AutoDeploy and Host Profiles

[mars0077]

Hello,

I'm planning on deploying vSphere 8 using AutoDeploy and Host Profiles for a stateful installation. So far the autodeploy piece works as expected as I am able to PXE boot my new host and download the correct image from the autodeploy. Once the image is downloaded and installed, I can see the new host show up in my vCenter inventory under maintenance mode.

When I remediated the freshly installed host for the first time, I noticed that although the hostname and IP changed as a result of the host customization settings which is expected but soon after the host lost connectivity and became unmanageable. The reason was due to the vmk0 interface being removed by the remediation.

I am not sure if this strange issue is a bug or a host profile configuration issue. Every time I try to apply the host profile that I created from a working host,  I noticed that one of the steps included is to remove the vmk0 adapter from the management network which basically disconnects the host and the remediation never completes due to this.

I have disabled the default gateway override for the vmk0 interface from the host I am using to extract the image from. I have also disabled IPV6 all together.  

Here are the steps I am seeing on the pre-check list:

The system will perform the following tasks:
Remove vSwitch vSwitch0
Create vSwitch vSwitch0
Create vSwitch vSwitch1
Remove port group VM Network from vSwitch vSwitch0
Remove port group Management Network from vSwitch vSwitch0
Create port group LAB_MGMT on vSwitch vSwitch1
Create port group LAB_VMS on vSwitch vSwitch1
Create port group Management Network on vSwitch vSwitch0
Create virtual network adapter for host on port group Management Network with IP address x.x.x.x and subnet mask x.x.x.x.
Remove virtual network adapter vmk0 from portgroup Management Network.
Configure physical network adapter vmnic4 to automatically negotiate the proper settings.
Configure physical network adapter vmnic6 to automatically negotiate the proper settings.
Configure time servers to be 0.north-america.pool.ntp.org,ntp1.goempyrean.com
Update the firewall configuration
Set option UserVars.HostClientCEIPOptIn with value 1
Set option Syslog.loggers.amsdv.rotate to its default value
Set option Syslog.loggers.amsdv.size to its default value
Set option Syslog.loggers.sut.rotate to its default value
Set option Syslog.loggers.sut.size to its default value
Setting the startup policy for service TSM to on
Setting the startup policy for service TSM-SSH to on
Setting the startup policy for service ntpd to on

Is there any other settings I need to change in order to remove or avoid the vmk0 removal as part of the remediation? The version of vCenter is 8 and the host I am using to extract the image from is also 8. Thanks!

[mars0077]

I ended up doing away with this setup as host profiles will be replaced by vSphere configuration profiles. What I did instead is rely on UEFI HTTP booting capabilities and was able to configure the host to a desired state by leveraging the time tested kickstart script.

[OsburnM]

Would you mind posting a little more detail on your solution?  I'm in the same boat.  Have been using AutoDeploy for many years but the struggle with going from Stateless to Statefull, along with converting to Single Image vLCM & using Host Profiles has just gotten to where it's not longer efficient.  Spending way more time troubleshooting autodeploy & host profiles required for autodeploy then if I just manually installed hosts.  Would like to look at either PXE or HTTP scripted installs but haven't been successful on the PXE side yet.  Struggling with the structure on the TFTP server.  My hosts are all modern HPE gear, so UEFI HTTP should also be doable but haven't made much progress there either.  Very interested in your solution if you don't mind.

[mars0077]

Hey Osburn,

I hear you on the AutoDeploy struggles. I have always been interested in the solution from it's inception but in the end, I always ended up frustrated and disappointed due to the technical challenges and limitations it has. You're also right about the time spent in troubleshooting it. Honestly, I wouldn't feel safe with it in a production environment at this point. 

In regards to remote installations, what has worked well for me at this point is relying on a kickstart script for my initial config and HTTP booting the hosts as long as you have a server that supports UEFI boot mode. Under the UEFI HTTP boot settings/Network Settings,  you should be able to configure which NIC you want to use for the initial boot.  You should also be able to specify if you want to use a static or DHCP IP for your IP settings.

I'm not sure if you're familiar with how to setup a kickstart environment but if you need help in this area then let me know. Basically, you need to centralize your ESXi custom ISO and configure a web server of your choice. The web server will be used for the HTTP URI or KS destination. I'm using Abyss Web Server as it is simple and light weight. 

#This is the UFEI BIOS method in the BIOS/Network settings
http://x.x.x.x/ESXi-Dell-8.0-x64/efi/boot/bootx64.efi --This provides the initial boot instructions.

#This is the CLI method if you don't care for the HTTP boot. You would kick off the ESXi install and press shift + O #Here I am specifying which vmnic and protocol to use to boot. 

netdevice=vmnic3 bootproto=dhcp ks=http://x.x.x.x/Kickstart/kickstart.cfg

#KB on the above instructions. 

https://docs.vmware.com/en/VMware-vSphere/7.0/com.vmware.esxi.upgrade.doc/GUID-A33C4CAB-18EE-4C8C-A5...

Let me know...

[OsburnM]

Appreciate the response.  Yeah Im in the same boat.  May be time to say goodbye to AutoDeploy.  Then again, with 8U1's host profile changes to json & clusters switching to desired state config, it may actually be more stable.  Will be interested to see when it gets released.

As far as the HTTP UEFI--  I think Im good now.  I battled with our Bluecat DHCP for a while until I realized I didnt even need to bother as our HPE gear could do it natively w/o any DHCP customization by simply adding the mboot.efi url as an HTTP Boot Option in the profile.  Works like a charm.

How (if you know) are folks getting kickstart customizations done if they have secureboot enabled?  Looks like nothing in %firstboot will run with secureboot.

Thanks again,

[EricJohnson2]

Try not removing vswitch0 at all in your host profile.  It will leave you with an empty standard switch but I think it will solve your problem. 

Removing it kills the vmk0 in my experience and you're toast. 

Autodeploy using DHCP for all your vmks kicks butt.  No host customizations and the thing just comes online.  Nice.