Desktop refreshes, nic not connected

[HendersonD]

We have a pool of non-persistent desktops that when someone logs off and the desktop refreshes, the nic is not attached. This is not happening on logoff all the time, but at the end of the day I usually see 10-20 desktops in this state in a pool where 150-200 are used during the day. They show in the View console as "Problem vCenter VMs" and in the vSphere console they all have an IP address of 169.254.243.23. When I edit settings on any of these desktops and click on the network adapter, the Connected checkbox is not checked.

Any idea why the nic fails to connect when the desktop refreshes?

[Chiefnmaster01]

Hello HendersonDHendersonD.

Are you using distrubted switch or are you using a local switch ?

Can you please check if there are free ports avaibible on the switch ?

Can you post the switch config ?

[HendersonD]

Distributed switch

Ephemeral - no binding so I should not be running out of ports at least that is my understanding

[amatt12]

Yep, we have had the exact same issue since our release. ~7% of VM's would be deployed with disconnected NIC's. This is with a Nexus 1000v.

We're upgrading vCenter/ESXi to 6.0 U2 soon, so I'll reply here if it fixes anything.

[Gezmonder]

Hello guys,

Did you ever get anywhere with this? We have this exact issue and I'm getting nowhere with support. Having to keep connecting the NICs on the VMs to bring them online. During and after recompose the issue is particularly bad.

[HendersonD]

I am the original poster and have had an open case with VMWare for two months on this issue. To date I have done the following items. These are items I did independent of the items VMWare asked me to do:

1. Upgrade our Juniper switch firmware
2. Upgrade the firmware on our hosts including nic firmware. We use HS23 blades in and IBM Bladecenter as our hosts
3. Upgrade the firmware on our Pure Storage array
4. We were already running ESXi 6 U2. VMWare released a few patches recently which we installed
5. We only have 4 master images, all Windows 7. We rebuilt these 4 images from scratch. It was a good time for us to do this anyways since we were deploying Office 2016 and pulling off some software like flash, java, etc. but this was still a good bit of work
6. Upgraded from hardware version 9 to 11
7. Checked my switches to see if any input or output errors were shown, none were found
8. The dvSwitch we were using was a bit funky, I could not upgrade it to enable advanced features like Advanced LACP. No we do not use LACP but the shear fact that I received an error message when I tried to upgrade the dvSwitch had me concerned. Installed a new dvSwitch, added new portgroups, and migrated desktops, iSCSI, and vMotion to the new dvSwitch.

VMWare had me do several other things:

1. Installed Microsoft Hotfix Hotfix KB2550978 which is described in VMWare KB1020078
2. Installed an older version of VMWare Tools
3. Inside of Windows, on the nic properties, unchecked the box allowing the computer to power down the nic
4. I sent VMWare an extensive set of log files including logs from vCenter, Composer, Hosts, and desktops where nics were connected and ones where nics did not connect

To date, we do not have a solution. At this point I do not think the issue is related to host firmware, switches, or storage. I think it lies with one of the VMWare components like Composer, VMWare Tools, Agent, or the Host. Unfortunately we have not been able to narrow it down any further.

I did ask if upgrading to View 7 might solve this problem but have not received an answer yet. I would encourage others to open a support ticket if you have not done so yet.

[amatt12]

Would you guys be able to send me your VMware case numbers? The engineer assigned to my case reports that they have an internal bug reported that their engineers are working, but they can skim your case notes and see if it is related.

[HendersonD]

My case number is 16968067904

[gshivaram]

Can you share the VMware's Case number

[amatt12]

15841019912

[Gezmonder]

I've had a couple of SR's 16952922904 to look at the vNetworking side, this has been open since early April. This week they cloned it to 16144303106 to look at the Horizon side but that went straight back.

We have a good 150-200 desktops that are unusable on a daily basis as a result of this issue.

We are using Horizon 6.2 with vSphere 6 U2 with ESXi 5.5 Express Patch 10

[Gezmonder]

I've heard this morning that It looks like this will be fixed in vCenter 60U3.

[amatt12]

Do you have a source on that that you are allowed to share?

[Gezmonder]

If you are already chatting to VMware just use my SR references above. I heard this from both engineers this morning saying pretty much the same thing. They reckon it's down to using ephemeral ports and that we should use static ports as a workaround. We've not tested this yet but I think we will as there's no telling how long we'll need to wait for U3 and if indeed it will be fixed for sure.

[HendersonD]

I switched from Static to Ephemeral port binding quite awhile ago since we were having this issue

https://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=10211...

Now to solve the nic not connected issue I may need to switch back to static. It sounds like the lesser of two evils. Which problem do I want to have?

I wish they had an ETA on ESXi 6.0 U3

[Gezmonder]

So we have switched from Ephemeral to Static port groups now and the NIC connection issue is now gone

Luckily for us, we didn't have the recompose issue.