HPE FlexFabric 5700 LACP MAD to Cisco Switch

[Eskimoe]

I've recently done some labbing with two HPE FlexFabric 5700 switches and activated IRF.

Everything is fine and the IRF works great, the question is how I should setup MAD when connecting the IRF switches to two other switches, Cisco catalyst 3750 in a stack.

Since MAD is required according to HPE's guide, I should configure it, right?

Attached is a simplified picture of the topology.

What is the best solution for this?

Thanks!

[parnassus · Accepted solution]

Probably the simplest way of adding a MAD mechanism on a IRF setup of only two IRF Members (that will not expand horizontally with other IRF Member in future) is to use the BFD MAD mechanism deployed using common Ethernet ports (one dedicated Ethernet port per IRF Member is sufficient, no intermediate device is necessary for the BFD MAD deployment): HPE FlexFabric 5700 Switch Series IRF Configuration Guide explains various supported MAD mechanisms, among which, there is BFD MAD.

[Eskimoe]

Thank you very much for the reply Parnassus!

Just a few more questions, I am sorry in advance if they are off topic.

What would happen if I had 3 or more units/in case it would expand in a couple of years?
Is BFD MAD not a suitable solution then?

The configuration guide explains a scenario where you have 4 switches in an IRF and connect one port from each to an intermediate device. At the same time it explains BFD MAD as a mechanism where you don't need an intermediate device.

It also mentions this:
"If no intermediate device is used, any two IRF members must have a BFD MAD link to each other."

Does this mean that you cannot have 3 units in an IRF without using an intermediate device with BFD MAD?
And if you would use 4 units, connecting them two by two is the best practice for using BFD MAD as I understand it, in case you don't want an intermediate device.
But how would the switches then communicate using MAD, if there is no MAD link between all the switches in the IRF, or am I missunderstanding this?

Thanks!

[parnassus]

The point is that especially the BFD MAD Mechanism is one of the simpler MAD Mechanism you can deploy but requires that each IRF Member is connected with a MAD dedicated link (on common Ethernet port) to any other IRF Member...so...this means that the case of just 2 IRF Members simply requires the minimum number of ports (a grandtotal of two, one per IRF Member) and link (just one which interconnect the IRF Members pair)...if you add IRF Members those numbers go high in geometric progression since you should mesh any with any (3 Members IRF cluster = 6 ports, 2 per each IRF Member, 3 links | 4 Members IRF cluster = 12 ports, 3 per each IRF Member, 6 links)...and, at that point, it's better to shift to a BFD MAD Mechanism with an Intermediate Switch deployment or move to LACP MAD using an intermediate Switch (physical or logical) supporting LACP MAD.

There is a very good explanation on the HPE FlexFabric 5700 Switch Series IRF Configuration Guide (Edition: 19/04/2017 Part Number 5998-5586s)...from Page 9 there are examples of BFD MAD Mechanism implemented with/without an intermendiate device.