Upgrade tests dispatch Delete calls to the wrong version

[t0yv0]

What happened?

There is an issue @flostadler noticed when upgrading EKS provider tests. It seems that the way we are overriding the provider to use the release candidate provider substitutes all the calls, which includes dispatching Delete calls to the WIP provider. When a resource is removed it can't service them anymore.

Example

I'd love a better example as we just discussed this informally.

Output of pulumi about

N/A

Additional context

N/A

Contributing

Vote on this issue by adding a 👍 reaction.
To contribute a fix for this issue, leave a comment (and link to your pull request, if you've opened one already).

[danielrbradley]

Is this not true for real users who are upgrading and find the old resource removed? Or does this only affect the test due to the gRPC attachment method intercepting all delete calls even if they were being sent to the previous version of the provider? If so, this might be an issue of how attach works - we need to target a single version of the provider and not others.

[t0yv0]

Or does this only affect the test due to the gRPC attachment method intercepting all delete calls even if they were being sent to the previous version of the provide

I was assuming this is indeed a problem but we need a repro from Florian or some more pointers to confirm

[t0yv0]

f so, this might be an issue of how attach works - we need to target a single version of the provider and not others.

Exactly my thought.