SMLT with LACP not working (SFP and RJ45)

[TomJ1]

Hello at all,

I try to establish a SMLT connection with LACP.
I do have two VSP-7400-48Y-8C VSPs and one Cisco UCS C240 server with 6 network interfaces (4 x 25GBit SFPs and 2x 10GBit RJ45 Ports).

On the server side I created a bond with one of the 25GBit ports and one 10GBit port (LACP 802.3ad).

On the VSPs I created a SMLT with LACP:
Connection 1 -> VSP1 (port 1/10) to server 25GBit port(SFP)
Connection 2 -> VSP2 (port 1/10) to server 10GBit port(RJ45)

But the SMLT will not work. I always got an error "GlobalRouter LACP INFO A churn condition has been detected for port 1/10 as the PARTNER is out of sync."

I got the error on both VSPs, depending which port do enable first.

In the VSP GUI inside the port/lacp tab the "ActorOperState" is "lacpActive,lacpShortTimeout,aggregation,synchronization" for one port an for the other one
"lacpActive,lacpShortTimeout,aggregation,synchronization,collecting".

I tried on two different servers and both are not working.

I also tried to setup an SMLT with LACP to two 25GBit ports - this is working fine.

What could be the problem?

Thank you so much.

[PeterK]

As Jave wrote,

for a Link-Aggregation the speeds of all ports need to be the same. If you use LACP or static doesn't care also if it's only a mlt or smlt. You must have same speed.

[RobertD1]

Hello,

Don't the speeds have to be the same (which you have proven with two 25G)?

Rob

[WillyHe]

Hello TomJ1,

The answer is already given, LACP checks for multiple parameters before adding ports to a link aggregation group, these parameters are full duplex, speed (must be equal), I think also vlan membership (if tagging is enabled) must be the same.
So combining 10G and 25G connections into one aggregation group is not supported.

When you use static link aggregation you can combine different speeds because there is no checking done, the only prerequisite to add a port to the configured link aggregation group is link-up, even when it is connected to another device.

From the load balancing point of view it is a VERY bad design to use different speeds in the same aggregation group.
The load balancing protocol used will distribute the load equally over all links, this means in your example, that e.g. a load of 30Gb distributed over two links is 15Gb/link oversubscribing the 10Gb link with 5Gb, on the 25Gb link you still have a margin of 10Gb.

Hope it helps
WillyHe

[TQU]

Hi, 

Not a recommanded design but it should work. 

Did you check the lacp smltsystemid ? > show lacp 

It must be the same on both VSP.

regards, 

Théo