Merge branch 'dev' into 'master'

6.3.1

See merge request SchedMD/slurm-gcp!82

Author: jvilarru

CHANGELOG.md

@@ -2,6 +2,11 @@
 
 All notable changes to this project will be documented in this file.
 
+## \[6.3.1\]
+
+- Add reserved property for nodeset_tpu
+- update lustre repository url
+
 ## \[6.3.0\]
 
 - Upgrade installed Slurm to 23.02.7

ansible/roles/lustre/vars/redhat-8.yml

@@ -13,7 +13,7 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-lustre_repo_url: https://downloads.whamcloud.com/public/lustre/latest-release/el8.8/client
+lustre_repo_url: https://downloads.whamcloud.com/public/lustre/latest-release/el8.9/client
 
 lustre_packages:
 - lustre-client

scripts/util.py

@@ -1148,6 +1148,10 @@ def enable_public_ip(self):
     def preemptible(self):
         return self._nodeset.preemptible
 
+    @property
+    def reserved(self):
+        return self._nodeset.reserved
+
     @property
     def service_account(self):
         return self._nodeset.service_account
@@ -1277,6 +1281,7 @@ def create_node(self, nodename):
             node.service_account.email = self.nodeset.service_account.email
             node.service_account.scope = self.nodeset.service_account.scopes
         node.scheduling_config.preemptible = self.preemptible
+        node.scheduling_config.reserved = self.reserved
         if self.nodeset.network:
             node.network_config.network = self.nodeset.network
         if self.nodeset.subnetwork:

terraform/slurm_cluster/modules/slurm_nodeset_tpu/README_TF.md

@@ -59,6 +59,7 @@ No modules.
 | <a name="input_preemptible"></a> [preemptible](#input\_preemptible) | Specify whether TPU-vms in this nodeset are preemtible, see https://cloud.google.com/tpu/docs/preemptible for details. | `bool` | `false` | no |
 | <a name="input_preserve_tpu"></a> [preserve\_tpu](#input\_preserve\_tpu) | Specify whether TPU-vms will get preserve on suspend, if set to true, on suspend vm is stopped, on false it gets deleted | `bool` | `true` | no |
 | <a name="input_project_id"></a> [project\_id](#input\_project\_id) | Project ID to create resources in. | `string` | n/a | yes |
+| <a name="input_reserved"></a> [reserved](#input\_reserved) | Specify whether TPU-vms in this nodeset are created under a reservation. | `bool` | `false` | no |
 | <a name="input_service_account"></a> [service\_account](#input\_service\_account) | Service account to attach to the TPU-vm.<br>If none is given, the default service account and scopes will be used. | <pre>object({<br>    email  = string<br>    scopes = set(string)<br>  })</pre> | `null` | no |
 | <a name="input_subnetwork"></a> [subnetwork](#input\_subnetwork) | The name of the subnetwork to attach the TPU-vm of this nodeset to. | `string` | `null` | no |
 | <a name="input_tf_version"></a> [tf\_version](#input\_tf\_version) | Nodeset Tensorflow version, see https://cloud.google.com/tpu/docs/supported-tpu-configurations#tpu_vm for details. | `string` | n/a | yes |

terraform/slurm_cluster/modules/slurm_nodeset_tpu/main.tf

@@ -111,6 +111,10 @@ resource "null_resource" "nodeset_tpu" {
       condition     = sum([var.node_count_dynamic_max, var.node_count_static]) > 0
       error_message = "Sum of node_count_dynamic_max and node_count_static must be > 0."
     }
+    precondition {
+      condition     = !(var.preemptible && var.reserved)
+      error_message = "Nodeset cannot be preemptible and reserved at the same time."
+    }
     precondition {
       condition     = !(var.subnetwork == null && !var.enable_public_ip)
       error_message = "Using the default subnetwork for the TPU nodeset requires enable_public_ip set to true."

terraform/slurm_cluster/modules/slurm_nodeset_tpu/variables.tf

@@ -77,6 +77,12 @@ variable "preemptible" {
   default     = false
 }
 
+variable "reserved" {
+  description = "Specify whether TPU-vms in this nodeset are created under a reservation."
+  type        = bool
+  default     = false
+}
+
 variable "preserve_tpu" {
   description = "Specify whether TPU-vms will get preserve on suspend, if set to true, on suspend vm is stopped, on false it gets deleted"
   type        = bool

terraform/slurm_cluster/modules/slurm_nodeset_tpu/versions.tf

@@ -18,7 +18,13 @@ terraform {
   required_version = "~> 1.2"
 
   required_providers {
-    google = ">= 3.53, < 5.0"
-    null   = "~> 3.0"
+    google = {
+      source  = "hashicorp/google"
+      version = ">= 3.53, < 5.0"
+    }
+    null = {
+      source  = "hashicorp/null"
+      version = "~> 3.0"
+    }
   }
 }