[Bug]: `Provider produced inconsistent result after apply` on `advanced_configuration.oplog_size_mb` when cluster scales tier/disk

[Dima-Diachenko]

Is there an existing issue for this?

• I have searched the existing issues

Provider Version

v2.9.0

Terraform Version

v1.14.8

Terraform Edition

Terraform Open Source (OSS)

Current Behavior

Atlas-side update completes successfully, but Terraform errors at the apply post-read:

Terraform configuration to reproduce the issue

Minimal reproducer (single-region Azure replica set with compute auto-scaling enabled). advanced_configuration.oplog_size_mb is intentionally not set in HCL:

resource "mongodbatlas_advanced_cluster" "this" {
  project_id             = var.project_id
  name                   = "repro-rfp"
  cluster_type           = "REPLICASET"
  mongo_db_major_version = "8.0"
  backup_enabled         = true
  pit_enabled            = true
  advanced_configuration = {
    minimum_enabled_tls_protocol = "TLS1_2"
  }
  replication_specs = [{
    region_configs = [{
      provider_name = "AZURE"
      region_name   = "US_WEST"
      priority      = 7
      electable_specs = {
        instance_size = "M50"   # <-- changed to M40 in the second apply
        node_count    = 3
        disk_size_gb  = 1024     # <-- changed to 512 in the second apply
      }
      auto_scaling = {
        compute_enabled            = true
        compute_scale_down_enabled = true
        compute_min_instance_size  = "M50"   # <-- changed to M40
        compute_max_instance_size  = "M60"   # <-- changed to M50
        disk_gb_enabled            = true
      }
    }]
  }]
}

Steps To Reproduce

1. Deploy the cluster via terraform apply the config above.
2. Change electable_specs.instance_size, electable_specs.disk_size_gb, and the auto_scaling.compute_min/max_instance_size to the lower tier (e.g. M40/512/M40/M50).
3. terraform plan shows an in-place update; the diff for advanced_configuration is hidden / unchanged
4. terraform apply does the the cluster scale down but fails with the error in the end:

╷
│ Error: Provider produced inconsistent result after apply
│ 
│ When applying changes to
│ mongodbatlas_advanced_cluster.this["production-rfp"], provider
│ "provider[\"registry.terraform.io/mongodb/mongodbatlas\"]" produced an
│ unexpected new value: .advanced_configuration.oplog_size_mb: was
│ cty.NumberIntVal(52429), but now cty.NumberIntVal(26215).
│ 
│ This is a bug in the provider, which should be reported in the provider's
│ own issue tracker.
╵

Logs

Code of Conduct

• I agree to follow this project's Code of Conduct

[github-actions]

Thanks for opening this issue! Please make sure you've followed our guidelines when opening the issue. In short, to help us reproduce the issue we need:

• Terraform configuration file used to reproduce the issue
• Terraform log files from the run where the issue occurred
• Terraform Atlas provider version used to reproduce the issue
• Terraform version used to reproduce the issue
• Confirmation if Terraform OSS, Terraform Cloud, or Terraform Enterprise deployment

The ticket CLOUDP-400267 was created for internal tracking.

[Dima-Diachenko]

I'm ready to add any additional info/logs if needed.

BTW, I've searched in Activity Logs and found that at cluster scaling was triggered not one but 2 events:

First one:

And after that, the second one:

But I didn't want to manage the oplog_size, I would like even don't touch it, I just need to change the cluster compute.

[EspenAlbert]

Hi @Dima-Diachenko
Thanks for the detailed repro and the activity-log screenshots, they really helped!

Atlas behavior: If oplog_size_mb was ever set (including outside Terraform or in an earlier version of the config), Atlas continues to apply oplog size rules tied to storage, including scaling with disk (for example toward 5% of disk size for General / Low-CPU tiers). You cannot fully “unset” oplog the same way as a cluster that never had a custom oplog, so it is not enough to simply omit oplog_size_mb in Terraform and expect a stable implicit value when disk or tier changes.

On your HCL: You have auto_scaling.disk_gb_enabled = true and also a fixed electable_specs.disk_size_gb. If your intent is manual disk size only, make sure that matches how you want autoscaling to behave, so Terraform and Atlas are not fighting over who owns disk size.

Why you see the inconsistent-result error: The provider reuses some values from state to avoid noisy diffs for attributes you did not set. oplog_size_mb is in that set. When Atlas changes the real oplog after your scale, the post-apply value no longer matches what we kept from state, so Terraform reports Provider produced inconsistent result after apply. The Atlas update itself succeeded but Terraform raises the error. After a failed apply, run terraform plan again, and you will often see no further changes, but the failed apply is still a bad experience.

What you can do for now:

1. Either set oplog_size_mb explicitly in advanced_configuration, for example: oplog_size_mb = 50 * var.disk_size_gb # 5% of disk size, notice MB not GB
2. Or add a lifecycle block so Terraform ignores drift on that attribute (same idea as in the cluster module: ignore_changes = [advanced_configuration.oplog_size_mb]). Requires a Terraform version that supports nested attribute addresses in ignore_changes.

resource "mongodbatlas_advanced_cluster" "this" {
  # ... other arguments ...

  lifecycle {
    ignore_changes = [advanced_configuration.oplog_size_mb]
  }
}

3. Or accept the inconsistency error on applies that change oplog until the provider fix ships.

What we are doing: We are looking at a provider-side change (for example keeping this attribute unknown in the plan when tier/disk changes can change oplog) so you do not hit this error when you update your cluster. But I don't know when that will be available 😅