topic EAPS between MLAG core switches? in ExtremeSwitching (EXOS/Switch Engine)

EAPS between MLAG core switches?

BigRic — Wed, 12 May 2021 19:07:45 GMT

Good morning. We have an older two-tier core/edge network where the core is a stack of SummitX series switches. There are two, 2-port LACP’s that act as ring in/out ports in an EAPS ring and the core stack is the master. We’re replacing the core with two new switches in an MLAG pair. Am I correct that you can’t mix EAPS and MLAG, making the two core’s essentially run as a new highly-available EAPS master? Is it easier to just break the ring and move to a hub and spoke with LACP to each downstream spoke (we’re going to lose the fiber that creates the current 3-node ring we have, breaking the ring path between the two downstream nodes). I’m wondering if I need to deal with that now, as opposed to waiting for that link to actually be removed.

Re: EAPS between MLAG core switches?

CThompsonEXOS — Thu, 13 May 2021 03:04:07 GMT

Hi Ric,

I think the main concern with EAPS(also STP) and MLAG is the concern over EAPS(or STP) blocking the ISC link:

MLAG Limitations and Requirements
The MLAG feature has the following limitations:
• MLAG peer switches must be of the same platform family. The following MLAG peers are allowed:
◦ ExtremeSwitching switches with ExtremeSwitching switches
◦ SummitStack with SummitStack
Note
In the case of ExtremeSwitching standalone switches, it is strongly recommended that
MLAG peer switches be of the same type.
In the case of SummitStack switches, we recommend that the MLAG ports be from slots of
similar capability.
• Layer 2 protocols such as EAPS or STP (Spanning Tree Protocol) will be configured to not allow the
blocking of the ISC.

https://extremeportal.force.com/ExtrArticleDetail?an=000084520

Although it sounds like MLAG without EAPS would still solve you problems moving forward.

Thanks,

Chris Thompson

Re: EAPS between MLAG core switches?

BigRic — Thu, 13 May 2021 03:41:22 GMT

Thanks @Christopher Thompson. The new core is made up of two stacks of 620s with a 40GB ISC between the stacks (it’s not a huge environment). Since there are only three nodes in the EAPS ring, in truth, a hub and spoke design with LACP across core peers is probably sufficient for us moving ahead. I don’t think EAPS supports any type of mechanism for a split-master running across cores - that’s what I was getting at. We may just move from EAPS to a backup path between the two slave stacks and STP blocking that path during normal operation, and as you said, be sure to exclude the ISC from any STP participation.