I figured the last two weeks a strange, reproducable issue with following constellation:
Hardware (tested with both Servers)
HP ProLiant ML110G7, SmartArray P410, 1GB FBWC, RAID5, 32GB HP ECC RAM
HP ProLiant ML310e Gen8, SmartArray P222, 512MB FBWC, RAID1, 16GB HP ECC SmartMemory RAM
Hypervisor (tested vice-versa on both Servers)
ESXi 5.1U1 HP Custom Image VMware-ESXi-5.1.0-Update1-1065491-HP-5.50.26.iso
ESXi 5.5 HP Custom Image VMware-ESXi-5.5.0-1331820-HP-5.70.38-Sep2013.iso
When I install a Windows 8.1 or Server 2012 R2 (NOT Windows 8.0 or Server 2012 w/o R2) with the suggested NIC E1000E, and I have a little bit "heavy" Network-Traffic from/to Windows 2012 or Windows 2012R2 Servers, an PSOD occurs with "pf exception 14 in world 107417".
Unfortunally, I didn't made a dump or Print-Screen of the issue, but I can write you what I've done (or make a little bit of load and then wait until it crashes again...)
Network-Installation of "Windows 8.1 Pro" or "Windows Server 2012 R2" from an Windows Deployment Service wich resides on Server 2012.
WDS Server is VM on same ESXi Host like the new VM. WDS boots, made Domain-Login for selecting the Images, started Copying --> PSOD.
Reproducable only with 8.1 or 2012 R2 with E1000E NIC
Network-Installation of "Windows 8.1 Pro" or "Windows Server 2012 R2" from an Windows Deployment Service, resides on Server 2012.
WDS Server is VM on different ESXi Host like the new VM. WDS boots, made Domain-Login for the Images, Started Copying --> ALL OK.
Then installed on Host 2 (2 x Server 2012 R2 and 1 x Windows 8.1 Pro) everything for a Domain (AD, DNS, WDS, WSUS) and when I first want to download the Windows-Update from the new Windows 8.1 Machine --> PSOD again.
Network-Installation of "Windows 8.1 Pro" or "Windows Server 2012 R2" from an Windows Deployment Service, resides on Server 2012 (VM).
WDS Server is VM on ESXi Host, WDS Installation on Physical PC, not VM.
WDS boots, made Domain-Login for the Images, Started Copying --> ALL OK, no PSD at all.
After discussing this issue with a collegue he said that he has similar Problems with bigger DL380 and Server 2012 R2 (wich were unavaiable a couple of pings) and he told me that he loaded VMXNET3 Driver.
I've done this too and until now, no new PSOD was here.
But the silly thing is that VMXNET3 is unable to install via WDS, so therefore I have to install my Test-Machines the "old" way with mounting ISOs.
My question to you - had anybody else this issue or am I maybe the only one?
TIA - Josip.
does this also happen with the older E1000 instead of the E1000E ?
Yes and No - as with E1000 Driver (without E at the end), I've not received until now an PSOD.
But I've not done stress-tests with the NIC, I just was able to fire up the VM an install Windows 8.1 without PSOD
Well I have the exact same issue here!
I have updated from ESXi5.0 to 5.5 so I could install Win8 + Server 2012 VM's but the host consistently purple screens during Windows 8.1 - roaming profile logon...
SuperMicro board + Xeon / 32gb ram
Windows 2008 R2 VM with Dell Perc6i passed through directly to the VM (2x VMXNET3 NIC's / 1 on a dedicated vLAN for iSCSI / NFS)
iSCSI LUN's provisioned through iSCSI for Windows and presented back to ESXi
Windows 8 VM - installed on iSCSI LUN
XBMC with radeon GPU / pci pass-through
Windows 8.1 VM currently has an E1000E NIC assigned
Another issue ESXi 5.5 introduced with my setup was the USB controller pass-through failed after the upgrade.
I get the PSOD too.
I have a setup that consists of 2012 servers and windows 8.1. WSUS is setup on the 2012 server and group policy is setup to update all computers on the domain. As soon as the windows 8.1 client attempts to pull an update from the WSUS server.....bam....it's all purple.
I'll try the change of driver now though and let you know how I get on.
i can confirm there is a correlation between PXE imaging with windows 2012 R2. Last week i have been building a LAB for the Win2012 R2 including 4 domain controllers, 2 SQL 2012 databases in a cluster and 2 SCCM servers CAS+Primary site. Once i configured PXE and started imaging the system crashes PSOD. As all of the other VM's were deployed using an ISO it has definitely something to do with PXE and associated network transport layer related components.
Will continue to do some troubleshooting around this and post my comments later on.
This problem is already fixed for 5.1 in 5.1 Update 2 that is already available for download and for 5.5 update 1 which is expected to be fixed in coming month or this quarter.
The Current workarounds available for the same are given in below KB
http://kb.vmware.com/kb/2059053 < ESXi 5.x host experiences a purple diagnostic screen with errors for E1000PollRxRing and E1000DevRx >
HAve the same problem of 5.5 crashing wenn tryning to push image.
So i tried to swap to vmxnet3 adapter.
Did it two ways (machine shut dwon and tehne booted up again):
Manually editing in the .vmx file e1000 to vmxnet3 and then i tried by removing e1000 adapters and adding vmxnet 3 adapters.
MAchien boots up but, static ips are assigned, as i can see using ipconfig but i cant ping out or get pinged.
Changing back to e1000 and everythings working agaien, escapt WDS of course..
Are there any chnges to be made to the vswitch or somethin else to be done to get the vmxnet3 adaptesrs to work (vm hardware version 7)?
If you have a Windows OS, you should first remove the E1000 via properties and after this add an VMXNET3 Card.
Please double-check that the cards are on the same internal Network.
This works even if your VM is running, so no shut-down nor manual editing of VMX Files...
Are you sure you assigned the proper named network when you added the vmxnet3 adapter? By default it grabs the first one (alphanumerically) in the list and if the VM operates on a specific network you need the one valid for the Guest and it's IP address. If it is a Windows guest, you might also want to go back and eliminate the "phantom" adapter from the OS; I used the method below to clean up those devices and their drivers that no longer exist:
Thanks for your replys.
Yes I took care for using the right network and checked at the vswitch.
Actually i have 2 vswitches connected to the virtual server.
I will try to remove the networkdapater first lke you suggested, but when I checked the nics in the deveice manager the used driver was vmxnet.
Is there any possibility to see in the logs if therer is something wrong?
Is there maybe an issue with the hardware version 7? Made an upgrade from ESXi 4.1 to 5.5 about 2 weeks ago and everything was working fine until i started to install a new machien by wds.