yamldecode error in the kubectl-apply

[yangligt2]

If you find a similar existing issue, please comment on that issue instead of creating a new one.

If you are submitting a feature request, please start a discussion instead of creating an issue.

Describe the bug

When deploying the gke-a4 blueprint, [gcluster] fails with a yamldecode error in the kubectl-apply module. The error "missing start of document" indicates that yamldecode is being called on an empty or invalid string that can(yamldecode(...)) fails to guard against adequately (or yamldecode itself doesn't handle empty input gracefully even inside can in some terraform versions/contexts, or the expression evaluation order causes issues).

The issue seems to stem from modules/embedded/modules/management/kubectl-apply/kubectl/main.tf.

Steps to reproduce

Steps to reproduce the behavior:

1. Use the gke-a4 example blueprint.
2. Run the deploy command:

   ./gcluster deploy -d examples/gke-a4/gke-a4-deployment.yaml examples/gke-a4/gke-a4.yaml

Expected behavior

The deployment should proceed past the infrastructure validation/initialization stage without Terraform errors.

Actual behavior

The deployment fails with multiple yamldecode errors:

Error: Error in function call

  on modules/embedded/modules/management/kubectl-apply/kubectl/main.tf line 50, in locals:
  50:         can(yamldecode(content)) && length(yamldecode(content)) > 0 ? content : null

Call to function "yamldecode" failed: on line 1, column 1: missing start of
document.

Version (gcluster --version)

gcluster version - not built from official release
Built from 'main' branch.
Commit info: v1.78.0-1-g1ea487b16

Blueprint

[examples/gke-a4/gke-a4-deployment.yaml]:

terraform_backend_defaults:
  type: gcs
  configuration:
    bucket: gke_toolkit_storage

vars:
  project_id: xxxxxxx-gke-dev
  deployment_name: gke-a4-euw4
  region: europe-west4
  zone: europe-west4-b
  static_node_count: 2
  authorized_cidr: 0.0.0.0/0
  spot: true
  system_node_pool_disk_size_gb: 100
  a4_node_pool_disk_size_gb: 500
  enable_periodic_health_checks: true
  health_check_schedule: "0 0 * * 0"

[examples/gke-a4/gke-a4.yaml]:
(Standard example content for gke-a4 with spot reservation)

Expanded Blueprint

N/A

Output and logs

Creating deployment folder "gke-a4-euw4" ...
Initializing deployment group gke-a4-euw4/primary
Testing if deployment group gke-a4-euw4/primary requires adding or changing cloud infrastructure
Error: exit status 1

Error: Error in function call

  on modules/embedded/modules/management/kubectl-apply/kubectl/main.tf line 50, in locals:
  50:         can(yamldecode(content)) && length(yamldecode(content)) > 0 ? content : null

Call to function "yamldecode" failed: on line 1, column 1: missing start of
document.

Screenshots

N/A

Execution environment

• OS: Linux
• Shell: bash
• go version: go1.24.8 linux/amd64

Additional context

Attempted fix:
Modifying [modules/management/kubectl-apply/kubectl/main.tf] to explicit check for empty content or use try fixes the issue.

The fix that worked:

try(length(yamldecode(content)) > 0, false) ? content : null

Instead of:

can(yamldecode(content)) && length(yamldecode(content)) > 0 ? content : null

[siddhartha-quad]

Hi @yangligt2 ,

management/kubectl-apply/kubectl: To resolve the kubectl-apply error, please upgrade the Terraform version to v1.12.0 or later. Please feel free to reach out if the problem still persists.

[yangligt2]

Thank you @siddhartha-quad !
I followed the GKE a4 guide: https://docs.cloud.google.com/ai-hypercomputer/docs/create/gke-ai-hypercompute, and used the cloud shell as recommended. The terraform version on cloudshell is v1.5. So maybe its better to mention the minimum version requirement in the document?

$ terraform --version
Terraform v1.5.7
on linux_amd64

Your version of Terraform is out of date! The latest version
is 1.14.3. You can update by downloading from https://www.terraform.io/downloads.html

[siddhartha-quad]

Hi @yangligt2 ,

Thanks for the suggestion.
I’ll check this with the team and get back to you with more details shortly.