Hello, I am looking for some information on setting up unattended installation of Harvester. I was looking at the Config parameters page on the docs site, while that's informative, there isn't enough to help with how does one use such a yaml file. Is it the standard way hosting the configuration in a tftp server for it to be picked up automatically? Are there any other unattented/automated ways of installing Harvester?

Typically you'd host the YAML file on a web server, not TFTP, then pass the URL into it during the netboot command sequence. We use iPXE over HTTPS to bootstrap to an unattended install. The YAML file is hosted on a "secrets" end point that auto-erases after the first access to deal with the fact that passwords are required to be stored in here for a true unattended install.

David, can you give please some pointers how did you achieve the erase-after-first-access behaviour ?

Thanks @full-night-53442. Has this been documented anywhere? Also, I am interested to know how the erase-after-first-access behavior as well.

Maybe use something like MAAS to deploy this? Or Redifish if your servers are new enough?

@glamorous-rainbow-20283 well all of those options will help get the installation started, however I am looking for a completely unattended to configure of Harvester and any future modifications to the cluster as well. In a way, I am building my own MaaS (in a manner of speaking).

Hmm, i still don't quite understand what would be different from MAAS here? You can pass the config from MAAS to harvester automatically as well?

How would a MaaS pass the config to an application packaged into an iso? That's what I am seeking here. It would help if you can elaborate here.

yeah thats only for loading the image, but you can set kernel paramaters in the MAAS settings and it will use that for the installation

Does harvester support that? Its important to note that Harvester is an application as opposed to an OS itself. So setting kernel parameters may not make sense.

Harvester can be installed by either booting the Harvester ISO, or via PXE Boot. When booting via ISO,

harvester-installer

runs interactively on the system console to allow you to configure the system. When booting via PXE, you don't get the interactive installer - instead you need to provide YAML files specifying the configuration to apply.

Looks like it! its confusing tho, could use a best practice or guide

Exactly why I am here...

after install you'd use the harvester terraform to set up the cluster itself i think

Harvester's terraform provider is to consume harvester cluster as you would consume any cloud. I don't see it helping configuring harvester cluster itself. Unless you have reason/information that helps otherwise.

There is no documentation out there that I'm aware of that will guide you end to end for harvester. I had to build the knowledge and systems to do this myself. Including building my own web application and iPXE image for the purposes of bootstraping into Harvester. As I said, it's really assuming you've got a complete system for managing this already.

Thanks @full-night-53442. I will start experimenting then! and get back here when I hit a roadblock.

Sorry if this is late :

@thousands-advantage-10804 Awesome! Thanks for sharing this.

If you hadn't seen, we used https://github.com/harvester/ipxe-examples/tree/main/vagrant-pxe-airgap-harvester for configuring the pxe boot server