Requesting: on error in refresh for unreachable cluster, allow marking as deleted

Hello!

• Vote on this issue by adding a 👍 reaction
• If you want to implement this feature, comment to let us know (we’ll work with you on design, scheduling, etc.)

Issue details

Hi,

I’d like to propose a flag in refresh to allow the pre #1522 behavior; unreachable cluster, pods, Helm releases, etc, should be marked in the Pulumi state as deleted. As it stands now, Pulumi returns an error, and I have to work through a manual delete process.

Scenario:

I have gotten a Kubernetes cluster into a bad state. Pulumi had the state marked as “pods exist”, “charts exist”, “cluster exists”, etc, but now is unable to access any information due to the bad state. The solution I want to effect is to remove and regenerate the cluster to, nominally, bring the system state into a known-good one. The current Pulumi process is to manually edit the state to remove the relevant resources, something regrettably error prone.

To be precise, the “loss of state data” behavior specifified here https://github.com/pulumi/pulumi-kubernetes/issues/1485#issuecomment-816970573 is precisely what I want to effect here.

However, as per #1485 , this is not appropriate always; hence the request for a flag. I would suggest perhaps something as --assume-errors-are-removals.

Affected area/feature

[lblackstone]

I think it would be easier to document if we include a config parameter. Assuming the provider can pick that up on a destroy operation, let’s do both.

[stack72]

That sounds perfect - with an env var it will be a strict opt-in, should we offer only env var and not a config param or should we have both?