Start a Conversation

Unsolved

This post is more than 5 years old

M

2841310

September 2nd, 2013 13:00

Opti 9020s with Intel I217-LM nics causing network issues

I have 9 Optiplex 9020 towers with Intel I217-LM nics.  When they are connected to my network, eventually the network suffers severe latency issues, making some of it unusable.  The problem disappears immediately when I disconnect the 9020s.  Problem is somewhat intermittent, resolution varies.  Usually unplugging any 3 or 4 of the workstations is sufficient for a while, but the remaining workstations eventually cause the same network problems.  I have isolated the problem to the workstations themselves by using different switches and physically moving the workstations to different wiring.

I have tried multiple fixed IPs as well as DHCP.  I have reimaged the systems, with only Win7 Pro and Windows updates installed, including the latest Intel drivers, the issue remains.  Dell support advised a BIOS version patch, and drivers off dell.com instead of windows update (an earlier version), but I don't hold much hope that will resolve the issue, as the BIOS patch has no network card changes, and I've tried multiple drivers already.

I've noticed the issue can occur while the workstations are "asleep," so I've turned off "respond to ARP requests without waking system" and couple similar sleep related settings.  Still waiting to see if that had any effect.


Any advice, correction, things to try, or even just sympathy would be most welcome!  Thanks for reading.

February 25th, 2014 17:00

The traffic that is occurring when the device goes to sleep appears to be a faulty handling ot the Energy Efficient Ethernet feature.  BIOS versions A02 and prior seem to have the problem.  A03 doesn't seem to but as mentioned elsewhere is unstable.  A05 is the current version and that seems to stop the problem.  Visually, the Ethernet lights will be out when asleep instead of on.

My understanding of how the mess works is that the device, when asleep, sents something out to an unknown address that turns that into a multicast.  This reoccurs thousands of times per second.  By itself, a single device might only degrade your network.  For us, when we had 3 devices asleep at the same time the switch and adjacent router were overwhelmed.  The router failed over to the redundant router which then also failed back to the orginal router and on and on.  Waking up the devices stops the noise.  Disabling the EEELinkAdvertisement (1 to 0 in a few places in the registry) also seemed to shut the device up, but the BIOS update seems like the right answer.

Remember, if you have Bitlocker protection for your drive to disable Bitlocker before doing the BIOS update or you will be looking for your encryption key.

manage-bde -protectors -disable c  (as admin in a DOS window)

I think the M4800 has the same NIC.  I'm not sure if there are good and bad BIOS versions for it.

8 Posts

February 26th, 2014 03:00

Hi All,

  we have been experiencing strange network issues. My collegues used a fluke to identify high traffink talkers which turned out to be OptiPlex 9020 AIO's using the I217-LM network card. I use the dell Cab files for the drivers & checked both A01 & A03 which both comntained driver version 12.6.47.0. I upgraded the drivers via windows update which went to v12.11.77.0 which was released 11/02/14. Once this was applied & the machines shut down they then dropped of from being a high talker on our network.

 

So windows update network drivers seems to resolve the issue

March 5th, 2014 01:00

Little late to this party; we're planning to roll out hundreds of 9020s in the next few months but our first batch have network issues. Less of the broadcast storm described here, more not networking at all. I think our switches might be shutting them down as soon as they start -- has anyone experienced anything similar?

1 Message

March 11th, 2014 18:00

Please retest using LAN driver version 19,0 available from the Intel support web site. Thanks!

1 Message

March 14th, 2014 10:00

Couple notes:  Many customers that I work with are experiencing this issue.  Based on the Intel Support Community post associated with this issue https://communities.intel.com/message/220048#220048 , Intel posted this message 

"Yes Intel is aware of this issue. Please retest using our latest LAN driver version 19.0 release, available from our Download Center, and let us know if this driver helps resolve your issue. Thanks!"

We have had customers test this version and it does appear to eliminate the problem without the need to turn off power saving and IPv6.  Please try this out and post your results..  

If your network teams are interested (and they should be), here is the actual packet debug of the IPv6 multicast .. 

------- dump of incoming inband packet -------
interface NULL, routine mistral_process_rx_packet_inlin, timestamp 10:29:55.297
dbus info: src_vlan 0x373(883), src_indx 0x1070(4208), len 0x5A(90)
  bpdu 0, index_dir 0, flood 0, dont_lrn 0, dest_indx 0x5802(22530)
  4E820400 03730000 10700000 5A080000 0C000030 07000004 00000000 5802C7C6 
mistral hdr: req_token 0x0(0), src_index 0x1070(4208), rx_offset 0x76(118)
  requeue 0, obl_pkt 0, vlan 0x373(883)
destmac 33.33.00.00.00.01, srcmac C8.1F.66.A8.72.7A, protocol 86DD
protocol ipv6: version 6, flow 1610612736, payload 32, nexthdr 0, hoplt 1
class 0, src FE80::CA1F:66FF:FEA8:727A, dst FF02::1

 

 

1 Message

April 10th, 2014 00:00

I've been having the exact same problem for a couple of months now and haven't found a resolution. After about 24 hours one of my cores will have 100% load, "system interrupts" are consuming a full core. Using device manager to disable devices I tracked the problem to my NIC. Using a program called latencytimer I verified this further, it shows the high cpu usage is related to the NIC. If I quickly disable/enable the NIC it will be good for approx another 24 hours, but then I loose all my active Connections (like ssh prompts). If I disable every option on the power mgmt tab the NIC will be good for up to a couple of days sometimes, but the problem WILL return. When the problem has returned it can be fixed by re-enabling(!) any of the power options. Each time I fiddle with the power options I will loose all my active connections. If i run diagnostics for the card (tab link speed -> diagnostics -> hardware ) CPU usage will come down to normal levels and I will NOT lose active connections. However, I'm not interested in doing this each and every day. Have tried drivers from Dell as first option. Have tried drivers from Intel (2 or 3 updates), have flashed BIOS every time a new version has been released. This problem is driving me mad!

