically).  How do you propose we address
in a timely manner the numerous installs that need to be done from now
until linux or cougar?

=20

Narayan Venkat=20
Vice President, Marketing=20
ONStor, Inc.=20
Tel: (408) 963-2404=20
Cell: (408) 221-4297.=20

________________________________

From: Tim Gardner=20
Sent: Friday, May 04, 2007 11:14 PM
To: Narayan Venkat; Raghu Murthy; Sandrine Boulanger; Paul Hammer; Ian
Brown
Cc: Andrew LeFebvre; dl-Delorean Core Team
Subject: RE: Minutes from 5/3 Delorean Team Meeting

=20

The current upgrade functionality already does a, b, and c. To make it
faster we will need to rewrite the

compact flash driver for openBSD. Or we could invest our development
time in the bobcat/linux and cougar

projects where we are already planning on rewriting the upgrade code.

=20

________________________________

From: Narayan Venkat
Sent: Fri 5/4/2007 10:58 PM
To: Tim Gardner; Raghu Murthy; Sandrine Boulanger; Paul Hammer; Ian
Brown
Cc: Andrew LeFebvre; dl-Delorean Core Team
Subject: RE: Minutes from 5/3 Delorean Team Meeting

Tim,

=20

I know this is about installing Delorean.  I presume that we can write a
generic program to address points (b) and (c) and run it on OpenBSD.  If
done well it could be usable for the new upgrade utility in Zonda.

=20

Narayan Venkat=20
Vice President, Marketing=20
ONStor, Inc.=20
Tel: (408) 963-2404=20
Cell: (408) 221-4297.=20

________________________________

From: Tim Gardner=20
Sent: Friday, May 04, 2007 10:48 PM
To: Narayan Venkat; Raghu Murthy; Sandrine Boulanger; Paul Hammer; Ian
Brown
Cc: Andrew LeFebvre; dl-Delorean Core Team
Subject: RE: Minutes from 5/3 Delorean Team Meeting

=20

But this email thread is about installing Delorean. It is not about
building the next generation

install/upgrade functionality.

=20

I just want everyone to understand that we are simply specing out the
requirements for a temporary standalone

utility to install 3.0. Once you have 3.0 installed, you will never
again use this standalone install hack.

You will use the existing (and much improved in Delorean) system upgrade
utility - at least until we

invest the time to rewrite it. Perhaps we will do that in Zonda or post
Zonda.

=20

________________________________

From: Narayan Venkat
Sent: Fri 5/4/2007 10:37 PM
To: Tim Gardner; Raghu Murthy; Sandrine Boulanger; Paul Hammer; Ian
Brown
Cc: Andrew LeFebvre; dl-Delorean Core Team
Subject: RE: Minutes from 5/3 Delorean Team Meeting

I understand the issues with pre 2.2 installed systems.  We can't do
much given that you are starting with a code that has constraints.  But
my point has to do with DeLorean install going forward.  Multiple
reboots and 45 mins upgrade times are unacceptable.=20

=20

Multiple steps (within reasons) are fine.  For e.g.,  loading and
updating the secondary flash prior to a reboot is acceptable.  At a
macro level the flow should be as follows:

=20

a)       Get the latest binary from the ONStor website (or from FTP
server / local server) and store on share / volume

b)       Compare and update the secondary flash

c)       Reboot of the secondary flash.

=20

Steps (b) and (c) shouldn't take more than 5 mins.

=20

Narayan Venkat=20
Vice President, Marketing=20
ONStor, Inc.=20
Tel: (408) 963-2404=20
Cell: (408) 221-4297.=20

________________________________

From: Tim Gardner=20
Sent: Friday, May 04, 2007 10:22 PM
To: Narayan Venkat; Raghu Murthy; Sandrine Boulanger; Paul Hammer; Ian
Brown
Cc: Andrew LeFebvre; dl-Delorean Core Team
Subject: RE: Minutes from 5/3 Delorean Team Meeting

=20

Is it a requirement that the install actually succeed?

You can have an install procedure that works that involves two reboots
or you can have

a single reboot procedure that frequently fails. Please pick one.

