VMWare ESXi 4.0.0 management IP address drops traffic

[RocketCBR]

I have a new HP DL380 G6 server, VMWare ESXi 4.0 with the HP customizations, fresh install and it's having some trouble with the management IP address only.

This server has 4 NIC ports in it, and initially I tried to add all 4 ports trunked to ESXi. But then I ran into problems with the vSphere Client and it losing communication with the ESXi server. So I tried setting up just 1 NIC port dedicated to the management network with no VLAN. It has a static IP Address. I tried swapping network cables, checking the switch port settings, etc.

Funny thing was that the VM's on this host all work fine, and respond well to network traffic.

I've also tried to update the firmware on the HP server and NIC with no change in improvement. I'm losing about 40% of the traffic on the management ip address and it's causing some issues when I need to manage this server. Again, the VM's seem to work fine. If I reset the management interface at the server management t does come back and starts to work right away. But given a short while it seems to hang for 5 minutes and then starts working again.

Any suggestions?

[AndreTheGiant]

Try to set the NIC to a fixed speed.

And check flow control setting on the switch port.

This server has 4 NIC ports in it, and initially I tried to add all 4 ports trunked to ESXi.

If you use LAG/Etherchannel on switch side you must also use IP based team policy on vSwitch side.

Andre

[a_p_ · Accepted solution]

I think there's a little bit of confusion on the wording.

When you talk about "trunking" on a HP Procurve, this actually compares with "Etherchannel" on a Cisco switch. And when you talk about a "trunk port" on a Cisco switch this is called a "tagged port" on the Procurve.

From what you write I assume that you use a Procurve and created a trunk for link aggregation. In this case you have to set the virtual switch policy to "Route based on IP hash" as AndreTheGiant already mentioned.

However, I'd suggest you leave the vSwitch policy at default "Route based on the originating virtual switch port ID" and set the physical switch ports to "tagged" (or "trunk" for Cisco). This way you can configure the VLANs on the virtual port groups and the uplinks attached to this vSwitch are assigned in a round robin manner.

For more information see:

http://www.vmware.com/files/pdf/virtual_networking_concepts.pdf

http://kb.vmware.com/kb/1006628

André