There was a recent thread in the VMTN community forums where a user wanted to provide user input prior to an automated ESXi installation. This may sound odd at first, especially when we are talking about an unattended installation, the last thing you want is any type of user interaction. The reason for this requirement was to maintain a generic kickstart configuration file and based on the site specific input (IP Address, Netmask, Gateway, Hostname and DNS Servers), the ESXi installation would be able to deploy and configure itself the same across multiple locations.
The other reason for this requirement is that all hosts in the user’s environment must be configured with a static IP Address, this is not an uncommon requirement for many production environments to not have DHCP enabled networks. I can only assume the initial network the host is being built is either a private build network or booting off of local media such as a USB or CD-ROM device.
Whether you are booting off of local media or via PXE over the network, you have the ability to specify some boot parameters which includes things like IP Address, Netmask, Gateway and DNS Servers. These “bootstrap” options are fully supported and documented in the ESXi Installation Section. You can also specify advanced VMkernel boot parameters which may not be officially supported by VMware, but you can take a look here for more details.The concept of specifying these boot parameters is nothing new and has been supported since the early days of classic ESX and other popular UNIX/Linux distros for PXE boot/installations.
When you boot the ESXi installer, you may have seen a screen similar to the following in which you have a few seconds to hit the “tab” key to edit the boot options.
Once you hit the “tab” key, you will be able to see what the default boot options are and if you are PXE booting, you will also see some IP information appended towards the end of the string. This is where you can append or update additional parameters and later read in by your kickstart script.
Here is an example of an ESXi installation being PXE booted over the network and I have added 4 supported boot parameters and 2 custom ones.
If you take a look at the screenshot, the “+++” IP information towards the end is what was given by the DHCP server but I am interested in specifying a different IP Address for the ESXi installer to boot from. I provided new entries for ip, netmask, gateway and nameserver. You can also see that I introduced two new variables called hostname and dc, these will be used to specify the hostname of the ESXi host and also the name of the datacenter which will be used later in the kickstart script to rename a datastore.
Note: If you redefine the IP information, you do not need to change the IP information found after the “+++”, these will just be over-written with the new IP information.
When specifying these boot parameters, you need to make sure it is after the “vmkboot.gz” but before “— vmkernel.gz”. To help make this clear, I have colorized the section that was appended to the default options.
vmkboot.gz dc=ghettoDC hostname=vesxi41-2.primp-industries.com ip=172.30.0.200 netmask=255.255.255.0 gateway=172.30.0.1 nameserver=172.30.0.100 ks=http://172.30.0.108/esxi41u1/ks.cfg — vmkernel.gz — sys.vgz — cim.vgz — ienviron.vgz — install.vgz
Note: You can create any custom variables, the key in using the custom variables will be parsing from the boot command line using vsish which will be defined in your kickstart script. This will work also work for local media, but you MUST use a kickstart to perform the actual installation or inject the kickstart script into a custom ISO.
After you have provided all the input, you will then boot the installer and the the following kickstart configuration file is used to parse the boot options using vsish.
The script does the following:
- Reads the boot options using vsish and temporarily stores the output to /tmp/cmdline for later use
- Extracts all the relevant boot parameters in /tmp/cmdline and stores them in variables to be used in the script.
- Creates the “network” stanza for static IP assignment of the ESXi host which includes: Hostname, IP Address, Netmask, Gateway and Nameserver
- Writing out the custom variable “dc” into /var/log/esxi_install.log because this will be persisted through the reboot and can be later read in for any %firstboot operations.
- Upon the reboot, %firstboot will execute and read from /var/log/esxi_install.log to extract the “dc” variable in which it will use to rename the local datastore
Note: The above is just an example of what you can do with custom parameters. You can easily add as many as you need for site specific configurations and then use those input in your post configure your ESXi host. Also note that you do not need to make sure of ip,netmask,gateway,dns parameters to use custom variables, these are all optional.
You can add as many custom entries as you would like, but this can easily get error prone due to the amount of typing required. One thing you can do to minimize the amount of typos is to pre-specify the custom variables in your PXE/TFTP configuration file.
To do so, your PXE/TFTP configuration file would look something like this:
This not only reduces the amount of typing but also lets the user know what variables must be defined in order for the installation to proceed.
I totally understand where the user is coming from and I have had this requirement in the past but I think having a static DHCP entry and maintaining a simple configuration file can easily solve this problem. This not only makes the deployment process hands-off as it should be in the first place but also removing the human factor out of the equation. Can you imagine deploying 100-300 ESXi hosts in a day? I sure can not if I had to manually type out all those addresses by hand.