refactor!: gce-container-declaration deprecation

[d-costa]

Due to the Compute Engine container startup agent deprecation on July 21 2025, GCP recommends now using a startup script or cloud-init to run containers in VMs.

To avoid impacting users that already have startup scripts, this commit uses cloud-init to spin up the container.

Changes:

• Remove container module
• Add service responsible for the Atlantis container
• Changed command and args default values to [] instead of null, since we can't iterate null values

Impact:

• google_compute_disk.persistent will be updated in-place
• google_compute_instance_group_manager.default will be updated in-place
• google_compute_instance_template.default must be replaced

Requires a couple minutes of downtime while the VM is restarted.

Sources:

• https://cloud.google.com/compute/docs/deprecations/container-startup-agent-on-compute
• https://cloud.google.com/compute/docs/containers/migrate-containers

---

I tested creating and updating from an existing deployment, and it went as expected. Nonetheless, it would be best if someone else could confirm it 😄

• Close Migrate to use Cloud-init instead of StartUp-Script for the startup script #185

[d-costa]

hey @Zurvarian, can you take a look? Let me know if you have any suggestions and whether this works for you 👍

[Zurvarian]

@d-costa I'll try to give it a try when I'm free (Possibly next week)

[Zurvarian]

@d-costa I've tested your fork and it worked fine, thanks!

[bschaatsbergen]

Thanks for working on this!

[AndreaGiardini]

@d-costa Thank you for your work 🙏

When testing this change i had to add

  command = []
  args = []

to the definition of my module. Otherwise terraform would fail with:

╷
│ Error: Invalid function argument
│ 
│   on .terraform/modules/atlantis/main.tf line 89, in data "cloudinit_config" "config":
│   89:           ExecStart=/usr/bin/docker run -u ${local.atlantis_uid} --rm --publish '0.0.0.0:${local.atlantis_port}:${local.atlantis_port}' -v ${local.atlantis_disk_mount_path}:${local.atlantis_data_dir} %{for key, value in var.env_vars} -e '${key}=${value}'%{endfor} --name=atlantis ${var.image} ${join(" ", var.command)} ${join(" ", var.args)}
│     ├────────────────
│     │ while calling join(separator, lists...)
│     │ var.command is null
│ 
│ Invalid value for "lists" parameter: argument must not be null.
╵
╷
│ Error: Invalid function argument
│ 
│   on .terraform/modules/atlantis/main.tf line 89, in data "cloudinit_config" "config":
│   89:           ExecStart=/usr/bin/docker run -u ${local.atlantis_uid} --rm --publish '0.0.0.0:${local.atlantis_port}:${local.atlantis_port}' -v ${local.atlantis_disk_mount_path}:${local.atlantis_data_dir} %{for key, value in var.env_vars} -e '${key}=${value}'%{endfor} --name=atlantis ${var.image} ${join(" ", var.command)} ${join(" ", var.args)}
│     ├────────────────
│     │ while calling join(separator, lists...)
│     │ var.args is null
│ 
│ Invalid value for "lists" parameter: argument must not be null.
╵

[d-costa]

thanks for the heads up @AndreaGiardini, I went ahead and changed the default values to [] 👍

[kpocius]

Let me know if this would rather be discussed on a separate issue than a closed PR, but I have just tried upgrading and after restart the disk is simply not mounted. Same if I create a new instance. I looked everywhere in the module and simply fail to see what is supposed to format and mount the disk.

Please ignore me, just found #195