cancel
Showing results for 
Search instead for 
Did you mean: 

EXOS MLAG and bonded S8s

EXOS MLAG and bonded S8s

Jake_Koch
Extreme Employee
Hello Extreme Hub,

I have a questions regarding EXOS MLAG and bonded S8s

1st here is the scenario:

I have a customer with bonded S8s that function essentially as a collapsed core (Distribution/Core and some direct server connections)
S8-Alpha is building 'A' and S8-Bravo is in 'Building B'
There is a Data Center in Building A and a Data Center in Building B, each Data Center server is single homed to its buildings S8. Also each buildings IDFs are also single homed to its local buildings S8.

The customer is going to be increasing their fiber infrastructure between building A and building B with the goal of dual homing the Data Center servers and some of the IDFs.
The customer also plans to add some EXOS x690s to move the Data Center servers off of the S8s. So all of the servers would connect to local MLAG'd x690s, and the x690s would then uplink to each S8 (local and remote) using the new fiber infrastructure.

My question is, are there any issues connected MLAG'd x690s to bonded S8's with regards to STP or any other loop prevention protocols? (XLRP)
Am i correct in saying, that from the perspective of the x690s, the ports facing the servers need to be set up as MLAG and the ports facing the bonded S8s need to be set up as MLAG ? (otherwise, the ISC between the x690s would not block looped traffic from the S8s ?)

Should each x690s connect to more than one blade per S8 chassis ?

The x690s would be set up as L2, VRRP and routing would still be handled by the S8s.

I am not familiar with S8 bonding and I am newer to EXOS... I just want to make sure I am not creating a problem with this design.

Thanks

Jake
11 REPLIES 11

I'm not going to recommend it, but 22.5 and 22.6 have some features that allow RSTP and MSTP respectively over MLAG.

thanks Erik, great info. I believe I am good to go

Jake_Koch
Extreme Employee
Rob, great reference articles and you walked thru your explanation very well and answered all my questions. I feel more comfortable with what we are looking to do with this customer. Eventually, we are going to start phasing in Fabric Connect, but that is a different topic all together 

Thank you

Largetrousers
New Contributor
This might help:

https://gtacknowledge.extremenetworks.com/articles/How_To/Design-Considerations-when-using-EXOS-MLAG

If the MLAG'd X690s are connected to the VSB'd S8's i.e. a square layout with each device in the corner and the lines of the square representing the interlinks, you'd effectively setup LACP from the S8's to the X690s, from the X690s those connecting ports would be configured in the same MLAG group (so effectively LACP back) using port sharing - just make sure it's using the appropriate algorithm to match that of the S8. My take would be within this S8 to X690 square setup you would avoid L2 loop protection to keep these links always forwarding, but then anything say on an edge/access side of things going to edge devices (e.g. users, servers) use either ELRP on purple switches copper ports (not uplinks) and use adminedge/spanguard on any Enterasys edge switch copper ports (again not the uplinks). If Enterasys edge is going to S8 consider using some additional elements like loop protect.

If you chose to try the STP route between the environments, obviously test offline if possible - would be interesting to see how this behaves when failures are simulated.

Again if the devices on the X690s are unique subnets I would strongly consider the use of L3 separation and implementing OSPF to hook them up to the S8s, keeps it nice and separated and saves some of the pain you are now trying to work through.

Jake_Koch
Extreme Employee
Thanks for the responses guys.

Rob, to your point I believe STP would need to be disabled between the S8s and the x690s and only enabled on the x690s 'down' links.

Andre, the S8 in Alpha is bonded to the S8 in bravo.

GTM-P2G8KFN