Hi
I wonder if anyone can help.
We are building two new vSphere clusters running ESXi 7.0.1 17551050 and vCenter 7 Update 2. Our network has no native VLAN and we've been having issues getting host profiles to apply properly.
Basically the hosts boot and download an image from Auto Deploy. We achieved this by configuring the management VLAN ID in the PXE interface BIOS settings. We at first had an issue when after the image was deployed the host could not see vCenter as the default image is not sitting on the management VLAN. This was resolved by configuring a native VLAN (we did this as a test and have since removed) and was ultimately fixed by using the Set-DeployOption "vlan-id" option on the vCenter appliance using PowerCLI. So at this point we can get a host to boot from PXE (with the VLAN configuring on the PXE device) and download the image from Auto Deploy. By setting the above option Auto Deploy inserts the VLAN ID into the host image and it can see our vCenter.
Now when ESXi loads you can see it apply the host profile. I should point out that we are using DHCP to provide the IP address and other settings to the host with a reservation configured. However when ESXi loads and the host profile is applied it takes a long time on "vmware-fdm" at the end. What happens then is that the host has it's correct IP and you can view it in vCenter. After a few seconds it loses this DHCP IP and gives itself an automatic configuration IP. It is as if something changes during the final stages of the boot/host profile processes that causes the host to lose connection with the network. We have tried various things:
- Changing the order things are done i.e. setting host customisations and rebooting before running a remediation.
- Setting the management port group (which is on a distributed switch) to use ephemeral rather than static port binding.
- Tried no end of different settings/options within the host profile.
One way we did have it working was to configure the host profile so that it is set to use the MAC address from the PXE device for the management VMkernel. This seemed to work better. However, the issue with this is that we are caching the ESXi image so that the hosts can load in the event of a Auto Deploy outage. If you turn off PXE, so that the host loads from the cache, then this setting fails as the host did not PXE boot.
The VMware documentation is very poor on Auto Deploy with Host Profiles. For example, it says that the Set-DeployOption "vlan-id" option is ignored if host profiles are used but from our testing this is not the case. It definitely works better, and avoids a failure after the initial ESXi image is loaded, with this setting in place and using host profiles.
Anyone got any ideas/answers on this? The only other suggestion I have seen, and it would be good to know how many others are doing it, is not to include your management port group in the host profile i.e. Auto Deploy the host, manually configure the management port group and then apply the host profile.
Thanks