toolkit: stacks can get stuck in ROLLBACK_COMPLETE

[rix0rrr]

If the first deploy of a stack fails, the stack gets stuck in ROLLBACK_COMPLETE and requires out-of-band user intervention to delete it and try again:

$ cdk -a 'node integ.autoscaling.js' deploy
 ⏳  Starting deployment of stack aws-cdk-dynamodb...
 ❌  Deployment of stack aws-cdk-dynamodb failed: ValidationError: Stack:arn:aws:cloudformation:us-east-1:993655754359:stack/aws-cdk-dynamodb/540d7d70-cd53-11e8-b11c-50d5cd265c36 is in ROLLBACK_COMPLETE state and can not be updated.
Stack:arn:aws:cloudformation:us-east-1:993655754359:stack/aws-cdk-dynamodb/540d7d70-cd53-11e8-b11c-50d5cd265c36 is in ROLLBACK_COMPLETE state and can not be updated.

This didn't use to happen because we created an empty stack and updated from there: rollback would update back to a safe point from which we could try to deploy again. But right now, you have to use the AWS CLI to delete the stack, wait for that, and only THEN can you try using the CDK again.

Pretty bad experience for first-time users who are almost guaranteed to get things wrong on the first try, and then get stuck in a position from which it's not obvious how to recover from.

[eladb]

Oouch. Definitely worth fixing. From: Rico Huijbers <[email protected]> Reply-To: awslabs/aws-cdk <[email protected]> Date: Thursday, October 11, 2018 at 3:58 PM To: awslabs/aws-cdk <[email protected]> Cc: Subscribed <[email protected]> Subject: [awslabs/aws-cdk] toolkit: stacks can get stuck in ROLLBACK_COMPLETE (#901) If the first deploy of a stack fails, the stack gets stuck in ROLLBACK_COMPLETE and requires out-of-band user intervention to delete it and try again: $ cdk -a 'node integ.autoscaling.js' deploy ⏳ Starting deployment of stack aws-cdk-dynamodb... ❌ Deployment of stack aws-cdk-dynamodb failed: ValidationError: Stack:arn:aws:cloudformation:us-east-1:993655754359:stack/aws-cdk-dynamodb/540d7d70-cd53-11e8-b11c-50d5cd265c36 is in ROLLBACK_COMPLETE state and can not be updated. Stack:arn:aws:cloudformation:us-east-1:993655754359:stack/aws-cdk-dynamodb/540d7d70-cd53-11e8-b11c-50d5cd265c36 is in ROLLBACK_COMPLETE state and can not be updated. This didn't use to happen because we created an empty stack and updated from there: rollback would update back to a safe point from which we could try to deploy again. But right now, you have to use the AWS CLI to delete the stack, wait for that, and only THEN can you try using the CDK again. Pretty bad experience for first-time users who are almost guaranteed to get things wrong on the first try, and then get stuck in a position from which it's not obvious how to recover from. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub<#901>, or mute the thread<https://github.com/notifications/unsubscribe-auth/AAkjDFXEfAl0AN83RncrLfmFViDNd4Gdks5uj0BjgaJpZM4XXfNF>.

[leandropadua]

I'm experiencing this issue in version 1.38.0:

❌  SampleApp-testingleandro-PipelineStack failed: Error: The stack named SampleApp-testingleandro-PipelineStack failed creation, it may need to be manually deleted from the AWS console: ROLLBACK_COMPLETE
--

What is the default behaviour? Is there a way to force the deletion of existent stuck stacks?