switchs topology

[Nicolas967]

We have 3 bays all equipped with HP switches.

The first bay contains 4 switches 1820-48g (J9984A) simply connected to each other.

The second bay contains two 1910-24G (JE006A) switches that must be redundant (LACP)

And finally the third bay which contains 2 switches 1950-12XGT (JH295A) with IRF activated.

What is the best way to chain Bay 1 to Bay 2 and 2 to 3 without creating a loop ?

Thank you for your advice

Nicolas

[akg7]

Hello,

Is Bay 1 connected in a daisy chain way?

Do you want to connect all Bay1 switches to Bay 2 and Bay 2 to Bay 3?

Any possibility to enable spanning tree?

Thanks!

[Nicolas967]

Hi ,

yes all switches in bay 1 are connected in a daisy chain way.

And yes i want to connect bay 1 to bay 2 and bay 2 to 3.

This is partialy in place but my problem is to make the 2 switches of Bay 2 redundant for a vmware vsphere cluster.

I tried to activate LACP between these 2 switches but it generated a loop ...

 

[akg7]

Hello @Nicolas967,

How are you are activiating LACP from Bay2 switches to  vmware vsphere cluster?

Please share a diagram for that as well.

Thanks!

[parnassus]

Hi @Nicolas967, you're creating a Network Loop when Bay 2's switches connect to the IRF on Bay 3. In other terms the IRF (it's a single logical entity for both Bay 2 switches) close the chain formed by Bay 1 switches connected, on their respective ends, to Bay 2 switch 1 and Bay 2 switch 2.

What do you mean with the statement: "my problem is to make the 2 switches of Bay 2 redundant for a vmware vsphere cluster." ? How and where your VMware vSphere Cluster's nodes are connected exactly? Are you using VSS on VMware vCenter side?

[Nicolas967]

i have an IBM Blade Center with two networks cards, each network card is connected to one of the switch of bay 2

i'm looking for a solution in case of loss of one of the 2 switches of bay 2...

Today, one of the 2 switches of Bay 2 is also the gateway for some vlan if I lose it, I lose the road ...

[akg7]

Hello @ Nicolas967,

It is a design issue.
Is Bay2 switches are in IRF?
Is it possible to form IRF between these two Bay2 switches?

Thanks!

[Nicolas967]

IRF does not exist for this switches model...

[akg7]

Hello @Nicolas967,

I apologies, you are correct 1910-24G (JE006A) switches do not have IRF concept.

LACP concept is to bundle two or more ports from a switch or logical switches to get the more speed, redundancy and loop avoidance..

In this scenario, I believe LACP  with 'vmware vsphere cluster' will work if you will bundle 2 ports from one switch not from different switches.

Thanks!

[parnassus · Accepted solution]

There are two three main points here:

1. Network Topology: there is a Loop and it should be avoided (or, by means of proper STP priority setup on offending link - I refer to links from Bay 3 to Bay 2 - one could set that link so STP will block it preferring the other...but I don't believe that is the desired solution).
2. The way VMware permits Links Aggregations (if you use Virtual Standard Switch - VSS - you can at best use Static Link Aggregations without any control protocol, if instead you use Virtual Distributed Switch - VDS - you can opt for LACP Link Aggregations, also called Dynamic). BUT - and it's the third point - in both cases the terminating (and orginating) switch entity must be a single logical entity (IRF) or a single physical standalone switch (Bay 2 Switch 1 or Bay 2 Switch 2, as example).
3. Static/LACP Links Aggregations require them to originate and terminate on single logical entity (IRF) or a single physical standalone switch...so, considering this restriction, I don't see how - with the current setup/topology - your Blade Center server could be uplinked with Static or LACP Links Aggregations to Bay 2 on BOTH switches...it can...but only terminating on one switch per Link Aggregation. So you don't solve your initial problem of enhancing its connection resiliency within Bay 2.

A different scenario could be: IRF on Bay 2...and so you could connect a Link Aggregtion originating on your IBM Blade Center with member links distributed on both IRF Members....and you can use LACP or Static (better LACP).