DvSwitch issues

[Sjmry1]

Hi

We have recently deployed DvSwitches on a new ESXi 5 cluster of 5 hosts.

We are having a few issues though.

We are connecting to Cisco nexus 2000's with 10GB networking

I have 2 ports from each host connected to the DvSwitch (these ports on each host are in an port channel group on the Cisco switches) and the switchports are in trunk mode.

I have set the load balancing method to use "Route based on IP Hash" on all port groups that have been created - the DVUplinks group remains "route based on originating virtual port" and is greyed out so cannot be changed - I know the port group settings should override this though.

The issue is that when any of the VM servers try to ping another server in the same port group there is substantial packet loss.

What could be the reason for this?

1 of the hosts that shows the most significant problems shows under Network Adapters > Observed IP Ranges a list of VLANs and IP subnets - whereas all other hosts display None in this field.

Any help or ideas would be great

Thanks!

[rickardnobel]

Sjmry1 wrote:

The issue is that when any of the VM servers try to ping another server in the same port group there is substantial packet loss.

What could be the reason for this?

The VMs on the same portgroup are they also on the same hosts?

[Sjmry1]

yes the vm's could be on the same or different hosts.  1 host in particular causes large packet loss

thanks for your response

[rickardnobel]

Sjmry1 wrote:

yes the vm's could be on the same or different hosts.  1 host in particular causes large packet loss

If they are on the same host and on the same portgroup then the Etherchannel / IP Hash setup should not be the cause for the packet losses in these cases.

Do you know if the host is very high on CPU? Or the virtual machines? And how is the network load during the times you see the network problems?

[Sjmry1]

Hi

no the VMs and hosts are using very little CPU - there is little to no network activity as we havent gone "live" with the systems as yet.

thanks

[vmoney]

Any setup differences between the Cisco switches? If possible, flash the troble switch with the IOS of the one working well. Sometimes it is not a VMware issue. Sometimes it is. Best of luck.

[Sjmry1]

we have checked and confirmed that it isnt a switch issue already, I wish it was!

The observed IP ranges on the affected ESX hosts list subnet ranges whereas the observed ip ranges on the working hosts are stating "none"