cancel
Showing results for 
Search instead for 
Did you mean: 

Link Aggregation negotiate in half-duplex

Link Aggregation negotiate in half-duplex

Formia
New Contributor
I configured two port in aggregation between a X440G2-24t-G4 and a Moxa EDR 810 both in static mode.
Everything seems working fine but the two ports of the Extreme switch negotiate in half-duplex (with show port information). However Moxa shows his two ports in full-duplex.

The command I used is: enable sharing 21 grouping 21-22

Is there any reason why this can happen?
Thanks
Andrea
1 ACCEPTED SOLUTION

EtherMAN
Contributor III
when you set up the aggregation on the EXOS side you can choose what algorithm to use for the share. This must be defined when you create the share. Default is layer 2 so looks at mac addresses only ... Best is layer 3 and 4 so it looks deeper into packets for the sharing hash. But ports share will never be equal and if you only have one client on one side and one on other side transferring file or streaming ect all flows will only be on one port not matter what. This is why we are moving from ten gig lag groups in our metro core to 100 gig links... even with 2500 active vlans and about 89000 macs in core you can still get enough traffic from some of the larger single ip routers to load up a single ten gig port in a 40 or 60 gig lag group.

View solution in original post

6 REPLIES 6

Formia
New Contributor
Thank you very much for the explanation. Actually I tested only with one client connected.

EtherMAN
Contributor III
when you set up the aggregation on the EXOS side you can choose what algorithm to use for the share. This must be defined when you create the share. Default is layer 2 so looks at mac addresses only ... Best is layer 3 and 4 so it looks deeper into packets for the sharing hash. But ports share will never be equal and if you only have one client on one side and one on other side transferring file or streaming ect all flows will only be on one port not matter what. This is why we are moving from ten gig lag groups in our metro core to 100 gig links... even with 2500 active vlans and about 89000 macs in core you can still get enough traffic from some of the larger single ip routers to load up a single ten gig port in a 40 or 60 gig lag group.

Formia
New Contributor
Yes it is in the section of link aggregation (or port trunking) but it is only possible in static mode (LACP is not present). So I have configured ports sharing in static mode also on the Extreme switch.
Anyway after manually force both interfaces in both MOXA and Extreme to 100 Full Duplex (the maximum speed in MOXA), all links are established in 100Mb/s Full duplex. Probably the auto negotiation between these two model/brand is not working fine with the aggregation.

However after solving that I have realized that the traffic load is always in one link so the bandwidth is still 100Mb/s (not 200 Mb/s as I was expecting). I tested even between two extreme switch with the same configuration of the ports (100 Full).
Is this normal in the Extreme link aggregation?

Appreciate your support

EtherMAN
Contributor III
Does the MOXA support a lag group? I just looked at their manual and did not see a way to set up a port lag group. Both side have to be able to do lag for this to work.. This is more than likely the issue
GTM-P2G8KFN