poor stack LAG Performance with M4300 24X24F

Hi ClemensD

Thank you for your message and sorry about your issue. M4300 series switches are proficient when it comes to NSF stacking and distributed LAG for both performance and redundancy. We got to understand why your file server runs so slow when its 2-port LAG goes to two switches across the stack, and so great when its 2-port LAG goes to the same switch. Across the stack, performance should be similar and redundancy / unstoppable operation should be the added benefit, should one switch fail.

At this point we have too litte information, some troubleshooting must be made. Either you want to continue with the Community, or you may open a case at Tech Support.

We need the Tech Support file out of your M4300-24X24F stack for getting system software version, detailed config and comprehensive logs. Can you send it to me via Private Message if you want to continue with the Community? Please log in the M4300-24X24F stack using CLI, either local console or remote telnet/ssh, go (M4300-24X24F) >enable and enter the command (M4300-24X24F) #show tech-support. When you'll see the message "TechSupport File created successfully", please log in the M4300-24X24F GUI (web interface) and go to Maintenance \ Export \ HTTP File Export and select "Tech Support" in dropdown File Type.

(there is no passwords in clear, they're encrypted in the configuration files).

Also, please confirm the topology:

- Are the three M4300-24X24F in a 1x10G full ring?

       24X24F(1) ______ 24X24F(2)______ 24X24F(3)

            |_______________________________|

- While the access switch is connected to the stack using distributed LAG across 24X24F units 1 and 3, how is the file server LAG distributed across the stack, units 1 and 2, 2 and 3, or 1 and 3? To limitate inter-switch traffic in the stack, it would make sense to use units 1 and 3 as well, for the file server incoming LAG. 

The Tech Support file will tell us, but you can monitor the "stack ports" utilization in the GUI going to System \ Stacking \ Advanced \ Stack-port Configuration. The Transmit data rate (Mbps) port by port usually helps to identify congestion when stack interconnect isn't sufficient. It depends on the traffic across the stack, hence the importance of calibrating the number of "stack ports" per switch.  M4300 10G models support up to 16 "stack ports" per switch. Error rate might also indicate a stacking issue that the Tech Support file will document.

Last but not the least, for correct traffic "parsing" across the LAG, it is important to have the same hashing algorithm in both LAGs (the file server LAG and the access switch LAG), this will be worth a check. For maximum compatibility, I would make sure the file server is configured the same (L2, L2/L3 or L3/L4 for source and destination etc.) and select the right hashing on the M4300 LAGs. 

Thank you and please let us know how you want to proceed.

Regards,