You can use either LACP or balance round-robin. Of course it is very important to know if your clients will be behind a single IP (for example behind router).
As most of the bonding hash policy depend on either MAC (L2) or IP + Port (L3) , you will have a hard time to load balance to clients with the same IP.
In my case, I have a team device with 6 x 1GbE ports.
As my setup is hyperconverged (replica 3 arbiter 1) , I had to use multiple gluster volumes (and each brick has different port) with 1 VM disk per volume and to stripe on VM level for better performance.
В сряда, 24 юли 2019 г., 11:02:48 ч. Гринуич+3, Edoardo Mazza <firstname.lastname@example.org> написа: