OPSAREA Open Meeting Area Directors: Dan Romascanu and Ron Bonica MONDAY, November 17, 2008 1520-1720 Afternoon Session II Notes takers: Balazs Lengyel and David Harrington Jabber Scribe: David Partain 1. Note well, agenda bashing, note takers, blue sheets - 5 min 2. Renumbering still works - Hannu Flinck for Brian Carpenter and Ran Atkinson - 20 min only 2.5 people read draft Hannes: gap analysis needs to be done Dan: what do the authors think should be next steps? authors are no clear about what they want to do Brian Carpenter: do people care? Margaret: Is renumbering a solved problem? No, authors agree. Document important. Dan R.: I agree, it is important. Transition is just starting. Margaret: problem is not Ipv6 specific. Ipv6 doesn?t solve this. Brian: is this opsarea work? Peter Koch: dnsop co-chair: needs work. Is the idea to split this up and distribute it to WGs. Or do we keep it in one? Users would like one, expertise spread out. Dan R. has operational considerations. RFC? For now it could be Opsarea document. Later it could be divided to other WG. DNSOPS offers a slot to present. 3. Application Performance Monitoring - Al Morton for Alan Clark - 20 min General approach PMol work item? Layer based proposed work item for PMOL (above tcp layer) was similar work ever done already? scott bradner: lot of products out there, waiting time This seems to be messy. IP, tcp seems to be confused port availability is a non-issue tcp performance is an issue too many interdependencies it would be difficult to separate the component layers Dan R.: echoing Scott. Web interaction, VOIP call are interesting cases. How much if at all is common? w3c or oasis already does some similar work. will try to find out PMOL might be closed, depends if it continues or not Hanes: Complicated. this is a lifelong activity. Is the autor willing to spend that time? Author might not understand how much work he needs to do. 4. WG Status reports: WG chairs, 3 min each adslmib: Dan R.: last mib document 200+ pages. Editor review happened. Will go to ietf last call. benchmarking: rusty: mpls forwarding draft closing. sip device benchmarking proposed, will be next item. convergence/protection switching times draft. Progressing. pmol: rusty: 2 drafts nearly completed. little bit behind schedule. Next steps considered. Ran D. What should pmol do now onwards? 2 chartered documents are nearly completed. Long lived group like ippm, but dealing with non-ippm issues? Or provide technical advice to other wgs? scott bradner: you want to keep pmol going. We could also close pmol and see if the existing works were useful. Dan. R.: yes that is an option capwap: 6 documents. Base prot, binding to ieee, dhcp options, security IESG approved 2 more working on dime: hannes: finishing diameter api, has comments, mobility extension, mobile IP related, diameter integrated (IETF last call), QS (ietf last call) revising diameter base protocol (wg last call again) lots of new comments rechartering last IETF: new items Dan. R: revision of base protocols includes IANA procedure changes, should be reviewed by broader community Hannes: diameter heavily used, lot of design help given dnsop: Peter: 1 rfc out, ext deals with changes to on-the-wire; dnsop deals with deployment. Some work on DNS-abuse vulnerability. Constantly in process of recharter, including DNS server control protocol. New work suggested on performance. Other areas include ipv4/ipv6 coexist, trust anchors, key and signature lifetimes, etc. dan Keminsky?s findings distracted us. Interacting with other WGs to understand DNS ops better. Ron B.: IESG queue is empty. send documents now. Who knows Kaminsky. 12 people? grow: 1 document ready for wg last call. One bis document BGP tagging. Other documents - BGP Monitoring protocol draft> LISP presentations on operational aspects. imss: dan: shut down ipcdn: dan: shutting down ipfix: browning lot of documents in editor queue xml descriptions are commented by IANA. 5 current draft after last call mediation work ongoing, started by operators, building tree of collectors manipulating the data as it is passed up the tree. Psamp: Juergen Quittek 3 documents finished psamp mib missing will close wg soon mboned: no news netconf bert: netconf/tls in AD review; partial locking in wglc; the monitoring draft is in a wait state, looking for experience in modeling to make sure we do not develop too many duplicate definitions of data types. New work on notifications content, and netconf-bis, and default capabilities netmod: David P: most active group, progressing well. Arch doc accepted as wg item; datatypes stable; YANG converging; initial DSDL draft in progress opsawg: 3 documents after wg last