roundrobin mode traffic reduction

[mzky]

Is it normal or abnormal for the network port team roundrobin mode to bind ports that are not connected to the network cable to cause a decrease in traffic? （at libteam1.31~1.32）

for example（use iperf3）：
eth0 eth1 are connected to the network, while eth2 eth3 eth4 eth5 are not connected to the network

1. Bind Ethernet ports eth0 eth1 to the polling team
   Normal traffic
   

2. Bind Ethernet ports eth0 eth1 eth2 to the polling team
   Total traffic decreased by 20%
   

3. Bind Ethernet ports eth0 eth1 eth2 eth3 to the polling team
   Total traffic decreased by 50%
   

4. Bind Ethernet ports eth0 eth1 eth2 eth3 eth4 to the polling team
   Total traffic decreased by 30%
   

5. Bind Ethernet all ports to the polling team
   The total traffic has decreased by 50%, with the same effect as the 3 item

Thanks.

[fuanan-3]

1.There are five network ports in total,two network ports are up and three network ports are down,you only need to pay attention to the traffic of the two network ports in the up state,isn't it?
2.enp125s0f4 and enp125s0f5 is the name of which two of the five network ports?

[mzky]

1.There are five network ports in total,two network ports are up and three network ports are down,you only need to pay attention to the traffic of the two network ports in the up state,isn't it? 2.enp125s0f4 and enp125s0f5 is the name of which two of the five network ports?
@fuanan-3
There are six nets in total.
There are two consistently connected network ports (any two).
If one or more down network ports exist in the team, then the total traffic to the two continuously connected network ports is reduced.
enp125s0f4 and enp125s0f5 are any two continuously connected network ports.

[fuanan-3]

I think you should test the traffic of the two network ports in the up state,not the enp125s0f4 and enp125s0f5 all the time in the screenshot.

[mzky]

我认为您应该测试处于 up 状态的两个网口的流量，而不是屏幕截图中的 enp125s0f4 和 enp125s0f5。

If in actual application, four up network ports are used to bind team, but one of the network ports fails due to abnormal force majeure, the traffic loss is 1.5 times that of one network port. If the failure is two network ports, only one network port left of the two normal network ports is available.
If everyone thinks this is normal, then my question can be closed.

[fuanan-3]

The total traffic = the sum of the traffic of the up network ports.
Do you agree?
If only one available network port is in the up state and other ports are in the down state,whether one network port bears all traffic.

[mzky]

The total traffic = the sum of the traffic of the up network ports. Do you agree? If only one available network port is in the up state and other ports are in the down state,whether one network port bears all traffic.

Maybe my description is not clear enough, refer to the original my screenshots and instructions:

team0 = 2up = 1 network port 925 M/s + 1 network port 925 M/s = the total traffic of two network ports and about 1850M / s (this is a completely problem-free option)

team0 = 2up + 1down = 2up = 1 port 940 M/s + 1 port 580 M/s = total traffic about 1420M / s

team0 = 2up + 2down = 2up = 1 port 930 M/s + 1 port 0 M/s = total traffic about 930M / s

team0 = 2up + 3down = 2up = 1 network port 940 M/s + 1 network port 290 M/s = total traffic about 1230M / s

team0 = 2up + 4down = 2up = 1 port 930 M/s + 1 port 0 M/s = total traffic about 930M / s

2up=2 continuously connected network ports, but the number of bound network ports determines that the sum of traffic is not the same.

I think that under normal circumstances, no matter how many down ports are bound to team0, the traffic of the two up ports that are continuously connected should not be affected. However, the actual total traffic will change as the number of down ports added to team0 increases.

--- Forgive my poor English, this is from Google Translate.