Merge pull request GoogleCloudPlatform#4853 from GoogleCloudPlatform/release-candidate

Release candidate: v1.73.0

Author: aslam-quad

.github/workflows/multi-approvers.yml

@@ -15,6 +15,15 @@
 name: 'multi-approvers'
 
 on:
+  pull_request:
+    types:
+    - 'opened'
+    - 'edited'
+    - 'reopened'
+    - 'synchronize'
+    - 'ready_for_review'
+    - 'review_requested'
+    - 'review_request_removed'
   pull_request_review:
     types:
     - 'submitted'
@@ -31,7 +40,6 @@ concurrency:
 
 jobs:
   multi-approvers:
-    if: github.event_name == 'pull_request_review' && (github.event.review.state == 'approved' ||  github.event.action == 'dismissed')
     uses: 'abcxyz/pkg/.github/workflows/multi-approvers.yml@main'
     with:
       org-members-path: 'GoogleCloudPlatform/cluster-toolkit/develop/cluster-toolkit-writers.json'

.yamllint

@@ -8,7 +8,6 @@ yaml-files:
 ignore:
 - modules/management/kubectl-apply/manifests/*
 - examples/gke-a4x/nvidia-dra-driver.yaml
-- examples/gke-a4x/nccl-rdma-installer.yaml
 
 rules:
   # Defaults

cluster-toolkit-writers.json

@@ -394,5 +394,8 @@
   },
   {
     "login": "akiki-liang0"
+  },
+  {
+    "login": "simrankaurb"
   }
 ]

cmd/root.go

@@ -53,7 +53,7 @@ HPC deployments on the Google Cloud Platform.`,
 				logging.Fatal("cmd.Help function failed: %s", err)
 			}
 		},
-		Version:     "v1.72.0",
+		Version:     "v1.73.0",
 		Annotations: annotation,
 	}
 )

community/examples/gke-tpu-v6/README.md

@@ -248,3 +248,48 @@ Once deployed, the `Lustre` filesystem is available to the cluster as a `Persist
     ```
 
 The pod will start, and the Managed Lustre filesystem will be available inside the container at `/mnt/lustre`.
+
+### Understanding Hyperdisk Balanced Integration
+The blueprint also supports [Hyperdisk Balanced](https://docs.cloud.google.com/compute/docs/disks/hyperdisks), Google Cloud's high-performance, persistent block storage solution.
+
+To enable Hyperdisk Balanced integration, you must make these changes before deploying:
+
+1. Ensure the GKE cluster is configured to support standard Persistent Disks (the Hyperdisk CSI driver runs automatically once enabled). Verify the `gke-tpu-v6-cluster` module setting `enable_persistent_disk_csi: true` is present.
+
+2. Find the section commented `--- HYPERDISK BALANCED ADDITIONS ---`. Uncomment the entire block containing the following two modules:
+   * `hyperdisk-balanced-setup`: This module creates a **StorageClass** and a **PersistentVolumeClaim (PVC)** that will dynamically provision a Hyperdisk Balanced volume in your cluster.
+   * `fio-bench-job-hyperdisk`: This job is pre-configured to mount the Hyperdisk volume and run performance tests.
+
+After making these changes, run the `gcluster deploy` command as usual.
+
+#### Testing the Hyperdisk Balanced Mount
+
+1. Connect to your cluster:
+
+    ```sh
+    gcloud container clusters get-credentials DEPLOYMENT_NAME --region=REGION --project_id=PROJECT_ID
+    ```
+
+    Replace the `DEPLOYMENT_NAME`,`REGION` and `PROJECT_ID` with the ones used in the blueprint.
+2. Apply the generated FIO Job manifest, whose path is provided in the final deployment instructions.
+
+    ```sh
+    kubectl apply -f <path/to/fio-benchmark.yaml>
+    ```
+
+    The job created in the cluster will be named `fio-benchmark`.
+  
+3. Monitor the job until it completes and obtain the list of pods:
+
+    ```bash
+    kubectl get jobs 
+    kubectl get pods
+    ```
+
+4. View the logs of the completed pod to check the benchmark results:
+
+    ```bash
+    kubectl logs <pod-name>
+    ```
+
+The logs of the pod verifies the disk is mounted successfully and performs a mixed I/O test to validate the disk's provisioned performance.

