Implement graceful shutdown for helper/resource waiter

[radeksimko]

For many resources it's currently impossible for Terraform to gracefully stop the execution mid-flight as helper/resource.StateChangeConf is not cancellable from above the resource/provider.

Mitchell has done the initial work in hashicorp/terraform#9536 (also includes changes in Azure provider + 1 resource). The goal is to make similar changes to (most) other providers - at least the ones we know leverage helper/resource.StateChangeConf and pass the StopContext down to the resource.StateChangeConf in the resource code.

Terraform Version

v0.8.5-dev (aa3eda76425a0b936192c6e95e45758e2727ba4b)

Affected Resource(s)

• any resource using resource.StateChangeConf, but mainly ones that have long timeouts, e.g. AWS RDS, AWS Elasticsearch Domain

Terraform Configuration Files

resource "aws_rds_cluster" "default" {
  cluster_identifier = "aurora-cluster-demo"
  availability_zones = ["eu-west-2a","eu-west-2b"]
  database_name = "mydb"
  master_username = "foo"
  master_password = "barbarbar"
  backup_retention_period = 5
  preferred_backup_window = "07:00-09:00"
}

Expected Behavior

aws_rds_cluster.default: Creating...
...
aws_rds_cluster.default: Still creating... (10s elapsed)
^CInterrupt received. Gracefully shutting down...
Apply failed. Interruption received, partial state saved.

Actual Behavior

aws_rds_cluster.default: Creating...
...
aws_rds_cluster.default: Still creating... (10s elapsed)
^CInterrupt received. Gracefully shutting down...
aws_rds_cluster.default: Still creating... (20s elapsed)
aws_rds_cluster.default: Still creating... (30s elapsed)
aws_rds_cluster.default: Still creating... (40s elapsed)
aws_rds_cluster.default: Still creating... (50s elapsed)
aws_rds_cluster.default: Creation complete

Apply complete! Resources: 1 added, 0 changed, 0 destroyed.

Steps to Reproduce

1. terraform apply
2. Ctrl+C (once)

References

• terraform: ResourceProvider.Stop terraform#9536

[sheerun]

Just like @radeksimko noticed it would best to put StopContext into schema.ResourceData as it is the easiest place to access it when implementing CRUD actions in providers.

[paultyng]

Should investigate this as part of #276

[paultyng]

context.Context support was just merged in to our v2 branch and will be included in that release.

[ghost]

I'm going to lock this issue because it has been closed for 30 days ⏳. This helps our maintainers find and focus on the active issues.

If you have found a problem that seems similar to this, please open a new issue and complete the issue template so we can capture all the details necessary to investigate further.