Merge pull request #2544 from GoogleCloudPlatform/release-candidate

Release v1.33.0

Author: tpdownes

cmd/root.go

@@ -52,7 +52,7 @@ HPC deployments on the Google Cloud Platform.`,
 				logging.Fatal("cmd.Help function failed: %s", err)
 			}
 		},
-		Version:     "v1.32.1",
+		Version:     "v1.33.0",
 		Annotations: annotation,
 	}
 )

community/examples/AMD/README.md

@@ -1,5 +1,10 @@
 # AMD solutions for the HPC Toolkit
 
+> [!NOTE]
+> This document uses Slurm-GCP v5. A newer [blueprint](./hpc-amd-slurm-v6.yaml)
+> using v6 has been published. This documentation will be updated and the v5
+> blueprint deprecated.
+
 ## AMD-Optimized Slurm Cluster
 
 This example provisions a Slurm cluster using the AMD-based Computed Optimized

community/examples/AMD/hpc-amd-slurm-v6.yaml

@@ -0,0 +1,227 @@
+# Copyright 2024 Google LLC
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+---
+blueprint_name: hpc-amd-slurm-v6
+
+vars:
+  project_id:  ## Set GCP Project ID Here ##
+  deployment_name: amd-v6
+  region: us-east4
+  zone: us-east4-c
+
+deployment_groups:
+- group: primary
+  modules:
+  - id: network
+    source: modules/network/vpc
+
+  - id: homefs
+    source: modules/file-system/filestore
+    use: [network]
+    settings:
+      local_mount: /home
+
+  - id: swfs
+    source: modules/file-system/filestore
+    use: [network]
+    settings:
+      local_mount: /sw
+
+  - id: spack-setup
+    source: community/modules/scripts/spack-setup
+    settings:
+      install_dir: /sw/spack
+
+  - id: spack-execute
+    source: community/modules/scripts/spack-execute
+    use: [spack-setup]
+    settings:
+      log_file: /var/log/spack.log
+      data_files:
+      - destination: /tmp/projections-config.yaml
+        content: |
+          modules:
+            default:
+              tcl:
+                hash_length: 0
+                all:
+                  conflict:
+                    - '{name}'
+                projections:
+                  all: '{name}/{version}-{compiler.name}-{compiler.version}'
+      - destination: /tmp/slurm-external-config.yaml
+        content: |
+          packages:
+            slurm:
+              externals:
+                - spec: slurm@23-11-3
+                  prefix: /usr/local
+              buildable: False
+      - destination: /sw/spack/openfoam_env.yaml
+        content: |
+          spack:
+            definitions:
+            - compilers:
+              - gcc@13.1.0
+            - mpis:
+              - openmpi@4.1.3+legacylaunchers+pmi fabrics=none schedulers=slurm
+            - packages:
+              - flex@2.6.4
+            - mpi_packages:
+              - openfoam-org@8 ^flex@2.6.4 target=zen3
+            specs:
+            - matrix:
+              - - $mpis
+              - - $%compilers
+            - matrix:
+              - - $packages
+              - - $%compilers
+            - matrix:
+              - - $mpi_packages
+              - - $%compilers
+              - - $^mpis
+            concretizer:
+              unify: when_possible
+      commands: |
+        spack config --scope defaults add config:build_stage:/opt/spack_build_stage
+        spack config --scope defaults add -f /tmp/projections-config.yaml
+        spack config --scope site add -f /tmp/slurm-external-config.yaml
+        spack config --scope site add concretizer:targets:host_compatible:false
+
+        # gcc 12.1.0 is known to have runtime failures with OpenFOAM 8
+        spack install gcc@13.1.0 %gcc@8.5.0 target=x86_64
+        spack load gcc@13.1.0 %gcc@8.5.0 target=x86_64
+        spack compiler find --scope site
+
+        if ! spack env list | grep -q openfoam; then
+          spack env create openfoam /sw/spack/openfoam_env.yaml
+          spack env activate openfoam
+          spack concretize
+          spack install
+        fi
+
+  - id: spack-startup
+    source: modules/scripts/startup-script
+    settings:
+      runners:
+      - $(spack-execute.spack_runner)
+      - type: shell
+        destination: shutdown.sh
+        content: |
+          #!/bin/bash
+          if [ ! -f /etc/block_auto_shutdown ]; then
