nfsv41 mount semantics question

[Benny Halevy]

On Mar. 10, 2008, 6:40 +0200, Spencer Shepler <Spencer.Shepler at Sun.COM> wrote:
> As for Solaris, all NFS mounts are of type "nfs".
> There will be a new vers=xx selection in the case that

And the default?  Are you going with 4.1, falling back to
4 down to 2 if not supported?   Or will you start at 4.0
unless vers=41 is specified?

Benny

> an admin wants NFSv4.1 specifically.   Also at this point,
> there will be no option to enable/disable pNFS functionality.
> If it exists and the server supports the clients layout type
> functionality, then the client will ask for and use layouts.
> 
> Spencer
> 
> On Mar 9, 2008, at 9:33 AM, Benny Halevy wrote:
> 
>> Muntz, Daniel wrote:
>>> How about we do what Solaris does?  I don't know what that is, but  
>>> it's
>> first, please s/do/look at/ :)
>> It will be nice to be compatible with Solaris, but you have to
>> take into account the existing infrastructure and linux admins'
>> established habits.
>>
>>> probably reasonable.  Might also be nice to have an option to disable
>>> pnfs (assuming it's enabled by default by being part of 4.1),  
>>> something
>>> like '-o nopnfs' (forcing write-through the MDS).
>> Agreed.
>>
>>>   -Dan
>>>
>>> -----Original Message-----
>>> From: Dean Hildebrand [mailto:seattleplus at gmail.com]
>>> Sent: Friday, March 07, 2008 11:34 AM
>>> To: nfsv4 at linux-nfs.org
>>> Subject: nfsv41 mount semantics question
>>>
>>> At our weekly linux pNFS meeting yesterday the issue of how to  
>>> mount a
>>> nfsv4.1 server arose.  We are looking for some clarification on  
>>> how to
>>> proceed with the implementation.
>>>
>>> The first issue is with regards to whether we should use a new mount
>>> type (e.g., '-t nfs41') or instead rely on the 'nfsvers' parameter
>>> (e.g., '-t nfs nfsvers=4.1' or '-t nfs4 nfsvers=4.1').
>>>
>>> It was pointed out that the long term goal is to remove the current
>>> 'nfs4' mount type and use only a single 'nfs' mount type and rely  
>>> on the
>>> 'nfsvers' parameter to differentiate between versions.  Is this true?
>>> (Note that the current nfs man page says the nfsvers parameter is not
>>> supported for nfs4.) If so, should we implement the nfsv41 code to  
>>> use
>>> the mount type 'nfs'
>>> and a nfsvers=4.1 (or 41)?
>>>
>>> Another issue is if we move to using a single mount type of 'nfs',  
>>> what
>>> should the default value of 'nfsvers' be?  I'm assuming that for the
>>> near term it would be 'nfsvers=3'.  Would it depend on what is  
>>> compiled
>>> into the kernel?
>>>
>>> Anyways, we just want to get the conversation started at this  
>>> point so
>>> we can see what lies ahead.
>>>
>>> In summary, a few sample options:
>>> '-t nfs nfsvers=[2,3,4,4.1]'
>>> or
>>> '-t [nfs,nfs4] nfsvers=[2,3] (for nfs) or nfsvers=[4,4.1] (for nfs4)'
>>> or
>>> '-t [nfs,nfs4,nfs41] nfsvers=[2,3]' - nfsvers for nfs type only
>>>
>>> Dean
>>> _______________________________________________
>>> NFSv4 mailing list
>>> NFSv4 at linux-nfs.org
>>> http://linux-nfs.org/cgi-bin/mailman/listinfo/nfsv4
>>> _______________________________________________
>>> NFSv4 mailing list
>>> NFSv4 at linux-nfs.org
>>> http://linux-nfs.org/cgi-bin/mailman/listinfo/nfsv4
>> _______________________________________________
>> NFSv4 mailing list
>> NFSv4 at linux-nfs.org
>> http://linux-nfs.org/cgi-bin/mailman/listinfo/nfsv4
>