Nexus 7k 5k VPc

[abdul basit]

Hi Every body,

Can anyone please tell me why Cisco use separated VPC keep alive link instead of sending the the keep alive updates on the peer links ? May be using some other frequency on the physical layer ?

Thanks and Regards,

Abdul

[Mark Malone]

Hi

its best practice for design to have the keepalive link separate, no data on it at all

In best practice doc this is in it link below

http://www.cisco.com/c/dam/en/us/td/docs/switches/datacenter/sw/design/vpc_design/vpc_best_practices_design_guide.pdf

 vPC peer keepalive link: The peer keepalive link monitors the vitality of a vPC peer switch. The peer keepalive link sends periodic keepalive messages between vPC peer devices. The vPC peer keepalive link can be a management interface or switched virtual interface (SVI). No data or synchronization traffic moves over the vPC peer keepalive link; the only traffic on this link is a message that indicates that the originating switch is operating and running vPC.

[abdul basit]

I agree with you but as to my knowledge, the Peer keep alives will not need a high bandwidth so we can possible carry them in a separate VLAN over the Peer links and that can be the easiest solution in term of module failure scenario case. The only case i can see it might be feasible to have separate link if some how traffic on the peer link congest due to some miss configured routing and it start dropping the keep alives but for that case QoS can be use.

Thanks and Best Regards

Abdul

[Steve Fuller · Accepted solution]

Hi Abdul,

The key point to what Mark posted above is "avoid fate sharing in case the vPC peer link goes down".

If you use the vPC peer link for peer-keepalives, and the peer link is lost i.e., a single failure of some kind, then how are the two vPC peers to know whether the other peer is alive? With one failure they have lost both mechanisms used to determine whether the peer is operational.

If peer-keepalive is carried on a separate link e.g., the management interface, then there has to be two simultaneous failures before we get to that position.

In effect vPC is a clustering technology and the worst scenario for any clustering technology is a "split brain". We need to avoid this at all costs.

Regards

[abdul basit]

Dear Steve,

Thank  you so much, it is quite clear now. I think over the split brain case and yes you are right if the peer link is down, without peer keep alive link, there is no way to stop that the split braining.

Thanks and Best Regards.

Abdul