+                  touch /etc/block_auto_shutdown
+                  shutdown -h +1
+          fi
+
+  - id: slurm_startup
+    source: modules/scripts/startup-script
+    settings:
+      runners:
+      - $(spack-setup.spack_runner)
+      # the following installation of AOCC may be automated in the future
+      # with a clear direction to the user to read the EULA at
+      # https://developer.amd.com/aocc-compiler-eula/
+      - type: data
+        destination: /var/tmp/install_aocc.sh
+        content: |
+          #!/bin/bash
+          source /sw/spack/share/spack/setup-env.sh
+          spack install aocc@3.2.0 +license-agreed
+          spack load aocc@3.2.0
+          spack compiler find --scope site
+          spack -d install -v openmpi@4.1.3 %aocc@3.2.0 +legacylaunchers +pmi schedulers=slurm
+      - type: data
+        destination: /var/tmp/openfoam_test.sh
+        content: |
+          #!/bin/bash
+          # the following line works around a problem activating environments
+          # before directory is accessed
+          ls -lha /sw/spack/var/spack/environments/openfoam/ &>/dev/null
+          spack env activate openfoam
+          DIR=$HOME/openfoam_test
+          mkdir -p $DIR
+          cd $DIR
+          cp -fr $WM_PROJECT_DIR/tutorials/incompressible/simpleFoam/motorBike .
+          cd motorBike
+          ./Allrun
+
+  - id: spack_builder
+    source: modules/compute/vm-instance
+    use: [network, swfs, spack-startup]
+    settings:
+      name_prefix: spack-builder
+      machine_type: c2d-standard-16
+      instance_image:
+        # these images must match the images used by Slurm modules below because
+        # we are building OpenMPI with PMI support in libraries contained in
+        # Slurm installation
+        family: slurm-gcp-6-4-hpc-rocky-linux-8
+        project: schedmd-slurm-public
+
+  - id: low_cost_nodeset
+    source: community/modules/compute/schedmd-slurm-gcp-v6-nodeset
+    use: [network]
+    settings:
+      machine_type: c2d-standard-4
+      node_count_dynamic_max: 10
+      bandwidth_tier: gvnic_enabled
+      enable_placement: false
+
+  - id: low_cost_partition
+    source: community/modules/compute/schedmd-slurm-gcp-v6-partition
+    use: [low_cost_nodeset]
+    settings:
+      partition_name: lowcost
+
+  - id: compute_nodeset
+    source: community/modules/compute/schedmd-slurm-gcp-v6-nodeset
+    use: [network]
+    settings:
+      machine_type: c2d-standard-112
+      node_count_dynamic_max: 50
+      bandwidth_tier: gvnic_enabled
+      enable_placement: true
+
+  # Because is_default is set to true, jobs will run on this partition unless an
+  # alternative partition is specified using, for example, "srun -p lowcost"
+  - id: compute_partition
+    source: community/modules/compute/schedmd-slurm-gcp-v6-partition
+    use: [compute_nodeset]
+    settings:
+      partition_name: compute
+      is_default: true
+
+  - id: slurm_login
+    source: community/modules/scheduler/schedmd-slurm-gcp-v6-login
+    use: [network]
+    settings:
+      # need at least 8 physical cores to run OpenFOAM test
+      machine_type: c2d-standard-16
+      name_prefix: login
+
+  - id: slurm_controller
+    source: community/modules/scheduler/schedmd-slurm-gcp-v6-controller
+    use:
+    - network
+    - homefs
+    - swfs
+    - low_cost_partition
+    - compute_partition
+    - slurm_login
+    settings:
+      machine_type: c2d-standard-4
+      login_startup_script: $(slurm_startup.startup_script)
+      login_startup_scripts_timeout: 21600

community/examples/hpc-build-slurm-image.yaml

@@ -23,7 +23,7 @@ vars:
   image_build_machine_type: n2d-standard-16
   build_from_image_family: hpc-rocky-linux-8
   build_from_image_project: cloud-hpc-image-public
-  build_from_git_ref: 6.4.3
+  build_from_git_ref: 6.4.4
   built_image_family: my-custom-slurm
   built_instance_image:
     family: $(vars.built_image_family)

community/examples/hpc-slurm-gromacs.yaml

@@ -110,7 +110,7 @@ deployment_groups:
     use: [network1]
     settings:
       name_prefix: login
-      disable_login_public_ips: false
+      enable_login_public_ips: true
 
