Minutes for: NFS version 4 ? Beijing Tuesday, November 9, 2010 0900 - 1130 - Jade 1 ------------------------- Intro/Blue Sheets WG Administrivia - Beepy provided a reminder about the Microsoft disclosure and interested parties should review the disclosure. Draft: Federated FS WG last call summary and discussion Trond provided a review of the last call results for James. Last calls are closed with two minor issues being handled on the working group alias. James wanted to note that all work items have been completed with respect to the bi-weekly calls that he has been hosting. These FedFS calls will come to an end. If other topics are of interest, they need to be taken up on the working group alias. Draft: NFS operation over IPv6 Bhagwat provided brief overview of the I-D content and mentioned that there are two I-Ds now as per feedback from the Stockholm IETF meeting. Spencer provided feedback that the I-D quality needed to be improved to assist in moving through last call effectively. NFS operation over IPv4 and IPv6 Bhagwat explained the content of the I-D as representing issues of dual-stack deployments. Again, the overall quality of the document needs to be improved to allow for effective last call. A question arose if some of the items should be included in 3530bis -- Bhagwat needs to provide the input to Tom Haynes for inclusion. pNFS Access Permissions Check Sorin Faibish covered reviewed problem addressed. Update to the I-D added covering issues of transient errors. Next steps include additional review of the scsi and object error handling. Soring would like to close this I-D by the next IETF meeting. NFS Version 4 Protocol Tom Haynes provided a brief overview of current status of 3530bis. The I18N area is still a large issue. The intent of the I18N updates are to semantically change the RFC text to match the current implementations that exist. Note: this semantic change needs to be change and the reasons for the change need to be captured in the I-D. NFSv4 Multi-Domain Access Andy Adamson provided a brief review of the I-D content. Next steps can be found in the presentation material. The working group co-chairs will work in determining how best to move this item into the working group charter. NFS Pathless Objects Dipankar Roy introduced the I-D contents. Content can be found in the presentation and I-D. A question arose about the applications that would use the new searchable attribute space provided by the proposal. No examples were provided during the presentation and suggested that examples be provided on the working group alias. Storage De-Duplication Awareness and Sub-File Caching in NFS Mike Eisler quick overview of content. Discussion arose about the lack of having checksums cross the wire -- avoiding the exposure of algorithms used for de-dup so that a proposal could move forward more effectively. Mike would like to see interest from 3rd parties before moving forward further with the I-D. Note that this proposal does not require an NFSv4.2 -- a new layout type is introduced to achieve the capability. Metadata Striping for pNFS Mike Eisler identified that NFSv4.2 is unnecessary for this functionality. Metadata striping is provided by the introduction of a new layout type. This I-D needs more discussion/review via the WG alias. There seemed to be many questions about complexity of the proposal and how/if the client should be involved with the algorithms for striping. Extending NFS to Support Enterprise Applications Mike Eisler provided content overview for these changes. Spencer asked if the INITIALIZE operation was supposed to allocate space for the range while not writing to the underlying storage. Mike said that the implication was not intended. Mike noted that this functionality could be integrated with other proposals to handle the range of allocation, initialization, and "hole" handling. NFS space reservation operations Mike Eisler covered the content of the I-D with no questions. NFS Server-side Copy Mike Eisler covered the content with no substantive comments. NFS Version 4 Minor Version 2 unstable file creation and attribute update improvements Trond reviewed the content of the I-D and motivations. There were no substantive comments. Storage Control Extensions for NFS Version 4 Dave Noveck covered the content. One comment about inheritance was that it should follow the same model as the ACL inheritance model that is already present in the protocol. pNFS block signature and disk protection issues Sorin Faibish covered the issues that have arisen with respect to lack of ability to identify a target as part of the pNFS solution. Sorin asserts that this is needed to avoid inadvertant destruction of data. A couple of questions arose about location of the signature on an SCSI device -- no clear answer. NFSv4.1 Linux Implementation Status Andy Adamson covered the current implementation of pNFS in the Linux base. Presentation materials contain data. Minutes for: NFS version 4 ? Beijing (meeting 2) Tuesday, November 9, 2010 3:20 - 5:00 - Jade 1 A second meeting slot was picked up on Tuesday. The discussion during that meeting slot was free form given that a agenda had not been set. The discussion focused on the potential content of NFSv4.2 and how best to manage the work. For those in the room, there was consistent that NFSv4.2, at a miniumu, should include the "bug fixes" for NFSv4.1. For example, this would include the pNFS access I-D and the layoutcommit issues that have been recently discussed. The other area of interest was potential development of a feature negotiation methodology to be included in the protocol so that each new feature did not have to build a method of their own. With respect to the requirements I-D that currently exists. There wasn't consensus about how best or if that I-D should be moved forward. More discussion on the working group alias is needed to best determine how best to build an effective NFSv4.2 todo list and move that work forward.