community/examples/gke-tpu-v6/gke-tpu-v6-advanced.yaml

@@ -170,6 +170,7 @@ deployment_groups:
       configure_workload_identity_sa: true
       enable_gcsfuse_csi: true
       enable_managed_lustre_csi: true
+      enable_persistent_disk_csi: true # enable Hyperdisk for the cluster
       master_authorized_networks:
       - cidr_block: $(vars.authorized_cidr) # Allows your machine to run the kubectl command. Required for multi network setup.
         display_name: "kubectl-access-network"
@@ -397,3 +398,75 @@ deployment_groups:
         rm -rf /{training,checkpoint}-data/fio-benchmarks-${TAG}
 
     outputs: [instructions]
+
+  # # --- HYPERDISK BALANCED ADDITIONS ---
+  # # To enable Hyperdisk-balanced support please uncomment this hyperdisk-balanced-setup and fio-bench-job-template module.
+  # # Set up storage class and persistent volume claim for Hyperdisk
+  # - id: hyperdisk-balanced-setup
+  #   source: modules/file-system/gke-storage
+  #   use: [gke-tpu-v6-cluster]
+  #   settings:
+  #     storage_type: Hyperdisk-balanced
+  #     access_mode: ReadWriteOnce
+  #     sc_volume_binding_mode: Immediate
+  #     sc_reclaim_policy: Delete
+  #     sc_topology_zones: [$(vars.zone)]
+  #     pvc_count: 1
+  #     capacity_gb: 100
+
+  # # This is an example job that will install and run an `fio`benchmark against the hyperdisk volumes.
+  # # For more FIO tests, see https://cloud.google.com/compute/docs/disks/benchmark-hyperdisk-performance
+  # - id: fio-bench-job-hyperdisk
+  #   source: modules/compute/gke-job-template
+  #   use:
+  #   - gke-tpu-v6-pool
+  #   - hyperdisk-balanced-setup
+  #   settings:
+  #     name: fio-benchmark
+  #     image: ubuntu:latest
+  #     security_context:  # to make sure the job have enough access to install the fio packages
+  #     - key: runAsUser
+  #       value: 0
+  #     - key: runAsGroup
+  #       value: 100
+  #     - key: fsGroup
+  #       value: 100
+  #     command:
+  #     - bash
+  #     - -c
+  #     - |
+
+  #       set -eux
+
+  #       cleanup() {
+  #         # This function will be called on script exit
+  #         if [ -n "${TAG:-}" ]; then
+  #           echo "--- Cleaning up temporary directories for tag ${TAG} ---"
+  #           rm -rf "/data/hyperdisk-balanced-pvc-0/fio-benchmarks-${TAG}"
+  #         fi
+  #         }
+  #       trap cleanup EXIT
+
+  #       export DEBIAN_FRONTEND=noninteractive
+
+  #       # Install fio
+  #       apt update -y && apt install -y fio
+
+  #       # Use a tag to create a unique path for tests
+  #       TAG=`date +%s`
+
+  #       # Verify mountpoints
+  #       df -h
+  #       mountpoint /data/hyperdisk-balanced-pvc-0
+
+  #       # Create temporary directory for fio benchmarks
+  #       mkdir -p "/data/hyperdisk-balanced-pvc-0/fio-benchmarks-${TAG}"
+
+  #       # Perform hyperdisk balanced performance (Mixed IOPS) test
+  #       fio --name=hyperdisk-balanced-iops --ioengine=libaio --iodepth=256 --rw=randrw \
+  #       --bs=4k --direct=1 --size=10G --numjobs=16 --group_reporting --time_based --runtime=300s \
+  #       --ramp_time=10s --iodepth_batch_submit=256 --iodepth_batch_complete_max=256 \
+  #       --directory="/data/hyperdisk-balanced-pvc-0/fio-benchmarks-${TAG}" --filename_format=fiotest-balanced-iops
+  #     node_count: 1
+
+  #   outputs: [instructions]

community/modules/compute/htcondor-execute-point/versions.tf

@@ -29,6 +29,6 @@ terraform {
   }
 
   provider_meta "google" {
-    module_name = "blueprints/terraform/hpc-toolkit:htcondor-execute-point/v1.72.0"
+    module_name = "blueprints/terraform/hpc-toolkit:htcondor-execute-point/v1.73.0"
   }
 }

community/modules/compute/mig/versions.tf

@@ -22,6 +22,6 @@ terraform {
     }
   }
   provider_meta "google" {
-    module_name = "blueprints/terraform/hpc-toolkit:mig/v1.72.0"
+    module_name = "blueprints/terraform/hpc-toolkit:mig/v1.73.0"
   }
 }

community/modules/compute/schedmd-slurm-gcp-v6-nodeset-dynamic/versions.tf

@@ -17,6 +17,6 @@
 terraform {
   required_version = ">= 1.3"
   provider_meta "google" {
-    module_name = "blueprints/terraform/hpc-toolkit:schedmd-slurm-gcp-v6-nodeset-dynamic/v1.72.0"
+    module_name = "blueprints/terraform/hpc-toolkit:schedmd-slurm-gcp-v6-nodeset-dynamic/v1.73.0"
   }
 }

community/modules/compute/schedmd-slurm-gcp-v6-nodeset-tpu/versions.tf

@@ -18,6 +18,6 @@ terraform {
   required_version = ">= 1.3"
 
   provider_meta "google" {
-    module_name = "blueprints/terraform/hpc-toolkit:schedmd-slurm-gcp-v6-nodeset-tpu/v1.72.0"
+    module_name = "blueprints/terraform/hpc-toolkit:schedmd-slurm-gcp-v6-nodeset-tpu/v1.73.0"
   }
 }