Feat/deploy rollback

• startupProbe for dev deployment, not prod
  • as long as the startup probe fails the container is not ready => the pod is not ready => the old deployment is not deleted
  • by using health route instead of the port only we should be able to catch some issues in application.properties
  • if not successful in a given timeout => undo deployment and exit with non zero to fail pipeline
  • reason why using curl and not httpGet directly? k8s offers httpGet but I don't know why its failing, we have no access to events so not possible to debug from my pov.
• cleanup:
  • pinned images to specific tags
  • build_sha_prod and build_latest_prod can be merged because destination can be provided multiple times, reducing CI runtime and also that both tags are really identical images.

assigned to @oc000045108253

changed the description

mentioned in commit cf15f53f

merged