Minutes taken by: Zartash Uzmi Forces: attendence (~20) Chair: Jamal Hadi Salim 1) Chair: Administrivia ======================= The chair went through the administrivia 2) Chair: WG Status =================== The chair went over the WG Status Chair noted the two outstanding drafts (LFBLib and CEHA) as items that need to be completed soon. Chair mentioned the Interop that happened in February. Based on the interop, RFC 6053 will be obsoleted and the results from the February interop will be merged pushing for a new RFC. 3) Kentaro Ogawa: February Interoperability report ================================================== Kentaro talked about the setup involving the different scenarios of the CE/FE builds of the three implementations. test scenarios: scenario 1: LFB operation - verification of RFC 5810 compliance scenario 2: TML with IPSec scenario 3: CE High Availability (HA) -- verification of CEHA mechanisms scenario 4: packet forwarding results from above test scenarios: scenario 1: success scenario 2: success in local conf but problem in distributed conf with Greece scenario 3: success, FE priority issue on messages incoming from multiple CEs scenario 4: success with Japan CE and China CE, minor problems on Japan FE (on OSPF not on ForCES) issues found: i) data encapsulation format ii) message handling prioritization in FE 4) Weiming wang: Update on ForCES LFB Library ============================================= should there be text on static routing ability? The view in the meeting is there should be. Weiming to come up with text. updates to the document since last time agreed on after discussions: 1) neighbor discovery (ND) for IPv6 doesn't need to be defined in the base LFB Jamal asked if there would be no IPv6 in this draft. Answer was only ND was not - IPV6 is on. 2) inclusion of IPv4PrefixLen and Mask, removal of MaxPort, removal of MTU in EtherMACIn LFP, replacing OperStatus by Adminstatus in EtherMacOut, etc. 3) Change names of some LFB ports Issues under discussion: - separate pieces for LPM and NH (LPM LFB and NH Aggregator) - previously discussed, asked for agreement? Jamal said he had no disagreement on the separation of LPM and NH. Genericity and Offload issues: narrated by Jamal on a slide Jamal: the document (in an effort to be generic) was difficult to read performing ARP as in the current architecture may lead to race conditions. Joel: it is preferable to keep the mac index table "media independent" Jamal: for consistency (with removal of ND): we will remove the ARP mechanism There was agreement to remove ARP. Weiming went over the next steps: updating the draft, more details in the used-case chapter, optimize description text in LFB lib XML file. 5) Evangelos Haleplidis: Intra NE HA draft =========================================== changes from 00 version to accomodate interop 1) FEPO model -- required changes in FEPO LFB 2a) CEHA init -- added initialization steps 2b) CEHA state Changes to CEHA flowchart? Additional comments: Jamal suggested those be discussed on the list CE HA Interop test: An interesting test on FE in Japan and backup CE from Greece There was bug in FE which added 1 sec delay in message handling while HBs was set to 0.5 sec. Result: FE couldn't successfully re-associate with a backup CE Solution: FE should check priority channel when associating with backup CE Joel: This solution does/should not belong to the stack (as it is an implementation issue that causes interop to break) Jamal suggested to move this discussion to the interop draft 6) An Implementation Demo by Evangelos ====================================== Evangelos demonstrated using virtual machines an active and backup CEs with one FE. He showed the active CE configuring an FE but rejecting configs from the backup CE. He then demonstrated the backup taking over the CE role and showed he could further configure the FE from the backup CE (which was previously rejected by the FE).