Today I vmotioned 5 MS SQL 2005 servers over to a new DRS Cluster. All SQL servers use the "Production_LAN" network and a single virtual NIC of type "VMXNET 3".
The first 4 SQL VM (Windows 2003 Standard or Enterprise x32 bit) vmotioned over without a hitch.
The last SQL VM I vmotioned (Windows 2003 Standard x64 bit R2) vmotioned over without error, but I upon completion, I could no longer ping the VM. I went into the VM and could not even ping the gateway, however I could ping the loopback. This SQL server is extremely busy in comparison with the previous 4 VM's. I restarted the server and it came back up with the virtual nic working just fine.
The build of both servers (vmotioner and vmotionee) is ESX 4.0.0 175625 - So, pre-update 1.
Should I suspect the network switch/VM for possibly not updating the mac table on the switch?
Anybody else ever have this issue or know what may have caused it?
Thank you!
I would say you are on track with suspecting the mac table on the switch. Moving a machine across hosts can sometimes cause this. Look in your ESX vSwitch config for a "notify switches" checkbox. Check your switch's arp cache for the mac address.
There's only one way to know for certain if it's a host, vswitch or VM problem and that's by testing the migration methodically as the problem could be with any of those objects.
If you have the following hardware:
The solution is to decrease the ARP table refresh rate on the Dell MXL Foce10 switch by issuing this command: