Terraform and NREC: Part VI - Handling the state files

Last changed: 2024-04-17

Handling the state file after creating resources is a known pain point, but still crucial in the life cycle management of the infrastructure. Using an object store as a terraform backend can mitigate this. In this example vi will define the NREC object store as an S3 compatible backend for the terraform state. For this to work, you must export a couple of environment variables, AWS_ACCESS_KEY_ID and AWS_SECRET_ACCESS_KEY. Create a file called for example provider.tf. The following code is suitable for terraform versions greater than 1.6.3:

terraform {
  backend "s3" {
    endpoints                   = { s3 = "https://object.api.bgo.nrec.no:8080" }
    bucket                      = "bgo-nrec-terraformstate"
    use_path_style              = true
    key                         = "bgo.myinfra.tfstate.tf"
    region                      = "bgo"
    skip_credentials_validation = true
    skip_region_validation      = true
    skip_requesting_account_id  = true
    skip_s3_checksum            = true
  }
}

Simply replace bgo with osl in order to choose another region. When using this configuration, you need to create the bucket before running terraform init, for example

$ s3cmd mb s3://bgo-nrec-terraformstate

If you use a terraform version older than 1.6, use the following example (version 1.6.0 to 1.6.3 will not work at all). The bucket will be created by terraform:

terraform {
  backend "s3" {
    endpoint                    = "https://object.api.bgo.nrec.no:8080"
    bucket                      = "bgo-nrec-terraformstate"
    use_path_style              = true
    key                         = "bgo.myinfra.tfstate.tf"
    region                      = "bgo"
    skip_credentials_validation = true
    skip_region_validation      = true
  }
}

When running terraform init the backend will be inititalized. After applying an infrastructure with terraform the state file will be stored in the NREC object storage.