Re: [nfsv4] More layout stuff: layout return on close

"Sorin Faibish" <sfaibish@emc.com> Thu, 25 November 2010 20:04 UTC

Return-Path: <sfaibish@emc.com>
X-Original-To: nfsv4@core3.amsl.com
Delivered-To: nfsv4@core3.amsl.com
Received: from localhost (localhost [127.0.0.1]) by core3.amsl.com (Postfix) with ESMTP id 4CF9D28C0CE for <nfsv4@core3.amsl.com>; Thu, 25 Nov 2010 12:04:10 -0800 (PST)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -6.599
X-Spam-Level:
X-Spam-Status: No, score=-6.599 tagged_above=-999 required=5 tests=[BAYES_00=-2.599, RCVD_IN_DNSWL_MED=-4]
Received: from mail.ietf.org ([64.170.98.32]) by localhost (core3.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id Q8wGRXuFcAfc for <nfsv4@core3.amsl.com>; Thu, 25 Nov 2010 12:04:09 -0800 (PST)
Received: from mexforward.lss.emc.com (mexforward.lss.emc.com [128.222.32.20]) by core3.amsl.com (Postfix) with ESMTP id 0AA4E3A6A6D for <nfsv4@ietf.org>; Thu, 25 Nov 2010 12:04:08 -0800 (PST)
Received: from hop04-l1d11-si03.isus.emc.com (HOP04-L1D11-SI03.isus.emc.com [10.254.111.23]) by mexforward.lss.emc.com (Switch-3.4.3/Switch-3.4.3) with ESMTP id oAPK57De011101 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-SHA bits=256 verify=NO); Thu, 25 Nov 2010 15:05:07 -0500
Received: from mailhub.lss.emc.com (mailhub.lss.emc.com [10.254.221.145]) by hop04-l1d11-si03.isus.emc.com (RSA Interceptor); Thu, 25 Nov 2010 15:05:00 -0500
Received: from sfaibish1.corp.emc.com (USSAGARGAL3C.corp.emc.com [10.4.131.21]) by mailhub.lss.emc.com (Switch-3.4.3/Switch-3.4.3) with ESMTP id oAPK4qUn026675; Thu, 25 Nov 2010 15:04:52 -0500
Date: Thu, 25 Nov 2010 15:08:36 -0500
To: david.noveck@emc.com, tigran.mkrtchyan@desy.de, nfsv4@ietf.org
From: Sorin Faibish <sfaibish@emc.com>
Content-Type: text/plain; format="flowed"; delsp="yes"; charset="iso-8859-15"
MIME-Version: 1.0
References: <1290552459.30508.4.camel@heimdal.trondhjem.org> <4CECC382.5020207@panasas.com> <4CECE104.5080009@desy.de> <1290610592.3234.20.camel@heimdal.trondhjem.org> <4CEE283D.5040903@desy.de> <1290692948.2911.7.camel@heimdal.trondhjem.org> <4CEE8AA9.1030109@desy.de> <BF3BB6D12298F54B89C8DCC1E4073D8002A5F2B6@CORPUSMX50A.corp.emc.com>
Content-Transfer-Encoding: 8bit
Message-ID: <op.vmqswmrhrwwil4@sfaibish1.corp.emc.com>
In-Reply-To: <BF3BB6D12298F54B89C8DCC1E4073D8002A5F2B6@CORPUSMX50A.corp.emc.com>
User-Agent: Opera Mail/9.63 (Win32)
X-EMM-MHVC: 1
X-EMM-MFVC: 1
Cc: Trond.Myklebust@netapp.com
Subject: Re: [nfsv4] More layout stuff: layout return on close
X-BeenThere: nfsv4@ietf.org
X-Mailman-Version: 2.1.9
Precedence: list
List-Id: NFSv4 Working Group <nfsv4.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/listinfo/nfsv4>, <mailto:nfsv4-request@ietf.org?subject=unsubscribe>
List-Archive: <http://www.ietf.org/mail-archive/web/nfsv4>
List-Post: <mailto:nfsv4@ietf.org>
List-Help: <mailto:nfsv4-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/nfsv4>, <mailto:nfsv4-request@ietf.org?subject=subscribe>
X-List-Received-Date: Thu, 25 Nov 2010 20:04:10 -0000

On Thu, 25 Nov 2010 13:56:19 -0500, <david.noveck@emc.com> wrote:

