feat(argo-workflows): Switch to full CRDs for workflows

[Joibel]

This is mainly a PR to change the default value of full CRDs to full, to match the expected release of workflows v4.0.0 shortly. Mostly this PR is about the values.yaml README.md change - the rest would happen via renovate I believe.

The CRDs have been pulled from v4.0.0-rc4 which will also be tagged v4.0.0 at release - I'm happy to separate this from the auto-generated renovate PR if you'd prefer.

This is the continuation of the #3559 PR.

Checklist:

• I have bumped the chart version according to versioning
• I have updated the documentation according to documentation
• I have updated the chart changelog with all the changes that come with this pull request according to changelog.
• Any new values are backwards compatible and/or have sensible default.
• I have signed off all my commits as required by DCO.
• I have created a separate pull request for each chart according to pull requests
• My build is green (troubleshooting builds).

[tico24]

Thanks. Looks good to me. Obviously can't test and merge until 4.0.0 comes out of the oven upstream.

[vladlosev]

I am quite concerned about making full CRDs the default. They take 7.5MB, and on many systems, will take a long time to deploy and slow down the Kubernetes API server in the process. In our system, we use minimal CRDs because using the full ones makes a single deploy so long, it trips the CD operator timeouts. For many users in similar circumstances, it will be a breaking chage.

[mkilchhofer]

I am quite concerned about making full CRDs the default. They take 7.5MB (..)

I am also pretty sure that the full CRDs will not work when the Helm chart is installed via a regular helm install/upgrade-approach or with FluxCD.

For tooling like Argo CD, it will work because Argo only executes helm template.

Our CI test uses the chart-testing action which make use of helm install.

[Joibel]

Upstream the full CRDs are the default in 4.0, and I'd like to deprecate the minimized CRDs in future releases.

How do other projects with big CRDs handle this?

[mkilchhofer]

Upstream the full CRDs are the default in 4.0, and I'd like to deprecate the minimized CRDs in future releases.

How do other projects with big CRDs handle this?

Tbh, I never saw such big CRDs in the past.
I once saw a discussion going on for the Kube-Prometheus-Stack chart in the Prometheus Community org:

• [kube-prometheus-stack] Long term fix for kube-prometheus-stack release size issues prometheus-community/helm-charts#3548

Some operators/controllers update the CRDs as "business logic" when the operator starts up. (Cilium for example).
And I also saw helm charts using a Job with a pre-install hook to update the CRDs.