we have IBM Bladecenter H and 3 HS21 XM as ESX Hosts, 2 Nortel 3/8 Ethernet Switches only for Vmotion between the 3 Hosts and 2 Nortel 1/10 Ethernet Switches.
Each ESX has 2 active NiCS on Vswitch1 for LAN connected to both Nortel 1/10. They have a LACP trunk to one (I know it´s spof ...) HP 5308 with 3x1GB.
Nortel and HP Trunk looks good.
Pings on VM´s work for 10-20 pings and get loss for a while and come back.
On HP and Nortel we see no errors.
1) Is this the way we should do Loadbalancing ?
2) What type of loadbalancing should we set on Vswitch ?
Please see my PDF attached.
trunk between nortel & hp is okay but not neccessary if you use the preferred vmware loadbalancing method "route based on the originating virtual port ID" !
but do never use a trunk to connect esx servers to a physical switch! only use the vmware default value @ vswitch "route based on the originating virtual port ID" ! it really works fine and you don't have to configure the physical switch !
do you really need more performance then 1Gbit/s in the other blades with windows ? hs21 blades default have broadcom NICs and it should possible to configure failover NICTeam only ...
OK, one step back.
the Nortel internal ports 1+2 which are connected to ESX1 are NOT Trunked !
Only 2 x 3EXT Ports whoch are connected to HP Switch !
Is this also no good for ESX ? I know that vSwitch do loadbalancing by itself.
sorry I don't have a solution for you but an idea to isolate the problem:
put your vm into the network managment vlan if it exist an vlans non routed.
do many pings parallel from the vm:
to your pc (from picture), to the nortel 1/10 management ip, to the hp53xx management ip ...
is there a timeout from vm to the nortel 1/10 managment ip ?