automatic vm migration on vswitch fail

[aralestrip]

Hi,

I have 4 nics in the LAN network. I successfully tested the failure of one or more of them (by shutting down the ports on the physical switch). since the nic teaming is set, no lost of even a ping in these cases.

but if I make to fail them all, the fallback becomes useless. the virtual machines that reside on this esx just aren't accessible anymore. the windows doesn't see any network card failure (or unplugged), the esx cluster doesn't take any action - the lan vswitch failure is no reason for HA actions.

so, how can you protect the virtual machines from network loss? is there any specific funcionality?

i forgot. the esx is part of a cluster whre drs and ha are enabled.

[weinstein5 · Accepted solution]

If you have all ports of a vSwitch connected to a single physical switch and that switch were to fail there is really nothing that the vSPhere environment to do - this one reason why redundant switches are critical in your design

[akshunj]

In addition to redundant physical switches, where possible divide your physical NICs between network cards in your server. For example, if your LOM card was dual port, and you also had a quad-port PCIe NIC, use two interfaces from each card to prevent a card failure from taking your LAN vSwitch down completely.

[rickardnobel]

Unfortunately there is not much else to do but what is stated above: make sure to have physical switch redundancy and if possible make sure a single physical network card will not be a single point of failure for a vSwitch.

It would be interesting to have the ability for the vSwitch to take some kind of action if all uplinks are lost, but as the vSwitch emulates a quite simple standard based Ethernet switch this is not available today.

[aralestrip]

thank you for the answers.

i found an option in the esx cluster: vm monitoring. as I searched in the documentation, maybe this feature could help. no vm hb received, make a vm reboot. i tested, but I couldn't make this feature take any action. i just enabled the vm monitoring, so after a 30seconds lost of network it should reset the vm - as I understand the manual. but this did not happen.

do you have any experience about this?

[rickardnobel]

That is unfortunately not the way it works. The VM monitoring is only to detect a total internal VM failure, such as a blue screen or similar. It will not help to detect anything network related.

[aralestrip]

thank you again.

I hope vmware'll create a network cluster too, in one of the next releases. anyone heard if they're already working on this?

[weinstein5]

But if you are still relying on a single physical switch you will still lose network connectivity even if it fails over to another host -