John, I tried that with a straight through cable thinking that the MDI-X would make it work, but sadly it did not; I didn't have the tools to create a cross over cable at the time I was physically on site. I was also a little confused if that was even possible with 10Gb (I know I have done it with 1Gb before, but never 10Gb).
On Tue, Jan 5, 2016 at 5:29 PM, Ronciak, John <john.ronc...@intel.com> wrote: > > -----Original Message----- > > From: Tom Walsh [mailto:expressweb...@gmail.com] > > Sent: Tuesday, January 5, 2016 1:24 PM > > To: e1000-devel@lists.sourceforge.net > > Subject: [E1000-devel] X557-AT2 Problems > > > > Okay, I am at my wits end here with these NICs. > > > > We have a total of six X10SDV-TNL4F motherboards (2 x i350 & 2 x X557) > > connected to a QuantaMesh T3048-LY9 running CumulusLinux 2.5.5 (the > > QuantaMesh is using Trident II chips if that matters). > > > > I am having a complete devil of a time getting the X557 network > interfaces to > > connect at 10Gb. It seems to be completely random. > > > > Out of the six servers currently here is what happened on the latest > rebuild > > (Ubuntu 14.04 LTS with 3.0.19.0-43. kernel and 4.2.5 driver): > > > > Server 1: eth0 is connected at 10Gb, eth1 is connected at 1Gb Server 2: > eth0 > > is not connected at all, eth1 is connected at 1Gb Server 3: eth0 is > connected > > at 10Gb, eth1 is connected at 10Gb Server 4: eth0 is connected at 10Gb, > eth1 > > is connected at 1Gb Server 5: eth0 is connected at 10Gb, eth1 is > connected at > > 10Gb Server 6: eth0 is connected at 10Gb, eth1 is connected at 10Gb > > > > Server 1 and 2 are configured identically (X10SDV-TLN4F, 32GB of RAM, > > RAID1 SSDs), Servers 3-6 are identical (X10SDV-8C-TLN4F, 64GB RAM, 16GB > > SATADOM) > > > > I have tried the following (in various combinations): > > > > * Multiple driver versions 4.1.5, 4.2.5, 4.3.9, 4.3.13 (just released > earlier > > today) none of which seems to make a difference. > > * Updated the BIOS on the motherboard > > * setting: pre-up ethtool -s eth0[1] advertise 0x1000 > > * Multiple cables (Cat 6 rated up to 550Mhz - cables themselves are only > 2 > > meters long - Don't have a 10Gb cable tester to certify them though) > > * Turned off jumbo frames > > > > On some servers, when I set the pre-up advertise command and reboot, the > > interfaces never come back online. > > > > As an added data point, we have three servers with two X540-AT2 > interfaces > > in them connected to the same switch. Bonded interfaces, they all work > > flawlessly on the same switch, they link up at 10Gb without an issue > using the > > same cables (using the in tree driver). > > > > I am struggling to find any commonality between any of the servers to > even > > start to debug this problem. > > > > As an additional aside, if I bounce an interface (ifdown eth0; ifup > eth0) the > > interface never reconnects, I have to reboot the server for it to > reconnect. > > > > Help me, please. > > > > Tom Walsh > Have you tried to see what happens if you connect any of the NIC's > back-to-back? What do they link at if you do this? I know you can't > connect all of them that way but you can narrow down where the issue may be. > > Cheers, > John >
------------------------------------------------------------------------------
_______________________________________________ E1000-devel mailing list E1000-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/e1000-devel To learn more about Intel® Ethernet, visit http://communities.intel.com/community/wired