=20

Unfortunately this is the situation we are in. We can't change the fact
that the pre 2.2 installed

code base has issues that require a reboot to clear to ensure that the
install works.

=20

Everyone should also get off this requirement of a "one step install".
Not gonna happen.

We will provide a procedure that minimizes the need to babysit the
install but I guarantee that

whatever we end up building is going to involve multiple steps at the
start of the install and probably

a few steps at the end.

=20

________________________________

From: Narayan Venkat
Sent: Fri 5/4/2007 9:42 PM
To: Raghu Murthy; Sandrine Boulanger; Paul Hammer; Ian Brown
Cc: Andrew LeFebvre; dl-Delorean Core Team
Subject: RE: Minutes from 5/3 Delorean Team Meeting

The more reboots you need the longer it takes to upgrade.  Let me be
very clear!   A filer should shutdown and reboot only once!  Just
because virtual servers are running on other nodes doesn't mean that
it's okay to have any number of reboots.  Let's not constrain ourselves
to the current procedure.  Would it be helpful if we wrote up the
desired procedure for you or do you guys prefer to design it yourself?

=20

Narayan Venkat=20
Vice President, Marketing=20
ONStor, Inc.=20
Tel: (408) 963-2404=20
Cell: (408) 221-4297.=20

________________________________

From: Raghu Murthy=20
Sent: Friday, May 04, 2007 8:09 PM
To: Narayan Venkat; Sandrine Boulanger; Paul Hammer; Ian Brown
Cc: Andrew LeFebvre; dl-Delorean Core Team
Subject: RE: Minutes from 5/3 Delorean Team Meeting

=20

=20

As long as the filer is going to stay down for the duration of upgrade,
as per the current upgrade procedure, it does not matter how many
reboots are needed per filer.

=20

As per current procedure, the first reboot of a filer is needed to xfr
all the vservers over to other filers and create a "clean" environment
for starting the upgrade process.  After that first reboot, the filer is
effectively down while the upgrade process builds the OS on the
secondary flash. When that is done, a second reboot ensures a boot-up on
the secondary flash.

=20

This means that single filer sites will stay down for 1-3 hrs for the
upgrade. In 2 filer configurations, the down time would be restricted to
the first reboot of first filer's upgrade process. Subsequently when the
2nd filer starts the upgrade process the first reboot of the second
filer would fail the vservers back to the first filer creating a second
down time window. Once the 2nd filer reboots, it moves to the new code
and at that time the customer can move vservers over for load balancing
purposes

Raghu=20

________________________________

From: Narayan Venkat=20
Sent: Friday, May 04, 2007 7:20 PM
To: Sandrine Boulanger; Paul Hammer; Ian Brown
Cc: Raghu Murthy; Andrew LeFebvre; dl-Delorean Core Team
Subject: RE: Minutes from 5/3 Delorean Team Meeting

=20

I would say that only one reboot should be required.  Can we do this in
the Delorean timeframe or is it too big a problem to address?

=20

Narayan Venkat=20
Vice President, Marketing=20
ONStor, Inc.=20
Tel: (408) 963-2404=20
Cell: (408) 221-4297.=20

________________________________

From: Sandrine Boulanger=20
Sent: Friday, May 04, 2007 7:12 PM
To: Paul Hammer; Ian Brown
Cc: Raghu Murthy; Andrew LeFebvre; dl-Delorean Core Team
Subject: RE: Minutes from 5/3 Delorean Team Meeting

=20

Raghu and I came up with the following requirements:

=20

Single step upgrade

Should handle error conditions

Should be able to upgrade to 3.0 from any release

Indicate once upgrade is complete

Make sure vsvrs have failed over after first reboot before starting the
upgrade (except for single node cluster), to ensure availability

If upgrading from release prior to 2.2, need to include compare step
after copying the data

=20

Optional:

Cluster wide

=20

________________________________

From: Paul Hammer=20
Sent: Friday, May 04, 2007 4:46 PM
To: Ian Brown
Cc: Raghu Murthy; Andrew LeFebvre; dl-Delorean Core Team
Subject: RE: Minutes from 5/3 Delorean Team Meeting

=20

4:00, hence the need to see it today and have time to create a design
and schedule.

