Hi all, With vPC on both pairs, all 4 links between the pairs can be aggregated into a single port-channel like the one pictured below. Practically you're going to setup two different vPC domains (vpc domain 1, vpc domain 2 for example) on the two pairs (11,12) and (13,14). With a port-channel configuration like the following example:
Assuming e1/1-2 are the downstream links (between pairs) and e1/10-11 are the links between vPC peer switches *# Setup vPC* sw11# vpc domain 11 ; peer-keepalive destination <mgmt-ip sw12> ; int e1/10-11 ; channel-group 1 mode active ; int po1 ; vpc peer-link ; peer-switch ; peer-gateway sw12# vcp domain 11 ; peer-keepalive destination <mgmt-ip sw11> ; int e1/10-11 ; channel-group 1 mode active ; int po1 ; vpc peer-link ; peer-switch ; peer-gateway sw13# vpc domain 13 ; peer-keepalive destination <mgmt-ip sw14> ; int e1/10-11 ; channel-group 1 mode active ; int po1 ; vpc peer-link ; peer-switch ; peer-gateway sw14# vpc domain 13 ; peer-keepalive destination <mgmt-ip sw13> ; int e1/10-11 ; channel-group 1 mode active ; int po1 ; vpc peer-link ; peer-switch ; peer-gateway ! Peer-switch will allow both vPC peers to propagate the same STP BPDU information by sharing a vMAC ! Peer-gateway allows the partner to route on behalf of the other, this gets around some non-RFC behavior seen in certain NAS vendors products when using FHRP ! It's perfectly fine to use mgmt0 interfaces for the peer-keepalive, it's a stream of UDP packets to src/dst port 3200. This link can go down with no ill effect on vPC ! Note conventions for numbering of port-channels and domains is just for example, use whatever numbering system works for you *# Setup the port-channels* sw11# int e1/1-2 ; channel-group 1314 mode active ; int po1314 ; vpc sw12# int e1/1-2 ; channel-group 1314 mode active ; int po1314 ; vpc sw13# int e1/1-2 ; channel-group 1112 mode active ; int po1112 ; vpc sw14# int e1/1-2 ; channel-group 1112 mode active ; int po1112 ; vpc [image: vpc.png] Regards, Jason. On Mon, 26 Nov 2018 at 11:10, Ahad Aboss <[email protected]> wrote: > Radeck, > > To avoid blocked ports, you will need to configure your upstream links as > follows; > > Built a port-channel from SW13 to SW11/12. > Build a port-channel from SW14 to SW11/12. > > Physically, SW 11 and 12 looks as 2 x Switches but logically, SW 13 and 14 > treats SW11/12 as a single switch. > > I suggest using at least 2 x 10GE ports between SW13 & 14 as SFP's do fail > from time to time. > You can use MGM for peer link keep alive or a dedicated link. > > You can increase your port-channel capacity as your traffic grows. > > Ahad > > On Mon, Nov 26, 2018 at 10:16 AM Radek Tkaczyk <[email protected]> > wrote: > >> Hi Guys, >> >> >> >> So from the feedback so far is that we should also link SW-13 and SW-14 >> directly – updated as below. >> >> >> >> The primary purpose for this design is to ensure redundancy across >> switches, and to be able to provide approx. 50 x 10Gbps ports, and 50 x 1 >> Gbps ports – this also leaves heaps of room for growth as well. >> >> >> >> Still looking for someone to go over the config with, so if you are >> interested (paid gig) please let me know. >> >> >> >> >> >> >> >> Thanks >> >> >> >> Radek >> >> >> >> *From:* AusNOG [mailto:[email protected]] *On Behalf Of *Radek >> Tkaczyk >> *Sent:* Sunday, 25 November 2018 4:34 PM >> *To:* Jacob Taylor <[email protected]> >> *Cc:* <[email protected]> <[email protected]> >> *Subject:* Re: [AusNOG] Assistance with Cisco vPC configuration on 4 x >> Cisco Nexus 3000 switches >> >> >> >> Hi Jake, >> >> >> >> That’s something that I wanted to check if it was needed/recommended. >> >> >> >> Can certainly put it in if it will help achieve better performance and >> redundancy >> >> Thanks >> >> >> >> Radek >> >> >> On 25 Nov 2018, at 4:26 pm, Jacob Taylor <[email protected]> wrote: >> >> Hi Radek, >> >> >> >> Not personally familiar with vPC, more so Arista MLAG and Juniper MC-AE. >> >> >> >> In the diagram there isn’t a peer link between 13 & 14 - is that a >> mistake in the diagram or the actual design? >> >> >> >> If you intend to build 2x20G bonds to two standalone nexus switches, >> that’ll work fine. >> >> >> >> If you are trying to achieve a 40G bowtie between the two pairs I’m >> fairly certain that won’t work (unless Cisco has some special black magic >> to transport signalling/MAC synchronisation over the bond itself). >> >> >> >> Cheers, >> >> Jake >> >> >> >> >> >> >> On 25 Nov 2018, at 15:04, Radek Tkaczyk <[email protected]> wrote: >> >> Hi Guys, >> >> >> >> I have a need to configure vPC on 4 x Cisco Nexus 3000 switches at one of >> our data centres – a design that we will replicate to other data centres as >> well. >> >> >> >> I think I have the config down pat, but I’d like someone with another >> pair of eyes to go over it with me to ensure it’s 100% correct. >> >> >> >> Is there anyone who can give me a hand here with this configuration, >> happy to pay for someone’s time to go over it to ensure we are doing this >> correctly. >> >> >> >> The Physical setup that I’m looking for: >> >> >> >> >> >> <image001.png> >> >> >> >> Thanks >> >> >> >> Radek >> >> _______________________________________________ >> AusNOG mailing list >> [email protected] >> http://lists.ausnog.net/mailman/listinfo/ausnog >> >> _______________________________________________ >> AusNOG mailing list >> [email protected] >> http://lists.ausnog.net/mailman/listinfo/ausnog >> > > _______________________________________________ > AusNOG mailing list > [email protected] > http://lists.ausnog.net/mailman/listinfo/ausnog >
_______________________________________________ AusNOG mailing list [email protected] http://lists.ausnog.net/mailman/listinfo/ausnog
