provider/aws: Troubles destroying large ASGs

[hashibot]

This issue was originally opened by @phinze as hashicorp/terraform#3495. It was migrated here as part of the provider split. The original body of the issue is below.

---

hashicorp/terraform#3485 attempted to fix force_delete, but @maratoid provided a test config that still yields errors.

Current status with his config is:

• we skip the Terraform-managed drain properly
• we specify ForceDelete on the API call
• AWS still seems to need to churn through terminations of all nodes before the ASG is deleted
• Terraform times out waiting for the ASG to show up as deleted
• The internet gateway also times out... still haven't figured out if that's related or not.

In order to fix, I think we need to:

• parameterize the timeout value for both ASGs and IGWs to allow users to crank that up
• potentially figure out the "not found" issue if it's a separate IGW bug

This is all a bit jargon-y, happy to clarify anything if someone else digs into this before I do.

[hashibot]

This comment was originally opened by @maratoid as hashicorp/terraform#3495 (comment). It was migrated here as part of the provider split. The original comment is below.

---

Hey guys. Just wondering if this is in any kind of a roadmap. We're still having to work around this in all sorts of clunky ways.

Thanks.

[hashibot]

This comment was originally opened by @phinze as hashicorp/terraform#3495 (comment). It was migrated here as part of the provider split. The original comment is below.

---

Acknowledging that this is still an issue. Been kicking around a core-supported solution for this, let me braindump some notes on that:

Terraform needs support for per-resource user-configurable timeouts. These timeouts should apply to Create, Update, Destroy operations.

Sketch of one possible config (not super sold on it - there might be a nicer spot to drop this):

resource "aws_autoscaling_group" "foo" {
  # ...
  timeouts {
    create  = "30m"
    destroy = "30m"
    update  = "10m"
  }
}

Resource implementations would need to opt-in for support of these config values so that if a user specifies them on an unsupported resource, a validation error is returned.

[hashibot]

This comment was originally opened by @maratoid as hashicorp/terraform#3495 (comment). It was migrated here as part of the provider split. The original comment is below.

---

Any news :) ?

[hashibot]

This comment was originally opened by @maratoid as hashicorp/terraform#3495 (comment). It was migrated here as part of the provider split. The original comment is below.

---

ping

[Ninir]

Hi everyone,

The ASG deletion timeout is now configurable as per #949.

I am going to close this one if you folks don't mind, let me know if this is still an issue!

[ghost]

I'm going to lock this issue because it has been closed for 30 days ⏳. This helps our maintainers find and focus on the active issues.

If you feel this issue should be reopened, we encourage creating a new issue linking back to this one for added context. Thanks!