[HP 5900] does IRF create L2 loops ?

[fab2]

Hi all,

I don't know very much IRF (I'm pretty new with HPE), in the first place I thought IRF was making a layer 2 loop inside IRF.

I mean, in the following senario for instance, I thought I should NOT configure my 2 5900 Switchs as IRF Cluster (see snapshot).:

Or do I have to use STP if I have a make a IRF Cluster, in such senario ?

In that case I need to make STP configure on WDM Switches as well, and I do not think it is possible (customer's Switches, I don't handle them).

Thanks for your help.

Fab

[parnassus · Accepted solution]

Speaking only about IRF (I ignore what is a WDM switch): I think the concept of IRF Stack that creates a Layer 2 loop inside itself is...somewhat...strange...why? because IRF technology is used to extend, through the usage of virtual switching, the switching capabilities through all of its IRF Members switches...so how a L2 loop between those members can be generated if the whole IRF Stack is properly formed and is acting as a logical virtual switch?

I'm not able to find the 4AA5-3379ENW (title: "HP Intelligent Resilient Fabric (IRF) Frequently Asked Questions" of April, 2015) document anymore ot the URL from which I downloaded it just few months ago (here). That FAQs was very helpful to familiarize with IRF (basically you can still see it here on a non HPE site).

Different is the case if we are questioning about an "external" loop that generates between non-IRF member switches and, as a side effect, involves the whole IRF Stack (let's say you have an IRF Stack - made of Switch A+B -, that IRF Stack is connected, respectively and separately, to both non IRF Member switches Switch C and Switch D, at this point you are OK...STP/MSTP a part...once you plain connect Switch C and D together you start having a loop but that loop originates externally to the IRF Stack if no STP/MSTP mechanism are involved to manage/prevent this looping).

[fab2]

Thanks Parnassus WDM switch is a (tranport network) layer 2 switch, so we are exactly in the senario A-B-C-D you described. IRF is clear to me at least. So in such case IRF without STP is definitively to be avoided. I'll go for vrrp I think.

[Mike79]

---

@fab2 wrote:
Thanks Parnassus WDM switch is a (tranport network) layer 2 switch, so we are exactly in the senario A-B-C-D you described. IRF is clear to me at least. So in such case IRF without STP is definitively to be avoided. I'll go for vrrp I think.

---

Hello,

Please also keep in mind you need still STP/MSTP when decide to use VRRP for your switch cluster. This is because C+D switches are L2.

I would recommend keep VRRP + having LACP between switches A+B for redundancy. IRF is coming complicated when failover occurs and you have cluster split (fiber links between switches broken).

Michal