Network receive problems with Jumbo Frames at 9014 on Intel PRO/1000 MT Dual Port Server Adapter. Recently we started having problems with iSCSI on multiple Dell PowerEdge 1850 servers when using an "Intel PRO/1000 MT Dual Port Server Adapter" 1 GbE NIC with Jumbo Frames set to 9014. When placing heavy receive pressure using iSCSI and IOmeter with 32 concurrent 64 KB or larger sequential reads, we see many TCP connection resets and iSCSI session drops repeatably. IOmeter iSCSI writes are fine. The problem happens on two identical Dell 1850 servers, one running Windows Server 2008 Standard, and the other running Windows Server 2008 R2 Standard. However, a third system running Windows Server 2003 Standard in the same configuration works just fine. We know that Jumbo frames is configured correctly, as Jumbo pings work, and the iSCSI writes work well. Multiple network switches show the same problem. Oddly, when using the Dell 1850 motherboard network ports instead of the PCI adapter ports, which are the nearly identical "Intel PRO/1000 MT Network Connection" devices (uses the same driver, etc.), the problem does *not* occur on either Windows 2008 system. Also, if I use a Jumbo frame size of 4088 instead of 9014, the problem does not occur on either Windows 2008 system. All Windows Updates are installed, and I've tried all possible network port configuration settings to work-around the problem. The only one that works is changing down to the Jumbo Frame size of 4088 on the adapter ports, leaving the motherboard ports set to 9011. We are not sure, but we believe that this problem started happening recently, so may have been introduced in Windows 2008 and Windows 2008 R2 recently with some Windows Update. However the NIC driver hasn't changed since 8.4.1.0 on 5/28/2008, so it would have to be some other change. I don't really need any support per say as I have a work- around, but would like the problem fixed so it doesn't cost me so much time when I encounter it in the future. I really just wanted to report this to Microsoft, but there doesn't seem to be anyway to do this without purchasing support, so I guess I have to report the problem here and hope that somebody at Microsoft in the right department will read it. Thanks. ~ Jeff Byers ~

Any update on your situation? We recently loaded one of our old PE6850s with Windows 2008 R2 Enterprise (x64) and are having similar iSCSI issues. It seems to be specifically with Jumbo Frames. None of our 2003 servers, 64-bit or otherwise seem to be having these issues. I've been trying to find a more recent driver for the pro 1000/mt- with no luck. ~M. McCabe--Systemstate

Hi, Glad to read this.. I had the most frustrating experience the last few days. I've installed a 1000MT dual port server card in one of our Poweredge 2800 servers, to connect it to our Equallogic. I re-installed the machine with 2008R2. The installation procedure for the iSCSI host integration tools had been performed at least 30 times before, on other servers, without any problem. I could connect to the Equallogic San, 2008R2 would see the defined LUN, and the volume appeared in the disk management. Problems started when we wanted to (quick) format the volume... It always ended with an error stating the disk could not be formatted... 4 reinstalls later, about the same amount of re-creating the LUN (checking our installation notes 100's of times to make sure we hadn't forgotten something), changing cables and switch(ports).... it came all down to the network card and its' jumbo frame settings... Set it down to 4K in stead of 9K, and all went fine... Tried to find updated 2K8R2 drivers for this card.. but no luck on Intel's site..

With the older Intel drivers that are part of the "in-box driver only" support category for Server 2008 R2 and later, a couple of things to note: 1) If you download the latest Pro 1000 Net Drivers from Intel, extract the zip file, then manually apply them to the NIC "Update Driver...," you will then be able to access all of the additional menus to manage your NIC. Choosing Configure Adapter will display the teaming, ISCSI (if applicable) ROM, VLAN, etc. menus that were previously unavailable. 2) Maxing out the Send and Receive buffers to 2048 has worked for me on that particular nic as far as jumbo frames are concerned.heuristik

Hi, I know this post is pretty old, but as there's no really answer of the root cause but just workaround, I will share what I've found so far. I've exactly the same issue on one of two Dell PE 2950, able to connect Lun on an EqualLogic, able to make the drive online, able to create new partition, able to name it, etc. The only thing I couldn't do is formatting (quick or not). Same useless error message popup without any entry in the event log of Windows or EqualLogic Pinging the SAN with don't fragment flag and a size of 8192 show randomly missing packets, changing buffers or interrupts moderation put the server to BSOD while applying changes. Finally, to make a long story short, the culprit was the two (yes two in a row) PRO 1000 MT Dual port. After researching a lot, those cards appears to be fake one. There should be a sticker on the back side of the cards with a 'YottaMark' for the guenine Intel cards. http://www.intel.com/support/network/sb/CS-031541.htm HTH jbongran