Is it planned to actually support helm releases? Currently when I have an interrupted helm deployment (process was killed) we are left with resources in the kubernetes that we have to manually deploy in order to run pulumi again. If the helm integration used proper helm releases it would be able to “upgrade” a failed release.
There was some discussion about it in https://github.com/pulumi/pulumi-kubernetes/issues/1092
But I didn’t see any resolution about this matter.
09/14/2020, 4:31 PM
As mentioned on that issue, it would be nice to improve integration with Helm, but we have no current plans to use the Helm release process directly. Transformations are a critical part of our Helm support, and would not be possible to support if Helm was managing the deployments directly.
That said, I’d be interested to know more about how you’re getting into a failure state in the first place. I can’t think of any reason we should need Helm to fix a failed deployment (given that Pulumi already handles state resolution), so perhaps you’re hitting a bug.
09/14/2020, 6:15 PM
The scenario I ran into was during a first time deployment of a specific chart and i had a configuration issue about image tag in one of the pods.
Because the pod was in image pull error state the chart deployment was “stuck” and eventually my pod got killed because of a new deployment from our CD flow.
That caused the chart to be partially installed and when pulumi tried to redeploy the chart I got helm errors that the resources already exist