Re: [nfsv4] New Version Notification for draft-haynes-nfsv4-flex-filesv2-00.txt

[Thomas Haynes <[e-mail]>]

On Aug 8, 2017, at 3:17 AM, David Noveck <[e-mail]<mailto:[e-mail]>> wrote:

> And Ben, the other reason not to use a typed opaque auth blob
> is because the  ffds_user and ffds_group are strings, and do
> not fit into AUTH_UNIX.

I don''t see how the fact that ffds_user and ffds_group, being strings,
do not fit into AUTH_UNIX can be argument for retaining them.

My argument would be the minimal set of changes necessary to make a ffv2 implementation.

One thing about the synthetic ids are that they are used to fence as much as they are used
as credentials. This may mean that we need both the ids and the auth field.



If
the existence of AUTH_UNIX is relevant to the matter, it would argue
for deleting ffds_user and ffds_group and subsuming them both
within ffds_auth.

The situation is the same regarding the fact there might be arrays of these
things.  Clearly, if one of these items needs to be an array, they all do..
If that's the case, it would be better to have one potentially troublesome
array rather than three.


After having some time to think about this, I believe I have difference between these and the other
fields. Both the stateid and file handles are generated by the storage device - with loose coupling,
we cannot do anything to modify these. However, the synthetic ids, the auth token, and the
mode bits (which are never discussed but which play an important role), can be set by
the metadata server. So we can mandate that they are all the same across each instance
of the data file. If they are not, then the client will inform the metadata server in a LAYOUT_ERROR
(or LAYOUT_RETURN) that it cannot access the data and then the metadata server
can repair the access (mode/ids/auth token).

The auth token is not directly modifiable, but as the storage device validates it to the private auth server,
this should be sufficient.





On Mon, Aug 7, 2017 at 8:25 PM, Thomas Haynes <[e-mail]<mailto:[e-mail]>> wrote:
So here is a first hack at Flex Files v2. I thought it would come in much smaller than 11 pages,
so perhaps some pruning will be necessary.

I’m aware Ben had some comments on:

   Can I suggest putting in a typed opaque auth blob instead of explicit
   synthetic uid/gid fields?

I have purposely done:

   /// struct ffv2_data_server4 {
   ///     deviceid4               ffds_deviceid;
   ///     uint32_t                ffds_efficiency;
   ///     stateid4                ffds_stateid<>;
   ///     nfs_fh4                 ffds_fh_vers<>;
   ///     fattr4_owner            ffds_user;
   ///     fattr4_owner_group      ffds_group;
   ///     opaque_auth             ffds_auth;
   /// };

Because of AI1:

   [[AI1: after the lesson learned from ffds_stateid, we either need to put an
   array here or define all of the file handles to share the same
   credentials.  And as Olga points out in her email, this gets big
   fast.  Especially if we throw in many mirrored copies!  --TH]]

I.e., the ffds_auth is inside an array of ffv2_data_server4 which is
in itself inside an array of ffv2_mirror4 AND it could be argued that
ffds_auth (and ffds_user/ffds_group) need to be an array.

Even if we decide to make all FHs share the same credentials, we
eventually can hit a limit of the number of mirrored copies.

And Ben, the other reason not to use a typed opaque auth blob
is because the  ffds_user and ffds_group are strings, and do
not fit into AUTH_UNIX. We could define AUTH_SYNTHETIC. :-)

Begin forwarded message:

From: <[e-mail]<mailto:[e-mail]>>
Subject: New Version Notification for draft-haynes-nfsv4-flex-filesv2-00.txt
Date: August 7, 2017 at 5:11:45 PM PDT
To: Thomas Haynes <[e-mail]<mailto:[e-mail]>>


A new version of I-D, draft-haynes-nfsv4-flex-filesv2-00.txt
has been successfully submitted by Thomas Haynes and posted to the
IETF repository.

Name: draft-haynes-nfsv4-flex-filesv2
Revision: 00
Title: Parallel NFS (pNFS) Flexible File Layout v2
Document date: 2017-08-07
Group: Individual Submission
Pages: 11
URL:            https://www.ietf.org/internet-drafts/draft-haynes-nfsv4-flex-filesv2-00.txt
Status:         https://datatracker.ietf.org/doc/draft-haynes-nfsv4-flex-filesv2/
Htmlized:       https://tools.ietf.org/html/draft-haynes-nfsv4-flex-filesv2-00
Htmlized:       https://datatracker.ietf.org/doc/html/draft-haynes-nfsv4-flex-filesv2-00


Abstract:
  The Parallel Network File System (pNFS) allows a separation between
  the metadata (onto a metadata server) and data (onto a storage
  device) for a file.  The flexible file layout type is an extension to
  pNFS which allows the use of storage devices in a fashion such that
  they require only a quite limited degree of interaction with the
  metadata server, using already existing protocols.  This document
  describes two extensions to the flexible file layout type to allow
  for multiple stateids for tightly coupled NFSv4 models and an
  additional security mechanism for loosely coupled models.




Please note that it may take a couple of minutes from the time of submission
until the htmlized version and diff are available at tools.ietf.org<http://tools.ietf.org/>.

The IETF Secretariat



_______________________________________________
nfsv4 mailing list
[e-mail]<mailto:[e-mail]>
https://www.ietf.org/mailman/listinfo/nfsv4