LAG between stack members with different port speeds

[Paul_Thornton]

Hi all,

Here's an annoying LAG problem. I have a stack which consists of an X670-G2 (slot 1) and an X460-G2 (slot 2). I want to create a 4x 1G LAG from this stack to another switch elsewhere. We're running 16.1.3.6 on this.

Unfortunately, this is the result:
* Slot-1 Switch1.17 # enable sharing 1:7 group 1:7,1:8,2:31,2:39 alg address-based L3_L4 lacp
Error: System cannot support Load Sharing among port media types with different maximum speed
Configuration failed on backup Node, command execution aborted!

Ah ha, I think, I need to fix the speeds. Fair enough. Can't have two ports at 10G and two at 1G.

Nope, doesn't work:

* Slot-1 Switch1.23 # config port 1:7, 1:8, 2:31, 2:39 auto on speed 1000 duplex full
* Slot-1 Switch1.24 # enable sharing 1:7 group 1:7,1:8, 2:31,2:39 alg address-based L3_L4 lacp
Error: System cannot support Load Sharing among port media types with different maximum speed
Configuration failed on backup Node, command execution aborted!

This seems to be very broken - obviously I can use just 4x ports on slot 2 (the X460-G2) but hey, the whole point of having the stack here is to help with some redundancy as well as port count.

I obviously know that members of a LAG group have to be the same speed, but I think that this should be allowed as the X670-G2 is a 1G/10G switch. Its almost as bad as not permitting a LAG between say a 10/100/1000M port and a 1G only port.

Is there any workaround for this? It feels like a bug or a feature that's being a bit too enthusiastic!

Paul.

[OscarK]

For that you could use redundant port, the lower speed port is redundant for the high speed lag.
https://gtacknowledge.extremenetworks.com/articles/How_To/How-to-configure-software-controlled-redun...

[Paul_Thornton]

This is still a bug in my opinion; as what both of us are trying to do is very sensible and fairly common. Especially if you have an X670 in your stack - you can't spread the LAG across stack members.

I can just about understand why it is implemented like this (to stop you having a LAG of 1x1G and 1x10G, which would probably have interesting real-world performance consequyences), but I can achieve exactly the same problem by having a LAG of 1x1G and 1x100M - without error - so all it really does is breaks functionality 😞

Perhaps having an option on the "enable sharing..." command to force the LAG creation, or treat it like the nagging EAPS config warnings that can (thankfully) be disabled would be a solution here.

Paul.

[M_Nees]

GTAC clarify for me that not the current speed of a port is diciding for being port of an LAG/Sharing it is the maximum media type speed.

In my current environment i have 2x 1GB GBICs - one in a SPF+ port and the other in a SFP port (Stack of X450-G2-10GE and X450-G2-GE). Bonding this both ports to one sharing port is currently NOT possible!

Still one question is open - is it possible to use both ports as a active / standby LAG ? Means one port stay as an active sharing port (port with higher prio), the other port stay in standby mode till the first fail (lower Priority).

Regards,
Matthias

[M_Nees]

Having the same issue !