Destroy fails when AWS resources cannot be found about pulumi-eks HOT 5 CLOSED

Does pulumi refresh help in this case?

Is the scenario that the InternetGateway (or other resource) is manually deleted from AWS behind the scenes? If so, curious what is the desired/suggested behaviour in this case?

from pulumi-eks.

I'm going to close this out - as I believe the original issue is expected, and requires that the user do pulumi refresh or pulumi state delete to reconcile their checkpoint with the cloud provider prior to pulumi destroy.

pulumi/pulumi#2247 is tracking doing a refresh automatically prior to preview/update/destroy. That would be the only thing that could fundamentally "solve" the original issue here.

The other issues referenced along the way sound like unrelated potential bugs, but there are not details on repro steps for those here. Let's open new issues on those if needed.

from pulumi-eks.

+1 got bitten by this

from pulumi-eks.

I got an EKS cluster into a similar status and also hit this that I can't destroy when resources can't be found. A refresh also fails.

Full refresh output - https://gist.github.com/clstokes/e7eb00ae363e9a1974ee5be3679b9944

from pulumi-eks.

In @clstokes' repo, the fact that pulumi refresh fails due to the cluster being unavailable seems like a separate bug we should open and track. Ideally this would not cause a failure - ideally the refresh would warn on these instead of failing.

I assume that the original issue here should be solved by pulumi refresh and that anything that prevents refresh from working here is a bug in refresh. With our current model of not running refresh implicitly, it will be the case that when resources are deleted manually behind the scenes, it will in general be necessary to run pulumi refresh to reconcile those changes into Pulumi.

from pulumi-eks.