Network issue with ESXi 4.1

[zenariga]

Hello all,

I had a little environment with ESX 3.5 running fine, all VMs connecting to network without any problem.

Old environment: 4 ESX 3.5.2 and VC 2.5

A project team update my environment to ESXi 4.1 and VC 4.1, after this upgrade my virtual machines don't connect automatically to network when I restart or when I power off and power on. After guest OS up I need to disconnect the virtual NIC, wait 30 seconds and connect again.

This problem happens with all guest OS: Windows 2000, Windows NT, Windows 2003, Windows 2008, Red Hat Linux and Suse.

This problem happens with VMWare tools update and out of date.

This problem happens with all VM hardware version, I tested version 4 and 7

I create a new virtual machine and the problem occurs with new virtual machine too.

If I upgrade physical NIC card driver in ESX hosts solve the problem?

Anyone have any idea what is the problem? In my old environment I never had this problem.

[Mouhamad]

I believe you are using E1000 NIC adapter, try using the VMXNET3.

[a_p_]

Was this just a software upgrade or did you also upgrade/use new hardware? To me this sounds as if the ports on the physical switches are not configured correctly. Please verify spanning-tree is configured correctly (e.g. spanning-tree portfast or portfast trunk depending on the configuration)

André

[thealco]

That is odd - I would try a couple of things that would force reconfiguration (it is sometimes worth just changing stuff to reset a particular config) if you cannot see anything specific in the logs. Try unregistering and re-registering the vmware machine, or re-creating it (recycle the VMDKs and note down the current hardware config). You may have to reconfigure the NIC if you're not using DHCP.

I do not believe it is a spanning tree problem as it is occuring on the virtualised machines, not on the physical server.

[zenariga]

Hello A.P.

One specialist check all configurations about physichal switchs. He didn't find anything wrong with any police including Spanning-tree.

Thanks,

[zenariga]

Hello thealco,

I already tried to unregister and register the virtual machines, didn't work too.

I already reacreated one virtual machine to performed a test and the problem persist.

After reboot when the SO on virtual machine is up if I performed a vMotion the virtual machine can communicate with network too.

It's important to talk, all environment is the same: Servers, Storage, Switches, cables.

No hardware upgrades are performed in environment.

[venkatfuture]

Hi,

can you share what upgrade process ur team has followed, to help you on this

[a_p_]

To find out what could be causing this issue, I need to fully understand your current network setup (physical and virtual). Please describe the configuration of the Virtual Machine's vSwitch and port group (number of uplinks, policies, ...) and attach the configuration of ALL involved physical uplink switch ports.

What server hardware do you use, which vendor/model of NICs?

André

[zenariga]

Hello Venkatfuture,

I didn't participate of upgrade process but I talk with team responsable for that and they use a install CD, reboot one ESX per time and update one per one.

After upgrade Virtual machines were migrate to upgraded servers.

Nothing out of usual.

[zenariga]

Hello A.P. I don't have access to physical switches, This is admin by another team.

I have access to virtual switches.

I have one virtual-switch to Management Network using VLAN - one vmnic

two virtual-switch to iSCSI connection without VLAN - one vmnic each

One virtual-switch to virtual machines - two vmnics

3 port groups using VLAN

Traffic Shaping - Disable

NIC teaming  - Route based on the originating virtual port ID

120 ports

Server:

Dell PowerEdge R900

[a_p_]

I could be wrong, but to me this looks like one uplink is connected to an 802.1Q (trunk/tagged) port and the other one to an access port. Maybe one of the uplinks was plugged into the wrong switch port!?

André

[sakibpavel]

Are you using any active directory domain?

Sakibpavel

[thealco]

Just a question - when the guest OS boots, does it show that the nic is disconnected or it is connected but the server cannot communicate? It could be the nic teaming depending on the symptoms - ie if it appears connected make sure src-dst-ip is configured on your switch (if it's Cisco)

Show etherchannel load-balance

[Josh26]

zenariga wrote:

Hello A.P. I don't have access to physical switches, This is admin by another team.

I have a feeling this team are involved in the problem, and would suggest getting a copy of the switch config.

[zenariga]

Thanks A.P. I will check with network team if have a way to reconfigure the interfaces.

But I don't undestand why work without problems with ESX 3.5.

[zenariga]

Hello sakibpavel,

I don't understand the motive of your question but no, I don't use an active directory domain.

[zenariga]

Hello Thealco,

When the guest OS boot, boot with NIC connected and server can't comunicate with network.

The guest OS reboot, start without problems but don't comunicate with network, only if I performed a vMotion, disable wait 30 sec and enable virtual nic or change the port group and change back.

It happens with all virtual machines.

[zenariga]

Hello Josh26,

I will request a config of physical switches, but take a long time to receive an answer from network team.

[CARNSKI]

Hi zenariga,

Did you find a fix for this. I am having the exact same problem with some windows 2000 VM's using flexible NIC on Esxi  4.1 U2

After a restart the guest VM NIC is connected but can't communicate until the adapter is disabled and then enabled...

Regards

Gary

[zenariga]

Hello Carnski,

Yes the problem was solved, my case is a little different of yours. In my case all virtual machines have this error, W2k3, W2k and Linux.

In my case the problem was the physical switch configuration, my virtual switches was with 2 network adapters and in switch was not configured to support trunk mode.