Feat: create a net-install ISO

[bk201]

Problem:

To create a net-install ISO.

• Save time for downloading the installation medium.
• (hopefully) Prevent issues on buggy BMC ISO redirection feature: [BUG] Virtual Media Installation Hangs For 2+hrs With "containerd.sock" Connection Error harvester#2651

Solution:

We can ship a net-install ISO without image tarballs since we already harvester/harvester#2670.

Related Issue:
harvester/harvester#4295

Test plan:

• Build the code
• Boot the net-install ISO (Its name is similar to harvester-74e9450-dirty-amd64-net-install.iso) and verify the installation. Both interactive mode and ipxe mode should work.

Additional context

• Add a kernel parameter harvester.install.with_net_images=true to GRUB menuentry in the net-install ISO. The installer only preloads the harvester-cluste-repo image if it sees the parameter. We still preload this image to ease development.
• The kernel parameter harvester.install.with_net_images=true also works for the ordinary ISO.

[tserong]

Neat!

I did a test build here and got an 839MB net-install ISO, vs. the 5.6GB regular ISO, and it runs through the interactive install quite quickly. After reboot it took a bit over 30 minutes to complete setup, although this is of course presumably due to downloading all those images over my Tasmanian internet connection :-)

A couple of questions:

• Do we need to document the net-install ISO somewhere?
• Can this work in an airgap environment, e.g.: via a proxy server?
• Is there a good way to get better deployment progress than just looking at the console and waiting for it to show "Ready"? (I know I can ssh in and do things like journalctl -f and kubectl get pods -A ; kubectl get deployments -A, or run k9s... but I don't know how to tell exactly how far along we are...)

[bk201]

Thanks for the reviewing!

A couple of questions:

• Do we need to document the net-install ISO somewhere?

Good idea, I'll mention the change in the release note (there is a section ISO, e.g.: https://github.com/harvester/harvester/releases/tag/v1.2.1) that we have such an ISO.

• Can this work in an airgap environment, e.g.: via a proxy server?

The user can mirror all images to a private registry and:

• Create /etc/rancher/rke2/registries.yaml for RKE2
• Create /etc/rancher/common/registries.yaml for wharfie

I guess this might work, but need to verify.

• Is there a good way to get better deployment progress than just looking at the console and waiting for it to show "Ready"? (I know I can ssh in and do things like journalctl -f and kubectl get pods -A ; kubectl get deployments -A, or run k9s... but I don't know how to tell exactly how far along we are...)

That's one way except there are also daemon sets. Another way is to check the fleet bundles:

kubectl get bundles -A
NAMESPACE     NAME                         BUNDLEDEPLOYMENTS-READY   STATUS
fleet-local   fleet-agent-local            1/1
fleet-local   local-managed-system-agent   1/1
fleet-local   mcc-harvester                1/1
fleet-local   mcc-harvester-crd            1/1
fleet-local   mcc-rancher-logging-crd      1/1
fleet-local   mcc-rancher-monitoring-crd   1/1

Fleet keeps monitoring all resources in a helm are ready.

[Vicente-Cheng]

LGTM, tested with installation and basic operations work well.