NETGEAR is aware of a growing number of phone and online scams. To learn how to stay safe click here.
Forum Discussion
LACP doesn't balance connections between links
Hello, I have some issues with LACP and a M4300-24X24F switch: - I set up LACP bonding between two linux/debian hosts. - I use two 10Gb optical links per machine. - linux hosts hash policy is lay...
Hi ASKRZYNIARZ
Thank you very much for confirming that changing port numbers allowed for correct parsing through the algorithm / LAG legs during your test.
The change you are suggesting is a great idea, but you need to know that hashing algorithm is in the hardware (underlying switch chipset) so it can't be changed without a new hardware.
I hope that this exploration will serve you and the rest of our great Community. Networking admins need to understand the issues and apply configuration to mitigate them. This is exactly what you have done, many thanks for that.
I think iPerf testing isn't a real-world behavior, and I believe that the Source/Destination IP and Source/Destination TCP/UDP algorithm (mode 6 in the LAG) is the most effective one for typical server installations. Having same IP sources and destinations alongside always even L4 source/destination ports is very unlikely - even with Linux in general. I believe you have spotted a typical iPerf behavior here.
Regards,
Thank you for your reply.
Could you please explain how the hashes are calculated? It would help define configurations that mitigate such issues.
NETGEAR Academy
Boost your skills with the Netgear Academy - Get trained, certified and stay ahead with the latest Netgear technology!
Join Us!
