Packet loss with LACP and Intel 82575EB network cards
I'm setting up M4300 switches (migrating from an externally managed network with cisco switches to a network we control) and I am experiencing packet loss when using servers with Intel 82575EB network cards. I've tested extensively and came to the following conclusion.
- Multiple different servers with Linux (Debian Buster, Proxmox on Debian Stretch, Windows Server 2016) are affected every time an Intel 82575EB network card is connected to the LAG.
- All these servers and operating systems work fine (without packet loss), when some other network card is used (e.g. Intel 82576, Intel X722, Intel I350, Broadcom BCM5709).
- When the LACP is set up over two different network cards (an Intel 82575EB and a some other mentioned above), then the packet loss is smaller (depending on which connections is used due to hashing).
- Hashing mode on Netgear was left untouched, which according to the manual should be load-balance mode 3, but in fact the default mode is 2 (visible in the web interface). When changing it to mode 3 it appears in the running-config (and thus does not seem to be default). When changing back to mode 2 it vanishes from the configuration. These load-balance mode tests were made with firmware 184.108.40.206.
- A firmware upgrade to 220.127.116.11 (and default load-balance mode) does not resolve the issue.
- I've tested multiple Intel 82575EB network cards, so it's not just one defective card.
- These same servers with the same configuration and network cards have worked on the cisco switches.
The configuration under Linux is always the same:
auto bond0 iface bond0 inet manual slaves enp2s0f0 enp3s0f0 up ip link set bond0 up down ip link set bond0 down bond-miimon 100 bond-mode 802.3ad bond-xmit-hash-policy layer2+3 bond-downdelay 200 bond-updelay 200 bond-lacp-rate fast
The Windows configuration is as follows:
teaming mode LACP load balancing mode dynamic
The interface and lag configuration on the switch are as follows:
(sw-test) #show running-config interface 1/0/1 !Current Configuration: ! interface 1/0/1 description 'server1-123-a' addport lag 1 exit (sw-test) #show running-config interface 2/0/1 !Current Configuration: ! interface 2/0/1 description 'server1-123-b' addport lag 1 exit (sw-test) #show running-config interface lag 1 !Current Configuration: ! interface lag 1 description 'server1-123' spanning-tree edgeport ip dhcp snooping trust switchport mode trunk switchport trunk allowed vlan 123 vlan acceptframe vlanonly exit
I haven't found any known problem with Netgear and Intel 82575EB cards on the internet, so I'm hoping someone here can point me into the right direction.
Kind regards and thanks