terraform-provider-aws: Error: aws_autoscaling_group - Provider produced inconsistent final plan

Hey, i can confirm we’re still facing this issue :

│ Error: Provider produced inconsistent final plan
│
│ When expanding the plan for aws_autoscaling_group.workers_3a to include new values learned so far during apply, provider "registry.terraform.io/hashicorp/aws" produced an
│ invalid new value for .tags: planned set element cty.MapValEmpty(cty.String) does not correlate with any element in actual.
│
│ This is a bug in the provider, which should be reported in the provider's own issue tracker.

Another kind of error that we’re facing:

│ Error: Provider produced inconsistent final plan
│
│ When expanding the plan for aws_autoscaling_group.workers_3a to include new values learned so far during apply, provider "registry.terraform.io/hashicorp/aws" produced an
│ invalid new value for .tags: length changed from 10 to 11.
│
│ This is a bug in the provider, which should be reported in the provider's own issue tracker.

These errors are only happening when we’re creating an aws_eks_cluster (which takes roughly 13min), an aws_launch_template and the aws_autoscaling_group on the same run. (3 resources)

Once the EKS cluster is created (and the failure of the apply), we can relaunch a terraform apply and create the launch template and the autoscaling group without any issues. Maybe related to something computed at the beginning of the run that is updated later on?

By the way : we’re using the latest version of the provider (“3.69.0”) and terraform version 1.1.0

I’ve seen this as well, but I think it’s because I’m using a value from random_string in the tags; the random value isn’t available until apply-time, at which point terraform freaks out that things have changed. In fact, the list of tags during the plan shows that some items are (known after apply).

However, since Terraform knows that the values won’t be known until they’re available, it should proceed without throwing the warning.

Here’s the actual output:

Error: Provider produced inconsistent final plan

When expanding the plan for module.example_asg.aws_autoscaling_group.app[0] to
include new values learned so far during apply, provider
"registry.terraform.io/-/aws" produced an invalid new value for .tags: length
changed from 6 to 8.

This is a bug in the provider, which should be reported in the provider's own
issue tracker.

This is the output of terraform -version:

Terraform v0.12.24
+ provider.aws v2.64.0
+ provider.random v2.2.1
+ provider.template v2.1.2

This happened to me because I was using both availability_zones and vpc_zone_identifier in aws_autoscaling_group.

The documentation mentions:

availability_zones - (Required only for EC2-Classic) A list of one or more availability zones for the group. This parameter should not be specified when using vpc_zone_identifier.

The error is gone if I don’t specify the two, but it took a little sleuthing to piece this together.

Looks very similar to https://github.com/terraform-providers/terraform-provider-aws/issues/10615

Sometimes just running terraform apply again will work. But if it doesn’t, as a workaround you can mark the resource as tainted with -taint and then run terraform apply again. This will force the resource to be re-created.

@justinretzolk I have not seen this problem in quite some time. For the record, we are mostly on aws provider version 3.54+ and Terraform 1.0.0+ at this point.