I have just built a ESXi 4 cluster and an ESX 4 cluster, all with update 1 in a brand new datacentre, to the same design I always use. I am using standard vswitches, and the VMtraffic switch comprises of 2 pNICs each going to a different switch which uplinks to two stacked core switches.
The vswitch is set so both NICs are active, Beacon Probing Enabled, notify switches set to yes and failback set to yes.
I have a standard set of tests I run before going into production, one test is where I shutdown the port on the switch for the VMtraffic network, then enable and fail the other NIC to test failover times for the VM by pinging the VM from an external host. usually this is seamless, but in this datacentre it takes 15 - 30 seconds for the VM to start responding.
I set the failure detection to link status only and this still took approx 10 seconds to failover.
I have seen the port config and portfast is enabled, but I dont know what else it could be?
This happened on both the ESX4 cluster and the ESXi cluster.
All NICs are on the HCL from VMware and are standard HP DL380 G6 servers and NICs, same as used in another datacentre.
