[Federated-fs] FedFS Meeting Minutes, 9/25/2008

James Lentini jlentini at netapp.com
Mon Sep 29 07:22:36 PDT 2008


FedFS Meeting Minutes, 9/25/2008
--------------------------------

Attendees
---------

Craig Everhart (NetApp)
Paul Lemahieu (EMC)
James Lentini (NetApp)
Manoj Naik (IBM/ARC)
Renu Tewari (IBM/ARC)
Robert Thurlow (Sun)

Minutes
-------

+ IETF publication plan

 The draft documents will be re-published tomorrow (9/26) as NFSv4 wg drafts.
 [Editors note: they were republished]

 The NFSv4 wg charter needs to be updated to include federated namespaces.

 James Lentini sent out a draft of the additional text for the charter. Please 
 reply with comments.

 Two design principals that we want to retain when moving to the NFSv4 wg:
 - ability to support other (non-NFS) protocols
 - no client-server protocol changes
 
+ root fileset

  This discussion was fairly free flowing. We discussed a number of issues:

  There had been some questions on the Push versus Pull model. Craig got 
  answers on these from Ellard that resolved his questions.

  What does the root fileset contain? The first level could be arbitrarily 
  large. There would be top level directories for each member of the federation. 

  What is the problem the root fileset addresses? We have multiple NSDBs and
  we want them to have a single root fileset. We don't want the root to be 
  controlled by one NSDB since this is a federation.

  If there is only one master nsdb, then we know how to implement this.

  Given there is not one master, then we need to have a replication.

  Suppose two companies merge and they have two namespaces that need to be
  merged. How can the root fileset address this issue?

  Does one NSDB root imply that the root NSDB must be replicated using 
  a single LDAP implementation? In other words, if there were only one root 
  NSDB then the particular LDAP implementation (e.g. OpenLDAP version X.Y) 
  used for that NSDB would need to be used throughout the federation since 
  LDAP replication is implementation specific. 

  Does a single root NSDB mean that there is an admin with more privileges 
  than other admins?

  The root admin has control of their area and other admins have control over
  their areas.

  How is the root fileset implemented? Does the fileserver export a pseudo
  filesystem or an on disk structure? The server needs to present the root 
  fileset, but how this is implemented does not need to be defined. 

  Need to resolve schema of root file set and resolve single NSDB versus
  replicated NSDB issue.

Next week:
 - security considerations and requirements
 - root fileset, particularly the issue of one root NSDB versus multiple root NSDBs


More information about the Federated-fs mailing list