I have 3 HP c Class blades all running 3.01 (No hotfixes) and VC 2.01.
When I Vmotion a guest from one ESX server to another, I get (about 50% of the time) network timeouts if I am pinging the guest OS from an external host. Should I ping an external host from the guest then I don't lose any connectivity at all.
I'm thinking that this is a network problem...
Now, I am using 1 NIC for Vmotion, 1 NIC for SC and 2 NIC's for the VM Network. Each NIC port goes into a seperate HP c-Class GbE2c Switch And the switches connect back into a core switch (Foundry FESX448-PREM).
I suspected that this network connectivity drop-out occured because the destination ESX server for the VMotion decided to use a NIC connected to a different switch for that and the ARP table on the core switch is not being updated.
I seem to remember that to counter this problem VMWare Tools is supposed to ping the default gateway of the VM to ensure that the ARP tables get updated (certainly by running a manual ping the problem is solved). However this doesn't seem to be happening or it doesn't happen at the right time.
Has anyone else had this issue with similar gear? Especially with NIC's connected to different switches (There is no interconnect between them).
Also to prove this, I set all three ESX servers to use only the 1st NIC for the VM network (which are all connected to the same switch) and put the other into the 'Standby' group. No problem whatsoever. Migration happened and lost only 1 ping coming in from an external host.
The VMotion switches are set to notify the blade switches.
Other than running a constant ping on my guest OS's, does anyone know of a solution to this?
I've trawled the forums all day and come across nothing that would help me.
After rebooting all my ESX boxes, upgrading to VC Patch 2 and reinstalling VMWare Tools on the client machine, the problem seems to have gone away...
I can only assume that when VMWare Tools was originally installed the client machine had no network configured and didn't pick up the right settings.
I don't know for sure, but I suspect it was the client update. I find it hard to think of a reason that the VC update should have had any effect at all, but then I'm no expert! But in case it's relevant, as part of my VC patch I accidentally recreated my VC database.
Good Luck in solving the issue! If you manage to narrow down the solution, I'd love to know.