What problem are you trying to solve?

Focus on the what and why of the need you have, not the how you'd like it solved.

With disk autoscaling enabled, changing disk_size_gb in terraform has no effect. The plan shows the change, apply completes, but Atlas keeps the old size and the next plan shows no diff. The docs say that Atlas keeps the current disk size and ignores the requested value when auto scaling is on. The documented way around it is to disable autoscaling, apply the change, and reenable. For our workflows that doesn't work, because disabling disk_gb_enabled would also require advanced_configuration.oplog_min_retention_hours = 0 in a separate apply first, and our clusters are also required to have non zero oplog windows.

What would you like to see happen?

Describe the desired outcome or enhancement.

TF applies a disk_size_gb increase even with autoscaling enabled. Increase only behavior would be enough, although both would be better. The minimum would be to apply the requested size when it is above the current effective size, and ignore it when it is at or below.

Why is this important to you or your team?

Explain how the request adds value or solves a business need.

We manage many clusters with terraform, and auto scaling must stay on for all of them. We have scenarios such as a scheduled increase load from a customer, a cluster warmup for onboarding, migration planning etc. we want to raise the disk by changing a config value and running the normal deploy sequence we have in place. Today that change silently does nothing, which is easy to miss. During a disk alert it also forces a manual change in Atlas, outside our normal reviewed workflow.

What steps, if any, are you taking today to manage this problem?

We added a step that runs before terraform on every deploy. It reads the requested disk size from our config, compares it to the live Atlas disk, and if the request is larger it grows the disk with a direct Atlas Admin API call. It waits for the cluster to be idle before terraform runs. This works with auto scaling enabled, so Atlas allows the resize, only the terraform path does not. It is extra step that is added to all of our deploys, whether or not a config changed, slowing down our workflows and adding cost to the setup. It also removes a part of the job we associate with TF.