Alert configuration state is not stable

[guyguy333]

Hello,

We use mongodbatlas_alert_configuration to setup our alerts with Opsgenie. It works fine. However, every time we plan / apply without any changes of HCL, terraform detects changes. Moreover, even if we apply, issue remains. I guess resource is not well set or state storing has an issue.

Terraform CLI and Terraform MongoDB Atlas Provider Version

Terraform: 1.0.6
Provider Atlas: 1.1.1

Terraform Configuration File

resource "mongodbatlas_alert_configuration" "host_alerts" {
  for_each   = toset(["HOST_DOWN", "NO_PRIMARY", "HOST_MONGOT_CRASHING_OOM"])
  project_id = var.project_id
  event_type = each.key
  enabled    = true

  notification {
    ops_genie_api_key = var.api_token
    ops_genie_region  = "EU"
    type_name         = "OPS_GENIE"
    interval_min      = 5
    delay_min         = 0
    sms_enabled       = false
    email_enabled     = true
    roles             = ["GROUP_CHARTS_ADMIN", "GROUP_CLUSTER_MANAGER"]
  }
}

As we can see, email_enabled is set as well as interval_min and roles.

Steps to Reproduce

1. terraform init
2. terraform apply

Expected Behavior

interval_min is set as well as email_enabled and roles and a plan/apply doesn't trigger changes after change has been set.

Actual Behavior

resource "mongodbatlas_alert_configuration" "host_alerts" {
        id                     = "xxxxxx"
        # (6 unchanged attributes hidden)

      ~ notification {
          ~ email_enabled          = false -> true
          ~ interval_min           = 0 -> 5
          ~ roles                  = [
              + "GROUP_CHARTS_ADMIN",
              + "GROUP_CLUSTER_MANAGER",
            ]
            # (19 unchanged attributes hidden)
        }
    }

Debug Output

Crash Output

Additional Context

References

[themantissa]

@guyguy333 let me have the team take a look.

[themantissa]

Internal ticket INTMDB-284

[themantissa]

@guyguy333 we determined the root cause. You cannot include both sms_enabled and email_enabled unless type_name is GROUP, ORG, or USER. The API will accept the parameters without error but it's why the state mismatches later. However the documentation is unclear on this so we will update that.

[guyguy333]

Thanks @themantissa. I read the updated documentation in PR. So yes, in that case, as we're using Opsgenie, we can remove both fields sms_enabled and email_enabled. It now ends with:

~ notification {
          ~ interval_min           = 0 -> 5
          ~ roles                  = [
              + "GROUP_CHARTS_ADMIN",
              + "GROUP_CLUSTER_MANAGER",
            ]
            # (20 unchanged attributes hidden)
        }

I think I understand roles issue according to updated documentation and if I understand right, it has no impact for Opsgenie and so should be removed. However, for interval_min, I don't get why this diff is remaining:

notification {
          ~ interval_min           = 0 -> 5
            # (21 unchanged attributes hidden)
        }

resource "mongodbatlas_alert_configuration" "host_alerts" {
 for_each   = toset(["HOST_DOWN", "NO_PRIMARY", "HOST_MONGOT_CRASHING_OOM"])
 project_id = var.project_id
 event_type = each.key
 enabled    = true

 notification {
   ops_genie_api_key = var.api_token
   ops_genie_region  = "EU"
   type_name         = "OPS_GENIE"
   interval_min      = 5
   delay_min         = 0
 }
}

Thanks

[themantissa]

@guyguy333 that's actually a slightly different issue corrected in #624. Will be included in the next release.

[themantissa]

Fixed in 1.2