Fails to create change sets when a stack is in the REVIEW_IN_PROGRESS state

[amcinnes]

Using stackup 1.2.0

If I do change-set create against a stack that doesn't exist, a stack is created in the REVIEW_IN_PROGRESS state, as expected.

If I then try to create another change set, I get ERROR: no such stack.

I believe this is because stackup uses a change set type of UPDATE when the stack already exists -- but for a stack in the REVIEW_IN_PROGRESS state, it's appropriate to use a change set type of CREATE instead.

[amcinnes]

(I concede that this isn't a particularly important issue, as it's always safe to delete a REVIEW_IN_PROGRESS stack and try again...)

[amcinnes]

A related observation:

Normally if I update a stack that's in the ROLLBACK_COMPLETE state, stackup will automatically delete the stack and recreate it.

But if I try to create a change set for a stack that's in the ROLLBACK_COMPLETE state, stackup will say ERROR: Stack:... is in ROLLBACK_COMPLETE state and can not be updated.

I wonder if it would be appropriate to automatically delete a stack in ROLLBACK_COMPLETE (or CREATE_FAILED) before trying to create a change set for it.

[mbelekar]

@amcinnes The stackup up/down command behaves like an upsert, where as the change sets are a CRUD operation. We can update the documentation to reflect this difference in behaviour. Right now we don't know when we will be supporting an upsert operation on change sets.

[stevemac007]

Just ran into this one myself - requires manual intervention by someone to delete these dangling stacks - It would be awesome if stackup could cleanup these REVIEW_IN_PROGRESS stacks as part of normal operation.

[mdub]

It would make sense to be consistent with the auto-cleanup on stack creation.