we recently set up an Eaton UPS (5PX) and installed the "IPP" agents (Intelligent Power Protector) on our machines to provide "graceful shutdown" in case of power failure. Connected to the UPS are four ESXi hosts. Everything just worked fine but we are getting spammed with alert notifications of the IPPs: "Communication lost".
Strange thing is: All in all we have 17 VMs (and one IPP per VM) on 4 ESXi hosts and all the IPPs of VMs on ONE particular ESXi host report communication failure and recovery (over and over).
All these VMs have in common is their ESXi host.
We have checked anything network related: Pinged the VMs, pinged the UPS from the VMs and not one packet was lost. We also checked firewall/routing protocols for any rejected packets (Port 4679/4680).
Networking configurations seem identical on our servers, but we found out that this particular server has a different networkcontroller: "Intel 82575EB".
All the other have "Intel 82576" or "Supermicro 80003ES2LAN".*
I compared Intel's chipsets an found out that the 82575EB doesn't have the "VMDc" feature - Virtual Machine Direct Connect, direct networking I/O between NIC and VM:
Could that be it? I couldn't find out about Supermicro's controller but it's pretty much the only clue we have got...
Do you have any recommendation on how to improve networking I/O on this ESXi or how to work out a solution? (We tried Eaton support but were told that the free version ESXi wasn't supported..)
We are looking forward to any suggestions on this!
*We use manufactory-build (VMware certified Hardware) and self-made servers (at least VMware ready) with ESXi 5.0.0., 469512.