call maybe one new document comments needed dan r. support the importance of the document, guidelines for other WGs about OAM issues opsec: no news radext: no news v6ops: no news 5. Open microphone: whatever time is left Dave Harrington: boiler plate updates needed due to new protocols modeling languages. I will work on it. Dan R. today boiler plat is only a web page. DaveH: Netmod architecture needs this update as we will have not one but more SMIs. Margaret: Do we need a boilerplate in all documents? Wes Hardaker.: middle ground needed. no boiler plate is bad, too much is also bad. Just a an overview needed, a short one Scott Bradner: middle ground needed. +1 for Wes. Margaret: we need a reference to many thing. Why not a pointer to te RFC list. wes: need a pointer to the an overview document. People will not read all the rfcs People will just take the last document they have read. An overview and guidance is needed. Margaret: Doubts we will ever reach consensus on this document. Dan R. Vienna BOF is an example that guidelines are needed. Dave H.: We do have a guidelines document. Please review it. It speaks about how to manage protocols 2nd document lists all protocols and their recommended use and all MIBs Currently we say one mib with multiple protocols. This is changing. Dan R. Where should this be discussed? Opsarea is good for boilerplate. If we want a document opsawg is better. Dave H.: Opsawg has documents. They should have the responsibility for boilerplate as well. Netmod architecture needs the boilerplate, so netmod will start it and take it to opsawg. Margaret: we didn't use SNMP to do provisioning of access points because it wasn't right Consensus: opsawg will prepare proposal for boilerplate jefrey house: need a better common practice document about routing mib dan r. we do have this document rfc on snmp mib jefrey: question: how to put our data into the mib? snmp structure data needed, like snmp-ds. Ron.B.: YANG could be used. BGP mib in snmp but only basic. Put complex stuff in YANG. Jefrey: routing table is one complicated example. want one OAM model. Ron B. SNMP used only for monitoring. For complicated stuff was CLI. Why not move this part to YANG, exclude it from mib. jefrey: want a common routing table Ron.B. Yes that?s what YANG is for. How does the group like an SNMP/Yang divide Scott B: Why are we mandating a MIB Dan R.: We don?t mandate a mib Scott: routing area does mandate a mib. This is a basic question Ron B.: Some operators don?t have netconf. We need some basic monitoring Scott: we should just ask that the stuff is manageable, not ask for a MIB Dan.R: That is the question for new WGs: how can stuff be managed. We do not ask specifically for MIBs. Ron B.: Maybe we do need a MIB, but not for all data item. Jefrey: That?s our question as well. The answer is don?t use snmp. DaveH: I am involved in all this. Guidelines we do have a document. This is what you are asking for. It helps protocol selection. Rfc4181 thats about mib syntactical issues mostly. It is not about how to design a good mib. However snmp is not so important anymore Dont start yang usage as yet, it is not stable yet, not ready, limited field experience. Some already use it DNS server management, ipfix Document: If I don?t use snmp what should I use? cops-pr obsolete. netconf recommended yang missing yet. Sometimes own design needed. Today no good document. Guidelines document Simon Linen: bgp mib is good example, we can not monitor it with ietf mibs. bgp mib is from 1994. Old missing features. ipv6 etc. configuration data was there then thrown out bgp mib is used for many thinks which snmp is bad at. snmp is good for monitoring tl1, netconf probably no so scalable as snmp. snmp is good for some things, bad for others. This should be documented somewhere!!! scott: how to manage a routing protocol? monitor snmp good. configuring: snmp bad. Netconf is more promising. To do a protocol you have to have a mib. this is a too strict guideline jefrey: extensibility is crucial. bgp is not extensible?. Today bgp covers only basic. It is difficult to add the extras to the mib. You need a complete new mib to extend stuff. DavidP: Thats what YANG is good at. People should read YANG and comment. DaveH.: I agree with DavidP. YANG is more extensible. If you have part of the management in snmp part in YANG, how do you use them together? Difficult problems! Ron B.: We need a nuance guideline to protocol designers. We need that document. DavidP. This document underway, but check against reality. DavidH. Work is underway: Balazs: The guideline is in WG last call, but it needs to address new stuff still: e.g. how to address/name the same object in snmp and netconf?