Auto Deferring MongoDB Atlas Maintenance Window Fails

[nleite]

Atlas terraform provider throwing back a 400 while creating a project maintenance window with auto_defer option set. Seems like there is an out-of-order (or something similar) API call that tries to set the autoDefer before the maintenance window is set or gets completed.

Terraform CLI and Terraform MongoDB Atlas Provider Version

    mongodbatlas = {
      source  = "mongodb/mongodbatlas"
      version = "1.2.0"
    }

Terraform Config

...
resource "mongodbatlas_maintenance_window" "cluster_maintenance_window" {
    project_id = mongodbatlas_project.atlas_project.id
    auto_defer = var.maintenance_auto_defer
    day_of_week = var.maintenance_day_of_week
    hour_of_day = var.maintenance_hour_of_day
}
...

Steps to Reproduce

terraform init terraform plan terraform apply

Expected Behavior

Should create the maintenance window for the project without any issues

Actual Behavior

│ Error: error auto deferring the MongoDB Atlas Maintenance Window (XXXXXXXX): POST https://cloud.mongodb.com/api/atlas/v1.0/groups/XXXXXXXXXX/maintenanceWindow/autoDefer: 400 (request "ATLAS_MAINTENANCE_WINDOW_NOT_DEFINED") No maintenance window defined for project. This action is only supported for projects with user defined maintenance windows.
│ 
│   with module.test-webtask.mongodbatlas_maintenance_window.cluster_maintenance_window,
│   on .terraform/modules/test-webtask/cluster/atlas.tf line 213, in resource "mongodbatlas_maintenance_window" "cluster_maintenance_window":
│  213: resource "mongodbatlas_maintenance_window" "cluster_maintenance_window" {

[themantissa]

@nleite we'll have the team take a look and get back to you.

[nikhil-mongo]

@nleite The terraform resource is designed correctly. In order to set the number of auto-deferrals, it is important for the maintenance window to exist.

Therefore, initially the resource should be as below -

resource "mongodbatlas_maintenance_window" "cluster_maintenance_window" {
    project_id = mongodbatlas_project.atlas_project.id
   // auto_defer = var.maintenance_auto_defer
    day_of_week = var.maintenance_day_of_week
    hour_of_day = var.maintenance_hour_of_day
}

- Run `terraform apply`

Once completed, alter the resource as below - 

```terraform
resource "mongodbatlas_maintenance_window" "cluster_maintenance_window" {
    project_id = mongodbatlas_project.atlas_project.id
    auto_defer = var.maintenance_auto_defer
    day_of_week = var.maintenance_day_of_week
    hour_of_day = var.maintenance_hour_of_day
}

• Now run the terraform apply again and this should do what you are trying to accomplish.

The above difference is because there are two different API calls required to accomplish this.

The create maintenance window call is -

PATCH /api/atlas/v1.0/groups/{GROUPID}/maintenanceWindow

While when the auto_defer is set it calls for the below API -

POST /api/atlas/v1.0/groups/{GROUPID}/maintenanceWindow/autoDefer

Since these two API calls cannot be combined, and when terraform finds auto_defer as mentioned in resource it straightaway hits the second API call which fails with 400 error as there is no maintenance window to autodefer.

I hope I have been able to answer your question.

Thanks

[themantissa]

Internal ticket INTMDB-305 to investigate solutions.

[martinstibbe]

@nleite This fix will be in version 1.3.1 pre release for testing soon

[themantissa]

@nleite Pre-release with the fix is out if you'd like to test it. We'll likely release Monday. https://github.com/mongodb/terraform-provider-mongodbatlas/releases/tag/untagged-bcc6453f1acaa220dd7e

[themantissa]

1.3.1 is released