Hoping someone can help.
Recently upgraded a server running Xen from 3.18.44-20.el6.x86_64 to 4.9.39-29.el6.x86_64 (did a yum update so other packages would have been updated to).
Since rebooting, we randomly get "e1000e: eth0 NIC Link is Down", this happens for only a second, but since we have this interface bonded it can be enough for it to bring down the interface fail over, and then come back again, which causes packet loss.
If tried turning pcie_aspm off, flow control off, nothing seems to help. The problem seems to be worse when high traffic is being used, but also happens randomly as well.
NIC Intel Corporation 82574L Gigabit on a Supermicro box. Driver 3.2.6-k
Have now upgraded to 184.108.40.206, the problem seems to be happening a lot less often, but I'm not sure if its a coincidence.
Thank you for posting at Wired Communities. What is the exact model of your Supermicro box? Where did you download the driver version 3.2.6-K and 220.127.116.11?
As this is an onboard NIC, have you contacting the Supermicro support for the driver? Please feel free to update me.
3.2.6-K Came from centos provides kernel the 18.104.22.168 came from intel source forge
The motherboard is X9DRL-3F/iF
I will try contacting supermicro also but hoping you might be able to help as this seems to be only since the upgrade.
Thank you for the information. I will try check and for the meantime, please do contact Supermicro and update me if there is any result. Thanks.
Just to double check is there any changes done on the system prior to experiencing the link down issue? Have you contacted Supermicro* and any updates?
As stated in my initial message, we ran yum update which updated the kernel as well as other packages (which in turn also provided a new version of the driver from centos than we were using previously)
Supermicro have just requested further information on there replies but that is all. Upon further research I found some mentions of using IntMode=1,1 I set this in modprobe.d and restarted the network, it has since been stable.