I've been running ESXi 5.5 for about a year now with four VMs. I needed to add two more, so I added additional memory on my host server and configured the two VMs. Both VMs are Server 2013 R2. One is running a new database - the other a web site. Since adding the two VMs, I've been getting the PSOD. If I leave the two new VMs off, I don't get the crash. I have tried removing the new memory, but the problem persists. Is it possible for my memory or processor allocation to be causing this? I'm using the free version of ESXi. The PSOD screen is below.
I've been working with Dell on this. They instructed to update all the firmware on the server, but this has not resolved the problem. They are unable to diagnose further.
The output looks alot like the E1000 bug described here: VMware KB: ESXi 5.x host experiences a purple diagnostic screen mentioning E1000PollRxRing and E1...
/Rubeck
I am using a single Broadcom NetXtreme BCM5720 Gigabit Ethernet Adapter. On VM it lists the adapter as Intel Pro 1000 MT network connection. Why the difference?
This is a Known problem, you need to change the E1000 to VMXnet3 or e1000e. The fix is expected to be in next month. ie 5.5 update 2
Thanks,
Avinash
The two original VMs were set at the E1000 adapter. All the other, newer, VMs were set for E1000e. The older VMs were configured at the original installation of ESXi 5.0. I have since upgraded to 5.5. I also cannot choose E1000e from the dropdown list of available adapters on the older VMs. Only the VMXnet3 is available. I'm assuming I still need to choose the VMXnet3 adapter for those two XMs and that this won't cause any additional problems because they are using VMXnet3 and the other VMs are using E1000e..
yes correct, you need to choose the VMXnet3, e1000e also have known to cause this situation and you may need to disable RSS inside the guest if your using the e1000e.
What is RSS. I'm not finding it in the settings of the adapter or the documentation. Thanks!
Never mind. I got it. Would it be better to put all my VMs at VMXnet3 for uniformity purposes?
To verify this in a Windows guest operating system:
Thanks,
Avinash
Yes changing the adapters to VMXnet3 would be the best.
Changing all my server network adapters to VMXnet3 has apparently fixed my problem. No PSOD for over a week will all servers running.
This issue occurs due to the ring buffer for e1000 NIC card gets full. changing NIC cards to VMXNET3 will give good performance and no purple screen
i never would use other Nic then vmxnet. Because for the Most OS this is the recommended Setting. (Reasons are High CPU Overhead,......)
you could check for each OS the Compatibility Matrix. example windows 2012 R2
VMware Compatibility Guide: Guest/Host Search
i would only change away from vmxnet3 in Case using PXE (maybe theire still problems with it).
PS: have also a Look if you could change you're Power Settings to "High Performance". the Dynamic Process also not very good and could cause Delays,.... . in an Critical Latency important Environment > theire should be Maximum Performance available.
Best regards
Max