[improvement] google_compute_disk support for resizing disk

[JorritSalverda]

Increasing the size of google_compute_disk forces recreation, however it's possible to resize a google cloud disk on the fly, see https://cloud.google.com/compute/docs/disks/add-persistent-disk.

Instead of recreating the resource just modifying the size would be ideal. Now I still have to do that outside of terraform and then update my tf file and the terraform.tfstate file to reflect the new size.

Terraform Version

Terraform v0.7.13

Affected Resource(s)

google_compute_disk

[paddycarver]

Hey @JorritSalverda! Thanks for bringing this up. Definitely a great feature that we'd love to add. Looking into it a bit:

• Modifying persistent disks is currently problematic. See 'google_compute_instance' adding 'google_compute_disk' forces new resource #6678. We need to come to a solution that allows us to edit modify disks without triggering a forced recreation of the instance, because currently we can't change anything about a disk without recreating the instance. This is a bit complicated, but @danawillow is thinking on it, and we'll continue talking about it.
• Even if Terraform supports resizing the disk itself, I can't think of a reasonable or scalable method for resizing the filesystem, which also needs to happen. I know the AWS provider just did this, so I'll look into that and figure out what they did.
• Finally, resizing the root disk gets even more tricky, as you need to restart the instance for it to take effect. Terraform doesn't know about started/stopped/restarting, it only really knows about creating and destroying. Again, I think the AWS provider just had to deal with this, so I'll look into what they did.

As for updating your tfstate, you should be able to just update the config, and refresh should update the state to show the new size. If not, that's a bug, and please let me know!

So this is a bit complicated, and has some work that needs to get done before it's feasible. Which isn't to say that it won't happen, but I just want to set expectations accordingly that it may not happen immediately, and if/when it does happen, there may be some caveats around the filesystem.

[bgt101]

Increasing the size of the disk using gcloud and then refreshing the state does not update the size in the state.

As for growing the filesystem, that can be done while mounted without a reboot. I suspect it could be implemented in a null_resource using a trigger and a remote-exec, without the need for additional Terraform changes.

[paddycarver]

Thanks for letting us know, @bgt101. This is probably related to #14136. We're looking into it to see what's going on.

[danawillow]

Support for this was added in #15077. However, it does not address the pitfalls noted in this issue, so please be very careful when using it.

[bgt101]

In Terraform 0.9.8 resizing disks works beautifully. Thanks!

To resize the filesystem online, I used a null_resource like this:

resource "null_resource" "fs_grower" {
  triggers {
    disk_size = "${google_compute_disk.data.size}"
  }

  provisioner "remote-exec" {
    connection {
      host = "${google_compute_instance.instance.network_interface.0.address}"
      type = "ssh"
      user = "xxxxx"
    }

    inline = [
      "sudo xfs_growfs -d /data",
    ]
  }
}

It gets a little trickier for devices which are partitioned, but doable.

[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 have found a problem that seems similar to this, please open a new issue and complete the issue template so we can capture all the details necessary to investigate further.