local.node_pool_image_ids returns null

[ecout]

Keep getting this error on resource manager. Nothing has changed on our end.

`Error: Error in function call

on .terraform/modules/oci-oke/oke.tf line 77, in resource "oci_containerengine_node_pool" "oci_oke_node_pool"

77: image_id = (var.node_pool_image_type == "oke"" && length(regexall(""GPU|A1"", var.node_shape)) == 0) ? (element([for source in local.node_pool_image_ids : source.image_id if length(regexall(""Oracle-Linux-${var.node_linux_version}-20[0-9].-OKE-${local.k8s_version_only}"", source.source_name)) > 0], 0)) : (var.node_pool_image_type == ""oke"" && length(regexall(""GPU"", var.node_shape)) > 0) ? (element([for source in local.node_pool_image_ids : source.image_id if length(regexall(""Oracle-Linux-${var.node_linux_version}-Gen[0-9]-GPU-20[0-9].-OKE-${local.k8s_version_only}"", source.source_name)) > 0], 0)) : (var.node_pool_image_type == ""oke"" && length(regexall(""A1"", var.node_shape)) > 0) ? (element([for source in local.node_pool_image_ids : source.image_id if length(regexall(""Oracle-Linux-${var.node_linux_version}-aarch64-20[0-9].-OKE-${local.k8s_version_only}",source.source_name))>0],0)) null

├────────────────

│ local.k8s_version_only is "1.26.2"

│ local.node_pool_image_ids is list of object with 104 elements

│ var.node_linux_version is "7.9"

Call to function "element" failed: cannot use element function with an empty

list.

`

[robo-cap]

The version 1.26.2 is no longer supported. The datasource we are using to fetch the latest images includes no OKE image based on OL 7.9 for K8s version 1.26.2.

If you intend to continue using version 1.26.2 with the current OS version for the worker nodes, I recommend you add the following attributes to the nodepool definition:

demo = {
#       mode               = "node-pool",
#       size               = 1,
#       shape              = "VM.Standard.E4.Flex"
#       ocpus              = 1,
#       memory             = 8,
#       boot_volume_size   = 50,
       image_type         = "custom"
       image_id             = "<current_image_ocid_used_for_the_nodepool>"
     }

You can run this command to see all the versions being returned:
oci ce node-pool-options get --node-pool-option-id all -c ocid1.compartment.oc1......4diqq | grep 1.26.2

[ecout]

Version 1.26.2 is a DEFAULT value the module the terraform code is invoke has. The module is the one throwing the error. The problem is indeed in the logic of the terraform code. The correct combination of worker-node-image and kubernetes_version needs to be selected to yield a valid image_id. Adding a custom image Id as on OCID is not how this repository is configured to function.

Please refer to the module repo for details:
https://github.com/oracle-devrel/terraform-oci-arch-oke/blob/main/oke.tf

I actually already did that and found a match to move forward the problem is the MODULE this repository is invoking does not necessarily find all the combinations this terraform code offers. Meaning this repo provides the possibility input values that are invalid, instead of validating them.

[ecout]

What we need to be able to do is keep upgrading the clusters as new versions of OKE are released by OCI and for that, this terraform code and the module it utilizes need to be able to find the available images that the kubernetes_version variable selects by not depending on the autoscaler_deployment build version alone, because those do not always match always.

The problem is on this file and the autoscaler_image lookup function:
https://github.com/oracle-devrel/terraform-oci-arch-oke-autoscale-gitlab-runners/blob/main/oke_autoscaler.tf

locals { autoscaler_deployment = templatefile("${path.module}/templates/autoscaler.template.yaml", { autoscaler_image = "${lookup(local.autoscaler_image, var.kubernetes_version)}" # BUG here min_nodes = "${var.min_number_of_nodes}" max_nodes = "${var.max_number_of_nodes}" node_pool_id = "${module.oci-oke.node_pool.id}" }) }

The Autoscaler Image path build counter does not always concides with the available K8s version in the images for a region:
https://docs.oracle.com/en-us/iaas/Content/ContEng/Tasks/contengusingclusterautoscaler_topic-Working_with_the_Cluster_Autoscaler.htm#contengusingclusterautoscaler_topic-Working_with_the_Cluster_Autoscaler-step-copy-CA-config-file

Take for example

US West (Phoenix) | Kubernetes 1.30 | phx.ocir.io/oracle/oci-cluster-autoscaler:1.30.2-10 -- | -- | --

There's no IMAGE available under the worker-node-image input that is version 1.30.2 and above. Only 1.30.1.

If I add this code to the oke.tf to generate the list of images, I don't see an image with that. See attached.

ocid1.ormjob.oc1.phx.amaaaaaavg3kmwyajc7dtjrw5sogpm25vptsr3e7mbpl3ylyx6cokryd45nq.log

data "oci_containerengine_node_pool_option" "oci_oke_node_pool_option" {
  node_pool_option_id = "all"
}

variable "node_pool_image_type" {
  default = "oke"
}

locals {
 node_pool_image_ids = data.oci_containerengine_node_pool_option.oci_oke_node_pool_option.sources # Oracle-Linux-8.10-2024.11.30-0
  k8s_version      = var.kubernetes_version
  k8s_version_length         = length(local.k8s_version)
  k8s_version_only           = substr(local.k8s_version,1,local.k8s_version_length)
  image_id = (var.node_pool_image_type == "oke" && length(regexall("GPU|A1", var.worker_shape)) == 0) ? (element([for source in local.node_pool_image_ids : source.image_id if length(regexall("Oracle-Linux-${var.node_linux_version}-20[0-9]*.*-OKE-${local.k8s_version_only}", source.source_name)) > 0], 0)) : (var.node_pool_image_type == "oke" && length(regexall("GPU", var.worker_shape)) > 0) ? (element([for source in local.node_pool_image_ids : source.image_id if length(regexall("Oracle-Linux-${var.node_linux_version}-Gen[0-9]-GPU-20[0-9]*.*-OKE-${local.k8s_version_only}", source.source_name)) > 0], 0)) : (var.node_pool_image_type == "oke" && length(regexall("A1", var.worker_shape)) > 0) ? (element([for source in local.node_pool_image_ids : source.image_id if length(regexall("Oracle-Linux-${var.node_linux_version}-aarch64-20[0-9]*.*-OKE-${local.k8s_version_only}", source.source_name)) > 0], 0)) : null
}

output "Selected_Image" {
  value = local.image_id
}

Also, the autoscaler_image lookup function wouldn't match if we added v1.30.2 as a key to match the autoscaler value posted in your OCI documents:
phx.ocir.io/oracle/oci-cluster-autoscaler**:1.30.2-10**

output "autoscaler_deployment" { value = "${lookup(local.autoscaler_image, var.kubernetes_version)}" }

So we have to alter the kubernetes_version to match the OCI OKE nodepool worker-image-node value to match the available images. See attached terraform logs from OCI's Stack Resource Manager