use lustre in a3u a4h

Author: saara-tyagi27

examples/machine-learning/a3-ultragpu-8g/a3ultra-slurm-blueprint.yaml

@@ -28,9 +28,7 @@ vars:
   image_build_machine_type: n2-standard-16
   build_slurm_from_git_ref: 6.10.10
   # Cluster env settings
-  # net0 and filestore ranges must not overlap
   net0_range: 192.168.0.0/19
-  filestore_ip_range: 192.168.32.0/24
   net1_range: 192.168.64.0/18
   rdma_net_range: 192.168.128.0/18
   # Cluster Settings
@@ -126,7 +124,7 @@ deployment_groups:
             "install_cuda": false,
             "install_gcsfuse": true,
             "install_lustre": false,
-            "install_managed_lustre": false,
+            "install_managed_lustre": true,
             "install_nvidia_repo": true,
             "install_ompi": true,
             "allow_kernel_upgrades": false,
@@ -326,43 +324,24 @@ deployment_groups:
         ip_range: $(vars.rdma_net_range)
         region: $(vars.region)
 
+  - id: private_service_access
+    source: modules/network/private-service-access
+    use: [a3ultra-slurm-net-0]
+
   - id: homefs
-    source: modules/file-system/filestore
+    source: modules/file-system/managed-lustre
     use:
     - a3ultra-slurm-net-0
+    - private_service_access
     settings:
-      filestore_tier: HIGH_SCALE_SSD
-      size_gb: 10240
+      size_gib: $(vars.lustre_size_gib)
+      name: $(vars.lustre_instance_id)
       local_mount: /home
-      reserved_ip_range: $(vars.filestore_ip_range)
-      deletion_protection:
-        enabled: true
-        reason: Avoid data loss
+      remote_mount: lustrefs
+      per_unit_storage_throughput: $(vars.per_unit_storage_throughput)
     outputs:
     - network_storage
 
-  # To use Managed Lustre as for the shared /home directory:
-  # 1. Comment out the filestore block above and the`filestore_ip_range` line in the vars block.
-  # 2. Uncomment the managed-lustre and private-service-access blocks
-  # 3. Change the value for "install_managed_lustre" in /var/tmp/slurm_vars.json above to true
-  # - id: private_service_access
-  #   source: modules/network/private-service-access
-  #   use: [a3ultra-slurm-net-0]
-
-  # - id: homefs
-  #   source: modules/file-system/managed-lustre
-  #   use:
-  #   - a3ultra-slurm-net-0
-  #   - private_service_access
-  #   settings:
-  #     size_gib: $(vars.lustre_size_gib)
-  #     name: $(vars.lustre_instance_id)
-  #     local_mount: /home
-  #     remote_mount: lustrefs
-  #     per_unit_storage_throughput: $(vars.per_unit_storage_throughput)
-  #   outputs:
-  #   - network_storage
-
   # The following four modules create and mount a Cloud Storage Bucket with
   # gcsfuse.  They are optional but recommended for many use cases.
   # (Optional) The following creates a GCS bucket that will be mounted

examples/machine-learning/a4-highgpu-8g/a4high-slurm-blueprint.yaml

@@ -30,7 +30,6 @@ vars:
   # Cluster env settings
   # net0 and filestore ranges must not overlap
   net0_range: 192.168.0.0/19
-  filestore_ip_range: 192.168.32.0/24
   net1_range: 192.168.64.0/18
   rdma_net_range: 192.168.128.0/18
   # Cluster Settings
@@ -49,21 +48,15 @@ vars:
   a4h_dws_flex_enabled: false
   a4h_enable_spot_vm: false
 
-  # To enable Managed-Lustre please uncomment this section and fill out the settings.
-  # Additionally, please uncomment the private_service_access and managed-lustre modules.
-  # Managed Lustre is only supported in specific regions and zones
-  # Please refer https://cloud.google.com/managed-lustre/docs/locations
-
   # Managed-Lustre instance name. This should be unique for each deployment.
-  # lustre_instance_id: lustre-instance
+  lustre_instance_id: lustre-instance
 
   # The values of size_gib and per_unit_storage_throughput are co-related
   # Please refer https://cloud.google.com/managed-lustre/docs/create-instance#performance-tiers
   # Storage capacity of the lustre instance in GiB
-  # lustre_size_gib: 36000
-
+  lustre_size_gib: 36000
   # Maximum throughput of the lustre instance in MBps per TiB
-  # per_unit_storage_throughput: 500
+  per_unit_storage_throughput: 500
 
 deployment_groups:
 - group: image-env
@@ -127,7 +120,7 @@ deployment_groups:
             "install_cuda": false,
             "install_gcsfuse": true,
             "install_lustre": false,
-            "install_managed_lustre": false,
+            "install_managed_lustre": true,
             "install_nvidia_repo": true,
             "install_ompi": true,
             "allow_kernel_upgrades": false,
@@ -329,43 +322,26 @@ deployment_groups:
         ip_range: $(vars.rdma_net_range)
         region: $(vars.region)
 
+  # To use Managed Lustre as for the shared /home directory:
+  # 1. Comment out the filestore block above and the`filestore_ip_range` line in the vars block.
+  - id: private_service_access
+    source: modules/network/private-service-access
+    use: [a4high-slurm-net-0]
+
   - id: homefs
-    source: modules/file-system/filestore
+    source: modules/file-system/managed-lustre
     use:
     - a4high-slurm-net-0
+    - private_service_access
     settings:
-      filestore_tier: HIGH_SCALE_SSD
-      size_gb: 10240
+      size_gib: $(vars.lustre_size_gib)
+      name: $(vars.lustre_instance_id)
       local_mount: /home
-      reserved_ip_range: $(vars.filestore_ip_range)
-      deletion_protection:
-        enabled: true
-        reason: Avoid data loss
+      remote_mount: lustrefs
+      per_unit_storage_throughput: $(vars.per_unit_storage_throughput)
     outputs:
     - network_storage
 
-  # To use Managed Lustre as for the shared /home directory:
-  # 1. Comment out the filestore block above and the`filestore_ip_range` line in the vars block.
-  # 2. Uncomment the managed-lustre and private-service-access blocks
-  # 3. Change the value for "install_managed_lustre" in /var/tmp/slurm_vars.json above to true
-  # - id: private_service_access
-  #   source: modules/network/private-service-access
-  #   use: [a4high-slurm-net-0]
-
-  # - id: homefs
-  #   source: modules/file-system/managed-lustre
-  #   use:
-  #   - a4high-slurm-net-0
-  #   - private_service_access
-  #   settings:
-  #     size_gib: $(vars.lustre_size_gib)
-  #     name: $(vars.lustre_instance_id)
-  #     local_mount: /home
-  #     remote_mount: lustrefs
-  #     per_unit_storage_throughput: $(vars.per_unit_storage_throughput)
-  #   outputs:
-  #   - network_storage
-
   # The following four modules create and mount a Cloud Storage Bucket with
   # gcsfuse.  They are optional but recommended for many use cases.
   # (Optional) The following creates a GCS bucket that will be mounted