=20

________________________________

From: Ian Brown
Sent: Fri 5/4/2007 3:37 PM
To: Paul Hammer
Cc: Raghu Murthy; Andrew LeFebvre; dl-Delorean Core Team
Subject: Re: Minutes from 5/3 Delorean Team Meeting

at what time on monday?=20

=20

Ian

=20

On May 4, 2007, at 3:05 PM, Paul Hammer wrote:

=20

Hey Raghu,

=20

Can you send the req. to the alias soon, we need to have this understood
and scooped by Monday.

=20

Thanks,

=20

-Paul

=20

________________________________

From: Raghu Murthy
Sent: Fri 5/4/2007 10:53 AM
To: Andrew LeFebvre; dl-Delorean Core Team; Ian Brown
Subject: RE: Minutes from 5/3 Delorean Team Meeting

Andy,

The action item on the single step Upgrade requirements from CS will be
handled by Ian. Ian already has already requirements documents from the
past and will update that document to make it current and he will
connect with Sandrine to close the loop on this action item.

Thanks

Raghu

=20

_____________________________________________
From: Andrew LeFebvre
Sent: Thursday, May 03, 2007 5:03 PM
To: dl-Delorean Core Team
Subject: Minutes from 5/3 Delorean Team Meeting

All,

If you missed today's meeting, then you can see what was presented by
using the following link:

\\mightydog\Program Management\Delorean\Weekly Team
Meetings\05-2007\05-03-2007\Team Meeting Agenda 5-3-07.ppt
<file://mightydog/Program%20Management/Delorean/Weekly%20Team%20Meetings
/05-2007/05-03-2007/Team%20Meeting%20Agenda%205-3-07.ppt>=20

Key decisions and actions from the meeting were as follows:

	1.	The CRG and SAG are ready to go to the printer.
Discussed whether or not Dennis should proceed given the expected FCS
schedule delay.  We did not come to a conclusion on this yet.=20
	2.	Jonathan offered to help look at the i-node and
subdirectory creation duration problem, so we'll leave this open for
another week as far as backing away from the test limit commitment we
made at the start of the project.=20
	3.	Shared the decision from the Monday upgrade meeting that
the product will initially be upgraded with flash cards until DEV
develops and QA tests a single step upgrade process.=20
	4.	Jay committed to developing a single step 35 minute
upgrade process.  He scoped the effort at 4 weeks of DEV time and 2
weeks of QA testing.=20
	5.	DEV work will not begin on the single step upgrade
process until the requirements are defined.  Sandrine and Raghu agreed
to document what is required.=20
	6.	Sandrine to develop a set of test cases for upgrade
testing.  Her initial focus with be on flash install.=20
	7.	There needs to be a detailed review of the Installation
Guide before Delorean ships.=20
	8.	DEV would like to keep the manual 2-step upgrade process
in the works as a fallback option for CS or SE.  However, they don't
want it to be an externally published upgrade mechanism.  This means
changes need to be put in place in 2.2.3 or whatever follows depending
on the final Delorean schedule.  Team agreed this was a good thing to
do, and Andy took the action to discuss this with Jerry.  Andy talked to
Jerry this afternoon about this tactic, and he agreed it was a good
idea.=20
	9.	Defect close rate is up and defect find rate is down!
But one week doesn't make a trend . . .=20
	10.	Based on the broad distribution of defects across the
code base, Sandrine announced QE plans to go through another complete
round of regression testing.=20
	11.	There was a lot of discussion about when we will be
ready for an RC build, but in the end, without Marketing and SE present,
we did not feel there was a quorum to make an official proposal.  The
plan is to discuss this more in tomorrow's special meeting with Jerry
and come up with a date that we all agree with.  It is important that
all team members attend.=20

=20

Please feel free to reply back to all with this message if you recall
anything important I may have missed.

Regards,

Andy

=20

