Linking an environment group detaches it from other services

[ahmadabdelhalim]

Hey team,

I'm facing an issue with linking environment groups. the current implementation attempts to link an environment group to the PR environment upon deployment. However, there’s a bug with the Render provider that detaches the environment group from the previous services and applies it only to the newly deployed one. This causes the previous services to fail, requiring you to manually link the environment group through the UI for both the web and worker services. My theory is that the render provider only gets the services from this array

  service_ids  = [
    render_web_service.pr_web_service.id,
    render_background_worker.pr_worker_service.id,
  ]

ignoring all other services. is there a quick solution for this issue?

data "render_env_group" "staging" {
  id = var.staging_env_group_id
}

resource "render_postgres" "pr_postgres_database" {
  name         = "${var.pr_branch}-database"
  plan         = "basic_256mb"
  disk_size_gb = 10
  region       = "frankfurt"
  version      = "16"
}

resource "render_web_service" "pr_web_service" {
  name               = "${var.pr_branch}-web"
  plan               = "standard"
  region             = "frankfurt"
  start_command      = "bundle exec puma -C config/puma.rb"
  health_check_path  = "/health"

  pre_deploy_command = <<-EOT
    DEMO_DATABASE_URL=${var.demo_database_url} bundle exec rake db:import_demo
  EOT

  runtime_source = {
    native_runtime = {
      auto_deploy   = true
      branch        = var.pr_branch
      build_command = "./bin/review-app.sh"
      repo_url      = "test"
      runtime       = "ruby"
    }
  }

  maintenance_mode = {
    enabled = false
    uri     = ""
  }

  env_vars = {
    DATABASE_URL = {
      value = render_postgres.pr_postgres_database.connection_info.internal_connection_string
    }
  }
}

resource "render_background_worker" "pr_worker_service" {
  name          = "${var.pr_branch}-worker"
  plan          = "standard"
  region        = "frankfurt"
  start_command = "bundle exec sidekiq"

  runtime_source = {
    native_runtime = {
      auto_deploy   = true
      branch        = var.pr_branch
      build_command = "./bin/review-app.sh"
      repo_url      = "test"
      runtime       = "ruby"
    }
  }

  env_vars = {
    DATABASE_URL = {
      value = render_postgres.pr_postgres_database.connection_info.internal_connection_string
    }
  }
}

resource "render_env_group_link" "pr_services_env_link" {
  env_group_id = data.render_env_group.staging.id
  service_ids  = [
    render_web_service.pr_web_service.id,
    render_background_worker.pr_worker_service.id,
  ]
}

[mdbenjam]

To make sure I understand what you're asking, you have already defined a env group in the Dashboard that is connected to a set of services, but when you apply this terraform it unlinks from those services and instead links to the services defined in your terraform file.

This is currently the desired behavior for the terraform provider. Merging the list of service_ids with existing service_ids means the resources described in terraform don't actually match the resources described in Render. If you're going to manage a resource within the terraform provider it needs to be completely managed within it.

That said I think there are a couple of workarounds here:

1. Define the services you made in the Dashboard in your terraform file. This way you can reference them all in pr_services_env_link.
2. Get the service_ids for the existing services in the Dashboard (if you visit the service, the service id is the part of the url that begins with srv-), and hard code those service ids in the pr_services_env_link block of the terraform file.

Please let me know if I didn't fully understand what you're asking for.

[ahmadabdelhalim]

thank you for the detailed answer. yes, your understanding is correct.
I’m using the Terraform provider to manage preview environments, and I’m trying to have an environment group that I can attach to all of them. I feel the environment group should still be able to retain the already attached services. otherwise, I’d have to ditch the environment group and define all environment variables directly in Terraform. the options you provided seem like they would require manual effort, which I’m trying to avoid because I want to automate the entire process.

[mdbenjam]

Where are the existing services in the service_ids list coming from? Are they defined in the Dashboard, or in another file in your terraform?

Another alternative would be to use the datasource you've already defined to get the existing list of service_ids. I haven't tested this but think something like this should work using the concat and distinct terraform functions:

resource "render_env_group_link" "pr_services_env_link" {
  env_group_id = data.render_env_group.staging.id
  service_ids  = distinct(concat(data.render_env_group.staging.service_ids, [
    render_web_service.pr_web_service.id,
    render_background_worker.pr_worker_service.id,
  ]))
}

[ahmadabdelhalim]

I thought that would work but I got the following error:

Error: Unsupported attribute

  on services.tf line 71, in resource "render_env_group_link" "pr_services_env_link":
  71:   service_ids  = distinct(concat(data.render_env_group.staging.service_ids, [

This object has no argument, nested block, or exported attribute named
"service_ids".