facilitate the use of Kustomization healthChecks on Helm-based units (and use it with Rancher)

Thomas Morinrequested to merge
facilitate-healthChecks into main

Closes #102 (closed)

This MR:

  • does the change proposed in #102 (closed)

  • leverages it to improve how we "wait for Rancher", to compensate the fact that additionally to the resources directly installed by rancher Helm release, the rancher Deployment pods create tons of other things, on that other units actually need

    • By waiting on the Cluster.provisioning.cattle.io CRD, we ensure that capi-rancher-import (which needs it) can avoid a longish period of CrashLoopBackoff. I opted to not list all CRDs, because I'm assuming that they should roughly be created at the same time.

    • By waiting on the rancher-webhook Deployment and Service we ensure that we don't get an unreliable tests (in the past we had a case of successful creation of a resource when done prior to rancher-webhook being in place, but failing if created afterwards - see #3 (closed)).

    • I'm also optimistic than just waiting on more things helps having first-rancher-login spend less time in retries, and has less risks of reaching the maximum number of Job retries (this MR seems to have less test-kubeadm-capd failure that what we see these days on main)

Edited by Thomas Morin

Merge request reports