vNic placement UEFI boot

[ericstottelaar]

Hi,

I have an issue with the vnic placement, when i changed the boot policy for native to uefi boot mode i noticed that the vnic order in esxi has changed for vnic1 and vnic2. This looks strange to me because i also attached the host interface placement policy to the service template which would correct the vnic placement but it does not work.

Does someone have an idea where i can change the vnic order for uefi?

[Riaan van Niekerk]

Hi @ericstottelaar . This is the known and expected behaviour because PCI reordering takes place when changing from Legacy BIOS to UEFI on some server models. See Broadcom KB: How VMware ESXi determines the order in which names are assigned to devices which also has some examples of renaming vmnics using the localcli command. There is no known way of tweaking the host interface placement policies to change the device (vNICs & vHBAs) order to what you had on Legacy BIOS. 

Even though only UEFI is supported with newer UCS firmware versions on M5 & M6 (older 4.1.3 firmware on M5 did support Legacy BIOS), from our experience, there is very low to no risk to staying on Legacy BIOS. We have upgraded dozens of hosts on Legacy BIOS from ESXi 7.0 U3 to ESXi 8.0 U3 and none of them have become unbootable or had stability issues (two of the possible adverse effects of staying on Legacy BIOS). Even though we have the commands to bulk rename vNICs, due to the number of hosts that we had to upgrade, we decided to leave our M5s on Legacy BIOS. Switching M6s to UEFI did not cause device reordering for us. 

Cisco UCS KB article: Cisco UCS vNIC/vHBA Placement : Issue with desired order in VMware ESXi discusses the issue in detail. 

[jn242]

Same Issue.

Vmware/Broadom states they will not support you if you are on Legacy and you have to switch to UEFI. 

Contacted TAC to confirm the process to switch over to UEFI and swtched over boot policies from Legacy to UEFI.  Boots up fine but NICs and network traffic is impacted.

I thought it was a result of upgrading to 8 but it isnt. Its UCS from what i can see. TAC later sent me a bug article confirming  and explaining it. There is no fix. There is a workaround of remapping all our vmnics to former mac addresses via pci slot addresses for each vnic but im not sure i understand it. I may give this a try or i may just reload everything from scratch and start over for all of our hosts. I dont know which woudl be quicker. The output of the commands in esxi appear to show you all your vnics and your vhbas. I dont want to fat finger something and start having storage problems on top of it.  The bug article (and the broadcom article) are not clear on how to find what your slot ids are for remapping. I need to study it more. I think whatever it did, it did it the same for all the nics across the hosts so the commands should be almost the same for each hosts vnic mapping. 

Other than fighting this, the option to just reload and start over is also there. Clearly not ideal 

[Riaan van Niekerk]

You can ignore the VHBAs reordering. It does not impact anything, other than being cosmetic.