Destroying resources on AWS fails due to delays in dependency removal

[arnuschky]

Destroying resources on AWS often fails due to delays in the removal of dependent resources.

For example, when removing security groups, terraform often fails because the dependent instances haven't yet been fully removed: even though terraform initiated the removal in correct order, the operation takes a while to complete.

Accordingly the second destroy run always completes successfully.

Same happens with other resources that depend on slow operations, eg vpcs or subnets.

[arnuschky]

Using 0.2.1

[mitchellh]

We should be waiting for the resources to clean up, but I'll take a look. You said instances are an issue. Any other specific cases?

[arnuschky]

Some output:

$ terraform apply destroy.tfplan
aws_instance.a: Destroying...
aws_instance.b: Destroying...
aws_instance.c: Destroying...
aws_internet_gateway.gw: Destroying...
aws_eip.main: Destroying...
aws_eip.main: Error: Address xx.xx.xx.xx is in use. (InvalidIPAddress.InUse)
aws_internet_gateway.gw: Error: Network vpc-xxxxx has some mapped public address(es). Please unmap those public address(es) before detaching the gateway. (DependencyViolation)
aws_instance.a: Destruction complete
aws_instance.c: Destruction complete
aws_instance.b: Destruction complete
Error applying plan:

1 error(s) occurred:

* Address xx.xx.xx.xx is in use. (InvalidIPAddress.InUse)

Terraform does not automatically rollback in the face of errors.
Instead, your Terraform state file has been partially updated with
any resources that successfully completed. Please address the error
above and apply again to incrementally change your infrastructure.

On a second run, all the remaining resources get destroyed.

[mitchellh]

Thanks, I'm working through this now.

[mitchellh]

So the provider does wait until AWS is reporting that the instance is in the "terminated" state, so I'm just going to add retry logic to the EIP destroy.

[mitchellh]

Fixed

[arnuschky]

Awesome, that was fast. Thanks a lot!