feat: Add global_upgradable variable to support major engine version upgrades to global clusters.

[theherk]

Fix #425. A much more thorough explanation is provided in examples/global-cluster/README.md.

I recognize this adds complexity just to ignore engine_version, but ... I don't know another way around this issue.

Upgrading major version of global clusters

Upgrading the major version of global clusters is possible, but due to a limitation in terraform, it requires some special consideration. As documented in the provider:

When you upgrade the version of an aws_rds_global_cluster, Terraform will attempt to in-place upgrade the engine versions of all associated clusters. Since the aws_rds_cluster resource is being updated through the aws_rds_global_cluster, you are likely to get an error (Provider produced inconsistent final plan). To avoid this, use the lifecycle ignore_changes meta argument as shown below on the aws_rds_cluster.

In order to accomplish this in a module that is otherwise used for non-global clusters, we must duplicate the cluster resource. The limitation that requires this is, terraform lifecycle meta-arguments can contain only literal values:

The lifecycle settings all affect how Terraform constructs and traverses the dependency graph. As a result, only literal values can be used because the processing happens too early for arbitrary expression evaluation.

That means, that to ignore the engine_version in some cases but not in others, we need another resource. So, if you intend to upgrade your global cluster in the future, you must set the new variable global_upgradable to true.

Migrating the resource

If you already have a global cluster created with this module, and would like to make use of this feature, you'll need to move the cluster resource. That can be done with the cli:

terraform state mv 'module.this.aws_rds_cluster.this[0]' 'module.this.aws_rds_cluster.global_upgradable[0]'

Or via a new moved block:

moved {
  from = module.this.aws_rds_cluster.this[0]
  to   = module.this.aws_rds_cluster.global_upgradable[0]
}

After that, changing the major version should work without issue.

Breaking Changes

This is not a breaking change, but if people start using this variable without exercising caution they could wipe a database, but that is the nature of it all I suppose. It could possibly be considered a bug fix on the other hand.

How Has This Been Tested?

• [✓] I have updated at least one of the examples/* to demonstrate and validate my change(s)
• [✓] I have tested and validated these changes against several internal existing implementations.
• [✓] I have executed pre-commit run -a on my pull request

[github-actions]

This PR has been automatically marked as stale because it has been open 30 days
with no activity. Remove stale label or comment or this PR will be closed in 10 days

[theherk]

My patch has been updated to resolve conflicts. That doesn't seem to remove the stale label, so I comment to notify.

[jeff-french]

We are currently facing issues with this in a client environment. Hoping this gets merged soon. 🤞

[github-actions]

This PR has been automatically marked as stale because it has been open 30 days
with no activity. Remove stale label or comment or this PR will be closed in 10 days

[theherk]

Still awaiting review.

[github-actions]

This PR has been automatically marked as stale because it has been open 30 days
with no activity. Remove stale label or comment or this PR will be closed in 10 days

[theherk]

I'll be happy to resolve conflicts once any action happens here. Until then I'll just continue to keep this alive and maintain the fork.