FAILED release status during Helm upgrade resource deletion

Resource Consistency in Failed Upgrades

In Helm 3, executing a helm upgrade that removes specific resources from the chart can result in a FAILED release status if subsequent resource updates fail. While the release is marked as failed, the deletion of the removed resources may have already been committed to the Kubernetes cluster.

The goal is to ensure a consistent state when reverting to a previous revision. There is uncertainty regarding whether helm rollback reliably recreates resources that were deleted during the failed upgrade attempt, as the cluster state may diverge from the intended revision manifest.

• Does helm rollback explicitly trigger the recreation of resources that were deleted during a failed upgrade?
• How does the --atomic flag handle the restoration of deleted resources compared to a manual rollback command?