Having issues with X710 card and ethernet bonding in Linux

[RHelf]

Hi all,

We've just gotten in a bunch of new Dell R840 servers. We run Oracle Linux 6.10, and we have the Intel X710 card.

We have multiple servers that have NICs that seem to be flapping up/down. From dmesg:

[ 3343.096396] bond0: link status definitely down for interface eth2, disabling it [ 3346.090416] bond0: link status definitely up for interface eth2 [ 3350.082850] bond0: link status definitely down for interface eth2, disabling it [ 3358.067763] bond0: link status definitely up for interface eth2

The same info shows up in syslog. We note in /proc/net/bonding/bond0, that "MII Status" seems to flap between up and down, though the speed and duplex always show correct values.

We have tried:

• Upgrading firmware/drivers using the latest Dell image
• Upgrading to the latest UEK4 kernel on the OS
• Tinkering with NIC bonding settings. We run a pretty standard "active/backup" bonding mode 1.

Driver info:

# ethtool -i eth2 driver: i40e version: 2.1.14-k firmware-version: 6.80 0x80003d74 18.8.9 bus-info: 0000:17:00.2 supports-statistics: yes supports-test: yes supports-eeprom-access: yes supports-register-dump: yes supports-priv-flags: yes

Kernel:

# uname -r 4.1.12-124.27.1.el6uek.x86_64

Any help would be appreciated!

[Caguicla_Intel]

Hello RHelf, Thank you for posting in Intel Ethernet Communities. Kindly share the details below for us to check on your query: 1.) Bond configuration found in the ifcfg-bond0. 2.) Does the issue occurs with the bonding with onboard NICs? 3.) Please provide the markings of the Intel NIC found in the white sticker of the adapter. You may refer to the link below on where to find the PBA and serial number of the adapter: https://www.intel.com/content/www/us/en/support/articles/000007022/network-and-i-o/ethernet-products.html You might also want to consider posting in the oracle community site as there might be other user who experienced the same issue using Oracle Linux. https://community.oracle.com/community/groundbreakers/server_&_storage_systems/linux/oracle_linux, Looking forward to your response. Best regards, Crisselle C Intel Customer Support Under Contract to Intel Corporation

[RHelf]

Sure thing:

#1:

# cat ifcfg-bond0 BOOTPROTO="none" DEVICE="bond0" ONBOOT=yes PEERDNS=no PEERROUTES=no DEFROUTE=no TYPE="Bond" BONDING_MASTER=yes NM_CONTROLLED=no BONDING_OPTS="mode=1 use_carrier=0 primary=eth0 arp_ip_target=10.10.10.254 arp_interval=1000"   # cat ifcfg-bond0.156 BOOTPROTO="none" IPADDR="10.10.10.104" NETMASK="255.255.255.0" GATEWAY="10.10.10.254" DEVICE=bond0.156 ONBOOT=yes PEERDNS=yes PEERROUTES=yes DNS1=10.10.10.21 DNS2=10.10.10.22 VLAN=yes

#2: Yes, it's the 4 port onboard NIC. Here is info from Dell's DRAC:

Intel(R) 10GbE 4P X710 rNDC

Number 3 is going to be harder to get. I will need to have someone at our data center open the case and get that information. I'll try to get it.

Also, we have a case open with Oracle support for this as well. I'll follow up if I get any further information from them. They had me add the following, which did not make any difference:

# cat /etc/modprobe.d/bonding.conf alias bond0 bonding options bonding max_bonds=2   ## ADDED THIS: options bond0 mode=balance-alb miimon=100

[Caguicla_Intel]

Hello RHelf, Thank you for the reply. We will check on this and get back to you as soon as there is any findings. Hoping for your patience. Best regards, Crisselle C Intel Customer Support Under Contract to Intel Corporation

[RHelf]

Hi all,

I really appreciate the answers. After a lot of troubleshooting, we have determined that the issue was our B-side top-of-rack switches. They were not passing traffic correctly. A reboot of the switch actually fixed our issue.

Thanks for the views and the replies!

[Caguicla_Intel]

Hello RHelf, Thank you for the update. We are glad to hear that the issue has now been resolved. If you have any other concerns or assistance needed from Ethernet support team, feel free to post on Ethernet Community. Best regards, Crisselle C Intel Customer Support Under Contract to Intel Corporation