   - id: slurm_controller
     source: community/modules/scheduler/schedmd-slurm-gcp-v6-controller
@@ -121,6 +121,6 @@ deployment_groups:
     - homefs
     - appsfs
     settings:
-      disable_controller_public_ips: false
+      enable_controller_public_ips: true
       login_startup_script: $(script.startup_script)
       login_startup_scripts_timeout: 21600

community/examples/hpc-slurm-local-ssd-v6.yaml

@@ -103,7 +103,7 @@ deployment_groups:
     settings:
       name_prefix: login
       machine_type: n1-standard-4
-      disable_login_public_ips: false
+      enable_login_public_ips: true
 
   - id: slurm_controller
     source: community/modules/scheduler/schedmd-slurm-gcp-v6-controller
@@ -116,4 +116,4 @@ deployment_groups:
         suspend_timeout: 300
         no_comma_params: false
       machine_type: n1-standard-4
-      disable_controller_public_ips: false
+      enable_controller_public_ips: true

community/examples/hpc-slurm-ramble-gromacs.yaml renamed to community/examples/hpc-slurm-ramble-gromacs-v6.yaml

@@ -1,4 +1,4 @@
-# Copyright 2022 Google LLC
+# Copyright 2024 Google LLC
 #
 # Licensed under the Apache License, Version 2.0 (the "License");
 # you may not use this file except in compliance with the License.
@@ -32,7 +32,7 @@ deployment_groups:
   # Source is an embedded module, denoted by "modules/*" without ./, ../, /
   # as a prefix. To refer to a local module, prefix with ./, ../ or /
   # Example - ./modules/network/vpc
-  - id: network1
+  - id: network
     source: modules/network/vpc
 
   ## Install Scripts
@@ -68,6 +68,7 @@ deployment_groups:
       commands: |
         ramble workspace create gromacs -c /opt/apps/gromacs_scaling.yaml -t /opt/apps/execute_experiment.tpl
         ramble workspace activate gromacs
+        ramble workspace concretize -f
         ramble workspace info -v
 
   - id: login-startup
@@ -81,7 +82,7 @@ deployment_groups:
           ramble:
             variables:
               mpi_command: 'mpirun -n {n_ranks} -ppn {processes_per_node}'
-              batch_submit: 'sbatch {execute_experiment}'
+              batch_submit: '{execute_experiment}'
               processes_per_node: 30
             applications:
               gromacs:
@@ -92,7 +93,6 @@ deployment_groups:
                         variables:
                           n_nodes: [1, 2, 4, 8, 16]
             spack:
-              concretized: false
               packages: {}
               environments: {}
       - type: data
@@ -109,37 +109,39 @@ deployment_groups:
     source: modules/scripts/startup-script
     settings:
       runners:
+      - $(spack-install.spack_runner)
       - $(ramble-setup.ramble_runner)
 
-  - id: compute_node_group
-    source: community/modules/compute/schedmd-slurm-gcp-v5-node-group
+  - id: compute_nodeset
+    source: community/modules/compute/schedmd-slurm-gcp-v6-nodeset
+    use: [network]
     settings:
       node_count_dynamic_max: 20
       bandwidth_tier: gvnic_enabled
 
   - id: compute_partition
-    source: community/modules/compute/schedmd-slurm-gcp-v5-partition
-    use:
-    - network1
-    - compute_node_group
+    source: community/modules/compute/schedmd-slurm-gcp-v6-partition
+    use: [compute_nodeset]
     settings:
       partition_name: compute
 
-  - id: slurm_controller
-    source: community/modules/scheduler/schedmd-slurm-gcp-v5-controller
-    use:
-    - network1
-    - compute_partition
-    - access-startup
+  - id: slurm_login
+    source: community/modules/scheduler/schedmd-slurm-gcp-v6-login
+    use: [network]
     settings:
-      disable_controller_public_ips: false
+      name_prefix: login
+      machine_type: n2-standard-4
+      enable_login_public_ips: true
 
-  - id: slurm_login
-    source: community/modules/scheduler/schedmd-slurm-gcp-v5-login
+  - id: slurm_controller
+    source: community/modules/scheduler/schedmd-slurm-gcp-v6-controller
     use:
-    - network1
-    - slurm_controller
-    - login-startup
+    - network
+    - compute_partition
+    - slurm_login
     settings:
-      machine_type: n2-standard-4
-      disable_login_public_ips: false
+      enable_controller_public_ips: true
+      login_startup_script: $(login-startup.startup_script)
+      controller_startup_script: $(access-startup.startup_script)
+      controller_startup_scripts_timeout: 1000
+      login_startup_scripts_timeout: 1000