VMware ESXi 5.x host experiences a purple diagnostic screen Workarounds

VMware issued a Knowledge base article detailing a known issue with ESXI Hypervisors failing due to virtual machines using the Virtual Network Adapter (KB2059053).

The ESXI host will show the following symptoms:

Through the DCUI you will see a Purple Diagnostic Screen showing the following information:

@BlueScreen: #PF Exception 14 in world wwww:WorldName IP 0xnnnnnnnn addr 0x0 PTEs:0xnnnnnnnn;0xnnnnnnnn;0x0; – This indicates the virtual machine via WorldName 0xnnnnnnnn:[0xnnnnnnnn]E1000PollRxRing@vmkernel#nover+0xdb9 – This indicates the issue with E1000 Virtual Network Adapter

0xnnnnnnnn:[0xnnnnnnnn]E1000DevRx@vmkernel#nover+0x18a – This indicates the issue with E1000 Virtual Network Adapter 0xnnnnnnnn:[0xnnnnnnnn]IOChain_Resume@vmkernel#nover+0x247 – This indicates the issue with E1000 Virtual Network Adapter 0xnnnnnnnn:[0xnnnnnnnn]PortOutput@vmkernel#nover+0xe3 – This indicates the issue with E1000 Virtual Network Adapter 0xnnnnnnnn:[0xnnnnnnnn]EtherswitchForwardLeafPortsQuick@#+0xd6 0xnnnnnnnn:[0xnnnnnnnn]EtherswitchPortDispatch@#+0x13bb 0xnnnnnnnn:[0xnnnnnnnn]Port_InputResume@vmkernel#nover+0x146 0xnnnnnnnn:[0xnnnnnnnn]Port_Input_Committed@vmkernel#nover+0x29 0xnnnnnnnn:[0xnnnnnnnn]E1000DevAsyncTx@vmkernel#nover+0x190 0xnnnnnnnn:[0xnnnnnnnn]NetWorldletPerVMCB@vmkernel#nover+0xae 0xnnnnnnnn:[0xnnnnnnnn]WorldletProcessQueue@vmkernel#nover+0x486

The Current workaround provided by VMware:

• Use the VMXNET3 virtual adapter
• Disable (Receive-Side Scaling State) on the Windows OS

Disabling (Receive-Side Scaling State) on Windows OS

Use the following Command to check to see if “Receive-Side Scaling State” (RSS) is enabled (usually enabled by default)

Netsh interface tcp show global

To Disable (RSS) use the following Command in (CMD):

Netsh interface tcp set global rss= disabled

Locating affected Virtual Machines:

To locate the effected Virtual Machines you would need to use the following PowerCLI script I have written.

# VMware ESXI Script to show Virtual Machines using E1000/e Virtual Network Adapters

# Created by Ryan Mangan on the 17/04/2014

param (

[Parameter(Mandatory=$TRUE, HelpMessage=”Enter the name of the ESXI Host FQDN”)]

[String]

$VMHost,

[Parameter(Mandatory=$TRUE, HelpMessage=”Host Root Password Dual control”)]

[String]

$HostPW,

[Parameter(Mandatory=$TRUE, HelpMessage=”Location of exported .txt output file”)]

[String]

$exportfile

)

Add-PSSnapin VMware.VimAutomation.Core

ForEach( $VM in (Get-VM) ) { $VM|Where{ $VM|Get-NetworkAdapter|Where{ $_.ExtensionData -like “*e1000*” } } }

Out-File $exportfile

 

This script would need to be run on a client or vCenter server which has PowerCLI installed.

The issue is resolved in the following updates:

• ESXi 5.5 Update 1, see  VMware ESXi 5.5, Patch ESXi550-Update01: ESXi 5.5 Complete Update 1 (2065832).
• ESXi 5.1 Update 2,  see VMware ESXi 5.1, Patch ESXi510-Update02: ESXi 5.1 Complete Update 2 (2062314).
• ESXi 5.0 Patch ESXi500-201401001, see VMware ESXi 5.0, Patch Release ESXi500-201401001 (2065814).

Knowledge articles:

VMware ESXi 5.x host experiences a purple diagnostic screen mentioning E1000PollRxRing and E1000DevRx (2059053)

Poor network performance or high network latency on Windows virtual machines (2008925)