fix: ignore disk_size when disk_autoresize is true #36
Conversation
Terraform Format and Style 🖌
|
| @@ -69,7 +69,7 @@ resource "google_sql_database_instance" "instance" { | |||
| disk_autoresize = var.settings_disk_autoresize | |||
| disk_autoresize_limit = var.settings_disk_autoresize_limit | |||
| disk_type = var.settings_disk_type | |||
| disk_size = var.settings_disk_size | |||
| disk_size = var.settings_disk_autoresize ? null : var.settings_disk_size | |||
There was a problem hiding this comment.
What should be starting value of disk? You should have the default value of disk as 10GB and ignore the changes to the attribute disk_size in the life_cycle hooks.
There was a problem hiding this comment.
Why would we ignore the disk size @sunilhonest ?
There was a problem hiding this comment.
life_cycle already ignores updates to disk_size. We don't need an initial value, it's optional and will default to the minimum size (10GB).
disk_size - (Optional) The size of data disk, in GB. Size of a running instance cannot be reduced but can be increased. The minimum value is 10GB.
If you think there's use-case where we need a manual initial size bigger than 10GB, we can only ignore it for read-replicas.
There was a problem hiding this comment.
Updated to ignoresettings_disk_size only for the read replica to allow a custom initial size for the master instance.
There was a problem hiding this comment.
@jai We have disk_autoresize enabled and hence GCP increases the disk size. TF apply will try to reset it to value passed(ex: 10 GB in our case) and hence we should ignore it in the life_cycle.
Terraform Format and Style 🖌
|
The only test for this case is a scenario:
Quite complicated and not really worth it just for this case IMO. |
|
🎉 This PR is included in version 1.10.1 🎉 The release is available on GitHub release Your semantic-release bot 📦🚀 |
Pull Request Submission Checklist
Please confirm that you have done the following before requesting reviews:
Description
settings_disk_sizewhensettings_disk_autoresizeis set to true to avoid failing to create a read-replica after the master instance has been auto-scaled up.This change is