Vmotion HELP!

[wgerrish82]

Hi Everyone,

just wondering if anyone has any thoughts on the the following problem?

Just trying to setup vmotion on a couple of HP BL460c blades on their on dedicated network using a 3com 5500g swtich and dedicated NIC's. tried the switch with the base config and also using VLans but neither config seem to work as i also get a timed out error when trying to VMotion. ESX seems to verify that vmotion will work and if i put a cable between the 2 NIC's for vmotion it works perfectly.both are on the same IP range and subnet but im wondering if i need to enable anything on the switch like adding in the IP route for the dedicated range on to the switch?

also something else, if i vmkping whilst connected on the 3com switch it fails but with the cable between the 2 esx servers again it works fine.

if anyone could suggest anything no matter how small that'll be great!

thanks in advance.

will

[admin]

Are you using nic teaming/load balancing for the vmkernel vswitch/portgroup? If so, what are the settings?

Do the switch port counters go up when you try to vmkping?

Do the nics show as up when connected to the switch? esxcfg-nics -l

[wgerrish82]

hi there,

no the VMkernal is not teamed/load balanced. how do i check the port counters after vmkping? (sorry still fairly new to ESX). yep all nics are showing as up.

also just tried a the 2 vmotion nics in a small netgear 5 port gig switch and it worked fine..

thanks for the help.

[admin]

I meant check the switch port counters on the 3com to see if they are receiving the data/pings from ESX.

I'm not familiar with the switch you are using but it definitely looks like a physical switch problem since the other switch worked fine.

Perhaps check the speed/duplex settings on both ESX and the 3com and make sure they both match.

[Johnnyz]

If you use port group on virtual switch, make sure set the port on physical switch as trunk port. if you don't use port group on ESX you need to set the port as access. trunk port = tagged on a HP switch. I'm not sure about 3 com, but you should see something close.

[jhanekom]

Do you have anything filled into the VLAN field of the VMKernel port group?

Many switches do not deal well with receiving frames tagged with the same VLAN ID as the native VLAN of that port. That is, if untagged frames received on that port are assumed to be VLAN 1 (usually the case) and you configure ESX to also use VLAN 1 (but to tag packets as VLAN 1), the switch may discard the packets.

Workarounds include:

- set the VMKernel port group to not use tagging by clearing any numbers in the VLAN field of the port group

- set the default VLAN ID (PVID) of the port on the switch to a value other than the VLAN that ESX is using

- set ESX to use a different VLAN ID and configure the switch ports to accept packets for those VLAN IDs (but obviously leave the PVID unchanged.)

It could also be as simple as the two ports on the switch just not being in the same VLAN, which would prevent them from being able to talk to each other...

[wgerrish82]

Hi Everyone,

just thought id let you all know how i got this problem fixed. All you have to do is add a static MAC address of the VMkernal that will be doing the VMotion to the port on the 3com switch for this to work.

took a while but finally got there!!!!

thanks for everyones help.

will

[jhanekom]

Just my 2c's worth, but I don't think you've gotten to the root of the problem. Be prepared for this (or related) networking problems surfacing again. It should never be necessary to manually hack port address tables, in particular not in a VMware environment.

Usually, if that kind of workaround works, it's an indication of a mismatch in load balancing setups somewhere, or that two NICs are attached to the same VLAN but that they're not in the same broadcast domain.

[wgerrish82]

no that's fair enough jhanekom....

although i only got to this fix after speaking with vmware support and they suggested trying this route because they've seen that sometimes that in some switches the mac-address-table doesn't get updated and they suggested to check if i could set static mac in the mac-addres-table of the 3com switch.

thanks