Minimal Example to Reproduce a Bug in the Terraform AWS Provider

This minimal example written in Terraform shows an inconsistent final plan for a CloudWatch alarm when a setting in the Beanstalk is changed.

Steps to Reproduce

1. Set the local variables vpc_id and subnet_ids (comma-separated as string) in main.tf matching the VPC and subnet IDs of your AWS account.
2. Run terraform init.
3. Run terraform apply. An Elastic Beanstalk with application loadbalancer is created. Additionally, a CloudWatch alarm monitoring a specific metric for this loadbalancer is created.
4. Run terraform apply -var='switch_instance_type=true'. This only changes one setting (InstanceType) of the Beanstalk. The plan shows that "data.aws_lb.beanstalk_lb will be read during apply" and "aws_cloudwatch_metric_alarm.loadbalancer_high5xx will be updated in-place". The apply phase will not be successful with the "Error: Provider produced inconsistent final plan".
5. Do not forget to run terrafrom destroy.

This bug is reproducible with

• Terraform: v1.6.4
• AWS Provider: v5.28.0