Resolve "ArgoCD cypress tests looping" | SKIP UPDATE CHECK

General MR

Summary

• 🏗 Moved helm tests for application create and delete from Cypress-based testing to shell-script based testing.
  • This will allow us to more quickly and reliably confirm that Argo's key featureset is working post-deploy.
  • Argo's UI changes its own URL with nearly every keystroke in the "create new application" form and this has proven very difficult to get Cypress to play along with.
• ✨ added a tests/wait.sh to confirm the argo CRDs are established and that the argocd web server deployment rollout is complete before proceeding to our helm tests.
  • ⚠ this script makes heavy use of curl retries ♻ as the argocd-server web service has been refusing connections left and right in CI despite being pretty good at eventually fulfilling requests.
    • 🔬 See this CI job log with 60 something "connection refused" log lines that nevertheless ran to completion in six minutes thanks to curl retries.
• 🔧 added a top-priority Service at chart/tests/templates that sits in front of argocd-argocd-server for the duration of our helm tests.
  • For some reason Argo's usual Service object isn't there during the current pipeline-templates CI flow's helm test runs so I'm adding this short-lived service as we have already done over in the gitlab package at chart/templates/tests.

Relevant logs/screenshots

Successful wait.sh Run

Failing wait.sh run if I specify a nonexistent CRD to wait on

Successful Cypress Run

Now with only a simple "can we log in?" test, but it still takes about 30 seconds locally.

Successful tests/scripts/test.sh run

Creates an API token, creates an application, waits for that application to reach Healthy status, and destroys that application. Elapsed time: 30 seconds.

Upgrade Notices

N/A

Closes #221 (closed)