Tony, Read this as well ( it talks about NOT using the mgmt0 for peer keep alives ) - we are trying this too
http://www.cisco.com/en/US/docs/switches/datacenter/nexus5000/sw/layer2/Cisco_Nexus_5000_Series_NX-OS__chapter8.html After figure 6, step 3 there is this text ; Note VLAN 900 must not be trunked across the vPC peer-link because it carries the vPC peer-keepalive messages. There must be an alternative path between switches NX-5000-1 and NX-5000-2 for the vPC peer-keepalive messages. The problem we are encountering is that if we drop the peer vlan from the 5k to 5k link then we get weird errors as well. I will STRONGLY suggest that you test any possible failure scenario that you can think of. Are you using the 5Ks/ FEXs in dual homed fashion ? I have an open case with Cisco on the use of Message: 7 > Date: Fri, 30 Apr 2010 10:45:52 -0500 > From: "Tony Varriale" <[email protected]> > To: "nsp-cisco" <[email protected]> > Subject: Re: [c-nsp] Nexus 5xxx VPC peer keepalives > Message-ID: <25315f3d0266408e937bfbd541923...@flamdt01> > Content-Type: text/plain; format=flowed; charset="iso-8859-1"; > reply-type=original > > ----- Original Message ----- > From: "Church, Charles" <[email protected]> > To: "nsp-cisco" <[email protected]> > Sent: Wednesday, April 28, 2010 12:35 PM > Subject: [c-nsp] Nexus 5xxx VPC peer keepalives > > > > Anyone, > > > > Coming up on a design issue with our upcoming first deployment of Nexus > > 5010s and 5020s in a new datacenter. It's recommended in the following > > doc to use the mgmt0 interface for peer keepalive messages: > > > > > http://www.cisco.com/en/US/docs/switches/datacenter/nexus5000/sw/layer2/Cisco_Nexus_5000_Series_NX-OS__chapter8.html#concept_47F7274E5FDA489884D0488BC491B066 > > > > We're doing a true out of band management approach on this new network, > so > > the mgmt0 interfaces all home back to an OOB switch/router (4507) which > > houses the NMS gear, etc. My concern is that a reload (or failure of > some > > type) on this OOB switch could cause a 'dual active' situation on all the > > Nexus pairs of devices . (6 pairs of 5010s, and the pair of 5020s that > > aggregate the 5010 pairs). I don't think I want that to happen. So the > > alternative seems to be a back to back non-VPC-peer link between the two > > devices using a VLAN interface, but I hate the idea of using a 10 gig > port > > just for keepalives. There are what appears to be additional copper mgmt > > ports on the boxes, but they're covered up, and not in the CLI. Any way > > to utilize those? Any other possibilities I'm overlooking? Or am I > stuck > > getting 1 gig copper SFPs and crossover cables for keepalives? > > > > Thanks, > > > > Chuck > > There are specific rules and actions that are taken when specific failures > occur and it's important to understand them. Also, if this is your first > go > at this, I would highly recommend testing the scenarios so you can get > comfortable. Part of the value in Nexus is the vPC. Unless you have a > very > specific reason that is broken by the vPC, get comfortable with it. The > dual active scenario isn't that bad. > > Oddly enough, I was testing a 350mbps multicast stream in a similiar > environment today. 2 5ks that were dual active - workstation on a single > attached 2k. The keepalive link was down between those 5ks and had > connection to network via peer link only. The 2nd 5k had only 1 uplink to > 7k-1. The source of the multicast was on a 2k hanging off 7k-2 and a > separate set of 5ks. > > All multicast packets were received and in order (reliable multicast > testing > with sequencing). > > tv > > > > ------------------------------ > > _______________________________________________ > cisco-nsp mailing list > [email protected] > https://puck.nether.net/mailman/listinfo/cisco-nsp > > End of cisco-nsp Digest, Vol 89, Issue 102 > ****************************************** > _______________________________________________ cisco-nsp mailing list [email protected] https://puck.nether.net/mailman/listinfo/cisco-nsp archive at http://puck.nether.net/pipermail/cisco-nsp/
