Network File System Version 4 (nfsv4) ------------------------------------- Chair(s): Robert Thurlow Brian Pawlowski Document Editor: Spencer Shepler AGENDA: Wednesday March 20th: 13:00pm - 15:00 (2 hours) Welcome and Introduction (Pawlowski) 5 min Agenda bash - BLUE SHEETS - NOTE WELL Introduced Rob T, new co-chair, replacing Brent Callaghan. beepy agenda bashed - no changes. Schedule and milestones (Pawlowski) 5 min - Charter Presented changes to charter and new milestones. Mentioned in passing the ROI BoF that may affect us (we might be a customer for using their output) - slight discussion on that. Covered milestones, mentioned migration/replication is V4 targetted. Newest item - API advancement - GSSAPI needs to advance, need a way to evaluate. Repl-Mig part of work going forward, completion of work in the environment - Rob Thurlow will spearhead - checkpoint with AD in July. Will also deal with minor revisions. Need to work other specs also. Connectathon results (Shepler) 5 min - Linux NFS Version 4 posting - Kerberized V2/V3 fallout - Next testing event target July - Most issues resolved Shepler talked about Connectathon testing two weeks ago. Finished Mar 8. Linux, NetApp, Sun, EMC, Hummingbird, python test client, nfsv4shell there, lots accomplished. Nice RPCSEC_GSS testing w/v[234] - 4 platforms! Got RPCSEC_GSS Linux diffs together with a plan to get it into the next development kernel. SPKM/LIPKEY work in good shape for Linux at CITI; another implementer has come to light also, which is cool. No new issues there, focusing on details. Delegation progress good now, some testing done with Sun and Linux. Named attributes good progress also. Interesting addition to implementers was Peter Astrand's Python-based NFS Version 4 tester. See archives or http://www.cthon.org/talks02/index.html Fallout of requirement to security in NFS V4 has resulted in most implementers coming out with GSSAPI and Kerberos. The security is being back ported to NFS Version 2 and 3. Stuff will come out. UMich (see http://www.citi.umich.edu) has posted a snapshot of their Linux NFS Version 4 work. They have a rough SPKM3/LIPKEY under GSSAPI solution also - and a second group (Ron Hoffman) has emerged working on the SPKM3/LIPKEY implementation. Good news in moving RFCs forward (by providing to independent implementations). Closure of specification updates (Shepler) 25 min No new issues emerged from Connectathon. Some clarifications, but no changes. Delegation testing and recovery was played with at Connectathon. Minor changes to help migration/replication. Through implementation issues of named attirbutes. One week testing proposed for July. Spencer expects one or two new groups with new NFS Version 4 implementations to participate. Spencer is targetting week of March 25 for submit of new draft. Working group last call to happen in April. Detailed issues list and resolution covering changes to RFC 3010 to simplify review. Issues are filehandle types, stateids, OPEN upgrade/downgrade. Anything controversial? Violation of protection semantics? Specification wording to explain what server allows (to restrict behaviour). Interpretation of FH types related to repl-mig, delegation ops need a change to include FH as well as stateid, identification of server's file locking type, special stateids. OPEN upgrade/downgrade path probably biggest now. Working draft and issues list at http://www.nfsv4.org/rfc3010updates. Migration/replication next steps (Thurlow) 25 min Rob Thurlow (new co-chair) covered progress on migration/replication. See slides. Con call alerts will be posted to the alias. beepy to talk to rserpool (Maureen) on coordination and notice of con calls. Rob summarized the requirements generated by the WG and off line work. To be transformed into an internet draft for AD review. Described thoughts on protocol. Restartability thought important (vs. journaling). Mark Wittle asked about four areas of focus on the migration/replication. beepy clarified a question that no admin changes expected if someone decides not to deploy migration/replication in their environment. Rob responded that NFS V4 core protocol is probably not sufficient to support the back end protocol. The question is: what is the back end protocol? Requirements to drive understanding whether any existing protocol can serve. Have any bulk data movement options been ruled out? Sort of; proposal will go for new protocol with a particular structure. FTP not there, NFSv4 not quite enough. Control protocol seems like it would be NDMP - will you work with them? Yes, when the NDMP folks are a working group, we can talk to them. Review of work items (Shepler/Pawlowski) 20 min - API GSSAPI advancement - MIB draft - RPC/XDR/RPCSEC_GSS RFC plans - ACL API advancement drafts - Scott Bradner described the rationale of what two API implementations. Related to MIB advancement. Scott further clarified that some may or may not go standards track in the IETF. http://search.ietf.org/internet-drafts/draft-pawlowski-apitest-00.txt Shepler talked to RPC advancement - Sun owns. API advancement draft, in process, working with AD. Had a MIB awhile ago, more than the six month timeout, we will refresh. Other RFCs: XDR a draft standard now. RPC: still need to recast registration to IANA and update RFC info on this, but this is not too bad. SPKM/LIPKEY - we will help, but not own. Might see: ACL mapping work, NFSv4 global namespace. Highest priority has been resubmission of RFC 3010. Open discussion (Pawlowski) 30 min Early end. Wrapup (Pawlowski) 5 min Priority is RFC 3010 resubmission with updates.