feat: DA updates:<br>- the logic to handle various kinds of machine_types has been improved to help prevent failures in certain use cases<br>- the default_secondary_storage input is now removed. As GPUs will be included in additional worker pools, secondary storage can be provided there. (#683)

Author: imprateeksh

ibm_catalog.json

@@ -274,9 +274,6 @@
             {
               "key": "default_worker_pool_labels"
             },
-            {
-              "key": "default_worker_pool_secondary_storage"
-            },
             {
               "key": "enable_autoscaling_for_default_pool"
             },

main.tf

@@ -43,7 +43,6 @@ locals {
 
 # Local block to verify validations for OCP AI Addon.
 locals {
-
   # get the total workers per pool
   workers_per_pool = {
     for pool in var.worker_pools :
@@ -56,8 +55,8 @@ locals {
   worker_specs = {
     for pool in var.worker_pools :
     pool.pool_name => {
-      cpu_count = tonumber(split("x", split(".", pool.machine_type)[1])[0])
-      ram_count = tonumber(split("x", split(".", pool.machine_type)[1])[1])
+      cpu_count = tonumber(regex("^.*?(\\d+)x(\\d+)", pool.machine_type)[0])
+      ram_count = tonumber(regex("^.*?(\\d+)x(\\d+)", pool.machine_type)[1])
       is_gpu    = contains(["gx2", "gx3", "gx4"], split(".", pool.machine_type)[0])
     }
   }

solutions/fully-configurable/README.md

@@ -65,7 +65,6 @@ The following resources are provisioned by this example:
 | <a name="input_default_worker_pool_labels"></a> [default\_worker\_pool\_labels](#input\_default\_worker\_pool\_labels) | A set of key-value labels assigned to the worker pool for identification. For Example: { env = "prod", team = "devops" } | `map(string)` | `{}` | no |
 | <a name="input_default_worker_pool_machine_type"></a> [default\_worker\_pool\_machine\_type](#input\_default\_worker\_pool\_machine\_type) | The machine type for worker nodes.[Learn more](https://cloud.ibm.com/docs/openshift?topic=openshift-vpc-flavors) | `string` | `"bx2.8x32"` | no |
 | <a name="input_default_worker_pool_operating_system"></a> [default\_worker\_pool\_operating\_system](#input\_default\_worker\_pool\_operating\_system) | The operating system installed on the worker nodes. [Learn more](https://cloud.ibm.com/docs/openshift?topic=openshift-vpc-flavors) | `string` | `"RHEL_9_64"` | no |
-| <a name="input_default_worker_pool_secondary_storage"></a> [default\_worker\_pool\_secondary\_storage](#input\_default\_worker\_pool\_secondary\_storage) | The secondary storage attached to the worker nodes. Secondary storage is immutable and can't be changed after provisioning. | `string` | `null` | no |
 | <a name="input_default_worker_pool_workers_per_zone"></a> [default\_worker\_pool\_workers\_per\_zone](#input\_default\_worker\_pool\_workers\_per\_zone) | Number of worker nodes in each zone of the cluster. | `number` | `2` | no |
 | <a name="input_disable_outbound_traffic_protection"></a> [disable\_outbound\_traffic\_protection](#input\_disable\_outbound\_traffic\_protection) | Whether to allow public outbound access from the cluster workers. This is only applicable for OCP 4.15 and later. | `bool` | `false` | no |
 | <a name="input_disable_public_endpoint"></a> [disable\_public\_endpoint](#input\_disable\_public\_endpoint) | Whether access to the public service endpoint is disabled when the cluster is created. Does not affect existing clusters. You can't disable a public endpoint on an existing cluster, so you can't convert a public cluster to a private cluster. To change a public endpoint to private, create another cluster with this input set to `true`. | `bool` | `true` | no |

solutions/fully-configurable/main.tf

@@ -157,7 +157,6 @@ locals {
       operating_system  = var.default_worker_pool_operating_system
       labels            = var.default_worker_pool_labels
       minSize           = var.default_pool_minimum_number_of_nodes
-      secondary_storage = var.default_worker_pool_secondary_storage
       maxSize           = var.default_pool_maximum_number_of_nodes
       enableAutoscaling = var.enable_autoscaling_for_default_pool
       boot_volume_encryption_kms_config = {

solutions/fully-configurable/variables.tf

@@ -126,6 +126,10 @@ variable "default_worker_pool_machine_type" {
   type        = string
   description = "The machine type for worker nodes.[Learn more](https://cloud.ibm.com/docs/openshift?topic=openshift-vpc-flavors)"
   default     = "bx2.8x32"
+  validation {
+    condition     = length(regexall("^[a-z0-9]+(?:\\.[a-z0-9]+)*\\.\\d+x\\d+(?:\\.[a-z0-9]+)?$", var.default_worker_pool_machine_type)) > 0
+    error_message = "Invalid value provided for the machine type."
+  }
 }
 
 variable "default_worker_pool_workers_per_zone" {
@@ -146,13 +150,6 @@ variable "default_worker_pool_labels" {
   default     = {}
 }
 
-variable "default_worker_pool_secondary_storage" {
-  type        = string
-  description = "The secondary storage attached to the worker nodes. Secondary storage is immutable and can't be changed after provisioning."
-  default     = null
-  nullable    = true
-}
-
 variable "enable_autoscaling_for_default_pool" {
   type        = bool
   description = "Set `true` to enable automatic scaling of worker based on workload demand."

variables.tf

@@ -116,6 +116,15 @@ variable "worker_pools" {
     ])
     error_message = "Invalid operating system for the given OCP version. Ensure the OS is compatible with the OCP version. Supported compatible OCP version and OS are v4.14: (REDHAT_8_64); v4.15: (REDHAT_8_64, RHCOS) ; v4.16 and v4.17: (REDHAT_8_64, RHCOS, RHEL_9_64)"
   }
+
+  validation {
+    condition = alltrue([
+      for pool in var.worker_pools :
+      length(regexall("^[a-z0-9]+(?:\\.[a-z0-9]+)*\\.\\d+x\\d+(?:\\.[a-z0-9]+)?$", pool.machine_type)) > 0
+    ])
+    error_message = "Invalid value provided for one or more machine type."
+  }
+
 }
 
 variable "worker_pools_taints" {