Intermittent connectivity issues with 2008 R2

[Sevich]

Got a weird one going on that I haven't been able to find a lot of info on. I have a mixed esxi 4 and esxi 4.1 environment. On a few of my guests, that seem to be all 2008 R2 domain controllers or database servers, all of a sudden every month or so they'll just stop responding on the network all of a sudden. We won’t be able to ping or connect to the servers.

If I open the vSphere console the box is still running just fine. As soon as I log into it, it starts pinging and responding again. Everything is running the E1000 network driver, I've made sure vmware tools is up to date, checked power settings on the nic within the OS to make sure it wasn't going to sleep.

Just curious if anyone else has seen this issue and/or might be able to offer me some advice.

Thanks in advance

[jcwuerfl]

Could you quickly explain your vSwitch configuration that the VM's are connected to. If its using VLAN's or is it just a straight access port for the uplinks? do you have multiple uplinks are they all active, standby? etc?

[Sevich]

Thanks for the reply. The one guest that I had flap last week is on esxi 4.1 and a regular vswitch doing vlan tagging. Right now that vswitch is hooked to a single physical uplink since I am in the process of getting things over to a distributed switch. This host is cabled to a Cisco nexus 7000.

The guest that I had issues with this morning is on one of the esx 4 boxes that I haven't migrated to our new infrastructure yet. This host is also using a regular vswitch without vlan tagging, 2 physical uplinks with route based on virtual port id. This host is cabled into a cisco 1gb switch that I can't remember the model number of right now.

[jcwuerfl]

Do you have portfast enabled with the port that the vlan tagging is on? I've seen STP cause some issues if that isn't quite setup right and know there is an example out there in the KB for that. I don't think it would be anything with having 2 active with vpid but you could always change it so you have 1 active 1 standby temporarly.

[Sevich]

I bounced this off our network guy and he told me we do have portfast enabled on everything in our environment. I'm going to try migrating one of the boxes from the E1000 to VMXNET nic and see if that helps out.