How to manage the life cycle of installed Helm charts?

[otavio]

There are two prominent examples in the project which demand long-term maintenance such as:

1. Ingress controller
2. Longhorn operator

How would upgrades for those be applied in an existing cluster?

[mysticaltech]

@otavio Both have HelmChart and HelmChartConfig definitions, you can view them with kubectl get helmchart(config) -A (or just use kubectl get crds if I got the orthograph wrong).

You can upgrade them by applying a HelmChartConfig; see examples in the templates folder. And more on the matter here: https://rancher.com/docs/k3s/latest/en/helm/.

Also, it seems possible to directly upgrade with the helm cli, but GitOps principles would dictate the former method.

Lastly, we export a kustomization_backup.yaml and save it at the root of the folder; this will also include all the generated HelmCharts and HelmChartConfigs and can be used to upgrade too, but changing the chart versions. Please see the upgrade section in the Readme.

[mysticaltech]

I am not that familiar with Flux @otavio; just know that it's close to ArgoCD, but it could be the case, so you can disable the above add-ons to install them freshly with flux after the cluster is deployed!

For ingresses:

enable_traefik = false
enable_nginx = false

For Longhorn:

enable_longhorn = false
extra_packages_to_install = ["open-iscsi", "nfs-client", "xfsprogs"]

The extra packages above are needed for Longhorn. I hope this helps! :)

[mysticaltech]

Our Longhorn config does additional steps like creating a /var/longhorn folder on the node or creating volumes and mounting those to /var/longhorn when the longhorn_volume_size attribute is present on the agent nodepool definition.

So if you want to take advantage of that, just keep enable_longhorn = true. The needed package will be installed automatically, and the proper node-level config will be done too. But on flux's side, you can start by deleting the HelmChart corresponding to Longhorn, leaving you with a fresh Longhorn-ready cluster!

Up to you, really, or you could have a separate TF module or Ansible script do the Longhorn prep work; that would work too.

[otavio]

We are preparing a testing deployment doing the installation using Helm and adding the extra packages as you suggested.

[mysticaltech]

Wonderful! If you manage to get this to work and are able to share, a PR adding to the examples folder, and a little note on the examples section in the readme, would be very much appreciated 🙏

[mrclrchtr]

An info for others, so they don't stumble over it like me: If you turn off longhorn the iscsid service is not activated anymore:
inline = concat(var.enable_longhorn ? ["systemctl enable --now iscsid"] : [], [

And

enable_traefik = false
enable_nginx = false

do not exist anymore.
Its now:
ingress_controller = "none"