1 Message

May 22nd, 2014 18:00

Hi. Dell have alerted me to this issue after supplying a few thousand of them to my organisation. Dell are unable to tell me whether we need to be using IPv6 addresses to experience the issues. Since it will be a hassle updating all machines; I'd be interested to know if anyone experienced the problem whilst using IPv4 addresses and simply had IPv6 protocol bound to the adapter which is how our machines are all configured (ie just ticked but really being unused). It sounds as though the answer to this is yes otherwise surely someone would have piped up about switching back from IPV6 to IPV4 would mean changes to DHCP servers and be a bit of a hassle. thanks GB

8 Posts

May 23rd, 2014 01:00

BROWNEACTION, I've got this issue on an IPV4 network so don't go to the effort of swapping back. try the latest driver or use the windows update drive as that sorts the issue.

May 23rd, 2014 07:00

@BROWNEACTION, We experienced this with IPv4 addresses and simply having IPv6 protocol also enabled on the clients. Hope that helps.

September 17th, 2014 10:00

Hey, just wanted to chime in here that I was having the same issue on my network with HP Elitedesk PCs with this same onboard NIC, and this was the first link Google suggested.  Sure enough, disabling AMT fixed it, and other strange issues I've been having on older machines were also resolved.


AMT sounds not ready for actual deployment in any sized business other than a lab.  Which is surprising, since it's had time to mature a bit.  It'ss interesting that this is ON by default, and that Intel doesn't advertise this ability to cripple the network and machines its used on in the product literature.

1 Rookie

 • 

29 Posts

September 21st, 2014 11:00

Hi!

I'm not really sure if my problem is related to the issues described so far in this thread but I have a very annoying problem with our Dell 9020 towers with I217-LM NICs running Windows 8.1. When the computers wake up from sleep (standby) all connections to open network shares and open Office 2010 documents are lost. All shares are on a Windows 2012 R2 ESSENTIALS server + domain controller which runs as a VM. So after waking up a Dell 9020 from standby I get my Win 8.1 login prompt, log in, but for any open shares that were open when the PC went to sleep in explorer, now I randomly get either a domain authentication error message or a new authentication prompt that asks me for domain username and password. The connection to open Word 2010 docs is always lost and Word tells me that the computer came back from standby but a connection problem occured and the document is now read-only, i.e. it must be saved as a new document. Very annoying!

I found out that similar problems date back to Windows 7 or even XP and people suggested to disable "computer can turn off this device to save power" on the network adapter in device manager. I tried this with the latest Dell NIC drivers but this did not help at all. I also tried installing the latest Intel NIC drivers but this did not help either. I also don't see the "computer can turn off this device to save power" in Intel's latest drivers. Maybe those NICs just don't support the option to stay on during standby? BTW I have not enabled AMT so it should not interfere.

Any help or hints are welcome. Thanks in advance

Anguel

1 Message

September 25th, 2014 16:00

Hi Guys

Basically I was having the same issues on my network that they were flooding it with broadcasts/multicast.

I had to enable storm control just to get by so it wouldn't crash my network or would take much longer.

I just want to confirm 3 things.

1.The IPV6 fix has that fixed the issue for you guys it seems too from what I read.

2.Disable Power saving features on adapter from what I can tell fixes the issue.

3.Update driver from Intel -- This is the one I would like to use as it's more permanent solution, Can anyone vouch for this solution in terms of it solved there problem without doing the above two steps.

6 Posts

September 25th, 2014 20:00

Hi Anguel,

The problem you are describing does not appear to be the same problem we were having with the multicast storms occurring. At first glance, it appears that it might be a software/windows issue and not specific to the NIC itself.

Do you see this behavior on other machines on your network that do not have this NIC?

-Gordon

6 Posts

September 25th, 2014 20:00

Apologies, I meant to update this post earlier. When this was happening for me, Dell worked with me over four months or so before we got a more permanent fix.

The latest drivers provided from Dell should fix the issue. I was receiving 9020s with older drivers from the factory, but I implemented the latest drivers from Dell into our SCCM task sequence. Since then, the new machines have not repeated the behavior.

I'm surprised that Dell still ships their factory image with the defective drivers. I suppose they expect that most enterprises will re-image with their own. But still....

Anyway, I'm thankful that Dell assigned someone to work with us directly. It took some time, but they did release a fix and I appreciated their attention to the matter.

-Gordon

September 26th, 2014 02:00

Everyone in this thread is seeing the symptoms of a common issue.  Your mileage varies depending on how much you have to go wrong when you get devices with this NIC.  Depending on the model you have, you may have a BIOS version that mishandles the Energy Efficient Ethernet feature.  You may have a driver that mishandles the EEEF.  You may have switches that get very confused by the TCP/IP v6 anycast traffic that results from these problems.

Bottom line....get the right BIOS and driver.....then when these devices go to sleep they wont spew data.  Enabling and disabling the NIC or messing with the power settings just restarts the timer until it will start spewing again.  Messing with the TCP/IP v6 settings might fool it is to not doing the anycast.  My situation was that things when VERY wrong when 3 or more slept on the same network switch.

This is one of those problems where you don't diagnose it....you just fix your systems and move on.

No Events found!

Top