windows network connections reset after change in initiator

barrysmoke · Tue Oct 21, 2014 10:19 pm

not an issue with starwind, but an annoyance none the less...I've noticed that in my lab, any change in microsoft's iscsi initiator( either adding, reconnecting, disconnecting targets, etc...to starwind iscsi targets) resets the network interface, so that any connections over other services are interrupted. is anyone else seeing this, and if so, should we submit a microsoft bug ticket?

thefinkster · Wed Oct 22, 2014 8:29 pm

I have never seen this with iSCSI Initiator from Microsoft. Using it with various devices: Drobo, Iomega IX12-300R, VNXe3150, NX4 on Windows 7, Windows 2008, Windows 2008 R2 and Windows 2012 R2.

I'd say there's an issue with the installation of the server (either Role, NIC driver, etc..), but that's just my gut feeling.

barrysmoke · Fri Oct 24, 2014 12:47 am

this is virtual environment, so could indeed be a vmxnet3 issue.
I'll see if I can replicate on my physical server with starwind, and determine if it is virtual only...
Thanks!

Sun Oct 26, 2014 12:00 am

I have never seen this with iSCSI Initiator from Microsoft.

+1 to that!

barrysmoke · Sun Nov 09, 2014 11:25 pm

This did appear on our physical box as well. I had a vmware engineer remotely logged in over teamviewer, and when I created a starwind iscsi target, and connected it with the ms iscsi initiator, the nic went dead.
the vmware engineer suggested I look at the arp table, and shure enough, the default gateway had been assigned to one of the other nics upon initiator reset. I finally ended up disabling all the extra nics, and disabled, and re-enabled the nic to get it back up. Does anyone know how to prevent the arp table from switching nic's for the default gateway?
I've never seen this kind of weird behavior in windows, and suspect it is a combination of everything, like teamviewer, and starwind, and ms initiator.

going to try to remove teamviewer, and see if that makes a difference.
I can reproduce this every time though, not good, but should be easier to troubleshoot now that I have a clue what is happening.

barrysmoke · Mon Nov 10, 2014 12:20 am

not arp, not teamviewer,
even though I'm using static ipv4, a ipconfig /release fixed the issue, networking came back for outside addresses....strange.

thefinkster · Mon Nov 10, 2014 1:17 pm

Hm. Static assignment on the NIC and when you do an ipconfig release/renew, you get an IP number change. That does not sound like static assignment of IPs. DHCP delivered reservation, sure. But hard-set IPs, no. Or you have multiple NICs plugged into the network.

You'll want to post your hardware setup as well as network setup (IP diagram would help) if you have more questions.

barrysmoke · Mon Nov 10, 2014 4:42 pm

yes there was a second nic plugged in as a backup, in case the primary goes out...and it did have a dhcp address, in my lab. this may be where the issue is.
We'll table this, and I'll post back if I ever have it pop up again.

barrysmoke · Mon Nov 10, 2014 8:38 pm

I recreated the situation, leaving one nic dhcp, and the other static.
the main nic was static, and the second nic was a backup...the moment I click refresh on the discovered targets under iscsi initiator, a ping-t 8.8.8.8 stops responding, until I ipconfig /release
then it starts responding again. do not mix dhcp for starwind servers, even under lab scenarios...you get problems!

barrysmoke · Tue Nov 11, 2014 3:36 am

nevermind, disabled all nics, but the 2 10gig nics my iscsi/jumbo connections are on for sync, and the 1gig management network, that has a static ip.
every time I go into the MS iscsi initiator, and click refresh, I lose internet connectivity, but the nic stays up, and my remote desktop session stays up.
only way to get internet back is to wait(I believe starwind tech support said they would wait 15 mins, or so, and the connection would be back),
or reboot.

looks like I'm going to need assistance in figuring this one out.

barrysmoke · Tue Nov 11, 2014 3:48 am

my next thought is to update the nic driver from the default windows, to the intel driver, however it is strange I would see the same behavior on physical windows, and a virtual windows 2012 r2 server, using vmxnet3.

barrysmoke · Tue Nov 11, 2014 4:13 am

sorry for so many updates, but I'm just posting info as I troubleshoot.

I can recreate this every time, and now have a sure fire way to get the wan back. I have to go into the adapter tcp/ip v4 settings, and change the subnet mask from 255.255.255.0 to 255.255.0.0, then back again...which seems like the only non-disruptive change I can make, while being enough of a change to force the nic to start acting right again.

downloaded latest intel driver for my i350 nics, software installed, driver kicked in, checked intel proset driver settings, everything looked good.
ran test again, clicking refresh in the ms iscsi initiator, bam, loss of wan, but nic still up.
change nic ipv4 mask, and goes right back up....so strange.

I would love to show this to bohdan, or someone, logged into my workstation, watching me recreate this, I'm scratching my head!

barrysmoke · Tue Nov 11, 2014 4:37 am

problem only occurs on refresh button click, simply disconnecting, and reconnecting targets causes no issues, maybe this is why no one sees this, I'm the only one that clicks the refresh button, lol.

Sat Nov 15, 2014 10:33 pm

Well, just tried that - haven`t seen anything.
Just wondering - have you asked this question at MS forums?

barrysmoke · Wed Nov 19, 2014 6:39 pm

will do...you tried on windows 2012 r2?
going to try different OS's, and see where my issue pops up at.