Apparent network issue with VM

[ashleymilne]

I have a single Dell T320 host running Esxi 5.5 build 1331820 that has two VM's on it. One vm is running SUSE Linux Enterprise 11 with Vsphere server and the other VM is running Microsoft SBS 2011.

At the moment we have an issue whereby when we ping the SBS 2011 vm from a management machine we get intermittent results, at the moment when I ping it I'll get these results - sent 39 received 32, lost 7 (17% loss), however if I ping the other VM running on this host, there is no packet loss at all, no matter how many tests I run. We have had this problem, intermittently for some time and it appears if we reboot the host server running Esxi, the problem goes away for awhile. Tonight, I was rebooting the SBS VM due to some network slow downs and windows updates, upon rebooting the packet drop problem has reappeared.

We have purchased two brand new network switches, this did not help. We have replaced network cables, this did not help. The problem seems to be specific to this VM but we don't know why, whatever the problem is it does not affect the other VM running on the host.

I just ran another ping test, this time 100 received, 82 lost =18 (18% loss)

The network card inside the SBS VM is an E1000

Any help would be greatly appreciated.

[cykVM]

Hi,

to narrow it down what does a ping from the Linux VM to the SBS and vice versa give you as soon as the issue appears from your management workstation? Can you ping the gateway from inside the VMs without issues?

Which and how many physical NIC(s) are built into the T320 host?

Is the SBS VM probably under (very) high load as the packet loss appears?

What you mean by "SUSE Linux Enterprise 11 with Vsphere server"? You running vCenter on the Linux VM or have a nested setup?

cykVM

[ashleymilne]

This morning the problem has disappeared on its own, which it has done before.

Now when I run a ping test, I get sent 377, received 377 lost =0 (0% loss)

Very odd.

[ashleymilne]

Users are still experiencing slow downs in applications, its clearly network related and I believe something to do with Esxi or the configuration of the VM, but I can not narrow it down. I did initially open a case with Vmware support however they believed the issue was external to Exsi. I have not re engaged their support as of yet.

Suggestions?

[cykVM]

It's pretty hard to help you without further details. It could be nearly anything. You have to narrow it down if it's the virtual network, the physical NIC of the server or the external physical network. Are the slow-downs the users experience specific to one appliacation (e.g. Exchange/Outlook, networks shares (fileserver) or ...)?

Could you probably put up a Windows 7 or 8 (or whatever you are using on the user's workstations) VM oin the server and try to do the same things that are slow to the users?

Maybe there are probably even any errors/warnings found in Windows event logs, or in VMWare logfiles and the SBS VM's logfile?

[cykVM]

Another thing comes to my mind: Is the SBS VM probably configured with 2 (virtual) NIC in teaming mode? This usually causes a lot of trouble with a DC. Best pratices from MS are to run only a 1 NIC configuration.

Maybe it helps to run BPA for SBS: Windows SBS Best Practices Analyzer (BPA)</title> <!-- integration code for PDF --> <...

[homerzzz]

I would try the vmxnet3 driver instead of the E1000 and see if that helps.

[ashleymilne]

Sorry I should of supplied more info, I apologize.

The SBS vm is configured with one network card. There are two vm's on this host, the other vm is a VSphere appliance, it is unaffected at least from what I can tell, when the dropped packet issue occurs, pinging the sbs VM results in wildly dropped packets, whereas pinging the VSphere appliance does not, ever.

I have checked the SBS windows logs and cannot find anything conclusive. I have not looked at any vmware logs as I do not know how.

The slow downs that users experience are across all applications, in some cases opening dialogue boxes within an application results in latency whereby users type a few words and have to wait for them to appear.

[ashleymilne]

What would be the correct steps to do this?

Add a second network card as VMXNET3 in host settings.

Once it appears in Windows, disable the current E1000 network card, change IP settings so that the new VMXNET3 adapter has the same settings as the old E1000, and possibly reboot? Remove E1000 adapter or leave it to see if problem has been resolved?

[homerzzz]

Yes, that method will work. I would just remove the E1000 adapter altogether once you have added the vmxnet3 adapter. I would reboot also.

[cykVM]

For me on a Windows 2008 R2 DC this method worked many times but sometimes it doesn't. I would be at least careful because with SBS not only AD is running on that VM but also Exchange. It might mess up things.

[ashleymilne]

I wonder if its a bug with this version of 5.5 that was perhaps addressed in a later update?

[cykVM]

That could be the case, you may also try to run an upgrade installation from a DELL customized VMWare 5.5 Update 2 installation image for example, see: VMware ESXi 5.5 Update 2 Driver Details | Dell US

You should also check for any BIOS/firmware updates for that specific server.

[homerzzz]

Didn't you say only one VM is affected?

[ashleymilne]

Correct, only the SBS VM is affected, whenever the packet loss problem occurs, we can ping the vsphere appliance vm no problem, it never drops packets, whereas the SBS is badly affected for some reason.

[ashleymilne]

Since I didnt use a customized Dell install disc when I initially installed 5.5 I wouldn't want to use one now unless I was doing a fresh install which I don't, that might mess things up if I did an upgrade that way.

[homerzzz]

Then I would start with fixing the VM and not the host.

[ashleymilne]

I agree, however I was just wondering if it might a bug with the E1000 driver in my installed version of 5.5 what was causing the issue. Because the vsphere appliance vm isn't used that often it might be affected and we just might not be seeing it, although it doesn't appear to be affected at the same time or in the same way that the SBS vm is.