Hi

 

i have a weird problem.

 

i have 2 nic in my unraid server, a normal 1gbps (onboard) and a 10 gbe nic (Intel Corporation 82599ES 10-Gigabit SFI/SFP+)

 

if i try an iperf from my computer to my server (iperf3 -s and iperf -c)

i got only 1gbps

 

i disable docker and vm services to go in the network config, i change mtu to 4088, and try an iperf and now i get my speed

 

[  4]   6.01-7.00   sec   126 MBytes  1.07 Gbits/sec
[  6]   6.01-7.00   sec   126 MBytes  1.07 Gbits/sec
[  8]   6.01-7.00   sec   126 MBytes  1.07 Gbits/sec
[ 10]   6.01-7.00   sec   126 MBytes  1.07 Gbits/sec
[ 12]   6.01-7.00   sec   126 MBytes  1.07 Gbits/sec
[ 14]   6.01-7.00   sec   126 MBytes  1.07 Gbits/sec
[SUM]   6.01-7.00   sec   758 MBytes  6.39 Gbits/sec

 

and now, i start my vm services, and i still get my bandwidth

 

if i try to start docker services, everything slowdown a 1gbps.

 

[  4]   9.00-9.73   sec  14.2 MBytes   165 Mbits/sec
[  6]   9.00-9.73   sec  11.2 MBytes   130 Mbits/sec
[  8]   9.00-9.73   sec  11.4 MBytes   132 Mbits/sec
[ 10]   9.00-9.73   sec  16.9 MBytes   195 Mbits/sec
[ 12]   9.00-9.73   sec  12.2 MBytes   142 Mbits/sec
[ 14]   9.00-9.73   sec  15.6 MBytes   181 Mbits/sec
[SUM]   9.00-9.73   sec  81.6 MBytes   945 Mbits/sec

 

 

 

if i shutdown docker service again and reset the mtu and not restart docker services, everything are fine, i got my speed back.

 

i dont know if this is a bug or an configuration problem.

 

thanks

fenrir-diagnostics-20221006-1027.zip

aahhh, from your diagnostics i see that you do more or less russian roulette.

You have set your 1G card to use dhcp, your 10G card to use a static address. 

So far, so good, but they are both connected to the same lan and get (dhcp) an address from the same subnet!

So what you have is a host with 2 seperate addresses within the same net.

 

This is not a general error, but it grants the clients to pick one of the cards randomly. So one may favour the 1G the other one may use the 10G for a while (and switch back somewhat later).

 

I cannot think that this is the behaviour that you have in mind?

 

Pull out the cable of the 1G and use the 10G only! (configure UNRAID to switch the cards and then deactivate eth1).

 

Theroreitcally you could put both cards into a bond, but thats stupid too. Again it is not predictable then which one is used, so you may realize the same strange behaviour as you have now. (never put members with different speeds into a bond (unless it is a backup only bond))