> The text says it is a "directive".  Unfortunately, it does not clarify
> that with a "MAY", "SHOULD", or "MUST".
>
> Still a "directive" seems quite a way from a "hint".
>
> The problem here is that the "MUST" which follows this is about the
> server, not the client.  Saying the server "MUST" recall, does suggest
> that the "directive" for the client is not all that forceful, and you
> could go all the way down to "hint" I suppose.  It's more a hint of the
> form "Why don't you return that segment (hint, hint), since if you don't
> I'm going to have recall it anyway."  If it is a hint, it is not a hint
> suggesting that you could somehow keep the layout segment.
>
> The question I think we have is whether we can justifiably address this
> via the errata mechanism.  Are we fixing a mistake, rather than just
> improving the protocol, which would be a nice thing for v4.2, and not
> something for v4.1.
>
> I would argue that that the "MUST" is inconsistent with the forgetful
> client model which the spec allows and so is we can treat this as a
> mistake.  Also, we are not making any existing client or server
> non-compliant.
>
> My understanding of the correction:
>
>      Flag still described as a "directive".
>
>      Follow-up is that client MUST forget (i.e. no longer use) the
> segments
>      marked that way.  Client MAY advise server of their non-use, by
> LAYOUTRETURN
>      but server has enough information to infer layout segment non-use
> without
>      that.
>
>      Server MUST assume such layout segments are not to be further used
> when
>      CLOSE occurs.
>     Server MAY recall such layout segments if they have not been
> returned, but
>      there is no requirement that he do so.  He SHOULD ensure that if
> they are
>      used subsequently by the client, this is treated as an error.
I am not confortable with the posibility of an "error" such as the
client will use the segments following a CLOSE. How would the server know
that the clients used the segments. You do not imply that the server has
to check anything.

>
> Comments?
>
>
> -----Original Message-----
> From: nfsv4-bounces@ietf.org [mailto:nfsv4-bounces@ietf.org] On Behalf
> Of Tigran Mkrtchyan
> Sent: Thursday, November 25, 2010 11:11 AM
> To: nfsv4@ietf.org
> Cc: Trond Myklebust
> Subject: Re: [nfsv4] More layout stuff: layout return on close
>
> On 11/25/2010 02:49 PM, Trond Myklebust wrote:
>> On Thu, 2010-11-25 at 10:11 +0100, Tigran Mkrtchyan wrote:
>>> On 11/24/2010 03:56 PM, Trond Myklebust wrote:
>>>> On Wed, 2010-11-24 at 10:55 +0100, Tigran Mkrtchyan wrote:
>>>>> We are in a quite weird situation here. From one side servers asks
> a
>>>>> client to return
>>>>> layout on close, but from other side you have to keep track of the
>>>>> layout and recall it to
>>>>> ensure that the client will not use it any more.  I would love to
> see
>>>>> this part of the spec to
>>>>> be more deterministic. But it we change it not to be a more strong
>>>>> requirements, then
>>>>> some of the current clients will become not complaint with the
> spec. Are
>>>>> we ok with that?
>>>> I fail to see your point. Which clients would become spec
> incompliant if
>>>> we relax the requirement that they must send a layoutreturn before
> close
>>>> when the server sets the logr_return_on_close flag? Relaxing a
>>>> requirement is not equivalent to prohibiting the behaviour.
>>> If I read it correctly, then spec change will allow servers to forget
>>> client layout on close.
>> Only for layout segments that have return on close set.
>>
>>> The new clients will simply forget layout on close. Nevertheless the
> old
>> The forgetful client model is already part of the spec. This is not
> new
>> client behaviour. Please see RFC5661 section 12.5.5.1.
>>
>>> ( existing ) clients
>>> are allowed to reuse it as return_on_close was ( are ) only a hint.
>> That is incorrect: layout return_on_close is not a hint. In RFC5661
>> section 18.43.3, the server is required to recall the layout if the
>> client doesn't return it.
>>
>> The point is that if that is _mandatory_ behaviour (i.e. the layout
>> segment MUST be returned) then why not just say that if the client
>> doesn't return it, then that layout segment is automatically revoked?
> No
>> recall should be necessary, since both parties know that it is being
>> revoked.
>>
>       it MUST be prepared to recall the layout in the
>      case in which the client fails to return the layout before close.
>
> As there is no 'client MUST return' I read it as a hint.
> Anyway, I think we are talking the same thing. So, yes, I am happy with
> that.
>
> Tigran.
>
>> Trond
>>
>
> _______________________________________________
> nfsv4 mailing list
> nfsv4@ietf.org
> https://www.ietf.org/mailman/listinfo/nfsv4
>
> _______________________________________________
> nfsv4 mailing list
> nfsv4@ietf.org
> https://www.ietf.org/mailman/listinfo/nfsv4
>
>



-- 
Best Regards
Sorin Faibish
Corporate Distinguished Engineer
Unified Storage Division

        EMC²
where information lives

Phone: 508-435-1000 x 48545
Cellphone: 617-510-0422
Email : sfaibish@emc.com