Merge pull request #2619 from GoogleCloudPlatform/release-candidate

Release v1.34.0

cmd/destroy.go

@@ -16,12 +16,15 @@
 package cmd
 
 import (
+	"bufio"
 	"fmt"
 	"hpc-toolkit/pkg/config"
+	"hpc-toolkit/pkg/logging"
 	"hpc-toolkit/pkg/modulewriter"
 	"hpc-toolkit/pkg/shell"
 	"os"
 	"path/filepath"
+	"strings"
 
 	"github.com/spf13/cobra"
 )
@@ -62,6 +65,11 @@ func runDestroyCmd(cmd *cobra.Command, args []string) {
 		group := bp.Groups[i]
 		groupDir := filepath.Join(deplRoot, string(group.Name))
 
+		if err := shell.ImportInputs(groupDir, artifactsDir, bp); err != nil {
+			logging.Error("failed to import inputs for group %q: %v", group.Name, err)
+			// still proceed with destroying the group
+		}
+
 		var err error
 		switch group.Kind() {
 		case config.PackerKind:
@@ -72,9 +80,16 @@ func runDestroyCmd(cmd *cobra.Command, args []string) {
 		case config.TerraformKind:
 			err = destroyTerraformGroup(groupDir)
 		default:
-			err = fmt.Errorf("group %s is an unsupported kind %s", groupDir, group.Kind().String())
+			err = fmt.Errorf("group %q is an unsupported kind %q", groupDir, group.Kind().String())
+		}
+
+		if err != nil {
+			logging.Error("failed to destroy group %q:\n%s", group.Name, renderError(err, *ctx))
+			if i == 0 || !destroyChoice(bp.Groups[i-1].Name) {
+				logging.Fatal("destruction of %q failed", deplRoot)
+			}
 		}
-		checkErr(err, ctx)
+
 	}
 
 	modulewriter.WritePackerDestroyInstructions(os.Stdout, packerManifests)
@@ -88,3 +103,31 @@ func destroyTerraformGroup(groupDir string) error {
 
 	return shell.Destroy(tf, getApplyBehavior())
 }
+
+func destroyChoice(nextGroup config.GroupName) bool {
+	switch getApplyBehavior() {
+	case shell.AutomaticApply:
+		return true
+	case shell.PromptBeforeApply:
+		// pass; proceed with prompt
+	default:
+		return false
+	}
+
+	reader := bufio.NewReader(os.Stdin)
+	for {
+		fmt.Printf("Do you want to delete the next group %q [y/n]?: ", nextGroup)
+
+		in, err := reader.ReadString('\n')
+		if err != nil {
+			logging.Fatal("%v", err)
+		}
+
+		switch strings.ToLower(strings.TrimSpace(in)) {
+		case "y":
+			return true
+		case "n":
+			return false
+		}
+	}
+}

cmd/root.go

@@ -52,7 +52,7 @@ HPC deployments on the Google Cloud Platform.`,
 				logging.Fatal("cmd.Help function failed: %s", err)
 			}
 		},
-		Version:     "v1.33.0",
+		Version:     "v1.34.0",
 		Annotations: annotation,
 	}
 )

community/examples/AMD/README.md

@@ -1,9 +1,8 @@
 # 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.
+> This document uses Slurm-GCP v6. If you want to use Slurm-GCP v5 version you
+> scan refer [blueprint](./hpc-amd-slurm-v5-legacy.yaml)
 
 ## AMD-Optimized Slurm Cluster
 

community/examples/AMD/hpc-amd-slurm-v6.yaml → community/examples/AMD/hpc-amd-slurm-v5-legacy.yaml

@@ -1,4 +1,4 @@
-# Copyright 2024 Google LLC
+# Copyright 2022 Google LLC
 #
 # Licensed under the Apache License, Version 2.0 (the "License");
 # you may not use this file except in compliance with the License.
@@ -13,36 +13,37 @@
 # limitations under the License.
 
 ---
-blueprint_name: hpc-amd-slurm-v6
+blueprint_name: hpc-amd-slurm
 
 vars:
   project_id:  ## Set GCP Project ID Here ##
-  deployment_name: amd-v6
+  deployment_name: amd-v5
   region: us-east4
   zone: us-east4-c
 
 deployment_groups:
 - group: primary
   modules:
-  - id: network
+  - id: network1
     source: modules/network/vpc
 
   - id: homefs
     source: modules/file-system/filestore
-    use: [network]
+    use: [network1]
     settings:
       local_mount: /home
 
   - id: swfs
     source: modules/file-system/filestore
-    use: [network]
+    use: [network1]
     settings:
       local_mount: /sw
 
   - id: spack-setup
     source: community/modules/scripts/spack-setup
     settings:
       install_dir: /sw/spack
+      spack_ref: v0.18.1
 
   - id: spack-execute
     source: community/modules/scripts/spack-execute
@@ -66,15 +67,15 @@ deployment_groups:
           packages:
             slurm:
               externals:
-                - spec: slurm@23-11-3
+                - spec: slurm@22-05-8
                   prefix: /usr/local
               buildable: False
       - destination: /sw/spack/openfoam_env.yaml
         content: |
           spack:
             definitions:
             - compilers:
-              - gcc@13.1.0
+              - gcc@10.3.0
             - mpis:
               - openmpi@4.1.3+legacylaunchers+pmi fabrics=none schedulers=slurm
             - packages:
@@ -101,8 +102,9 @@ deployment_groups:
         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
+        # gcc 10.3.0 is the earliest copy of gcc with Zen 3 support
+        spack install gcc@10.3.0 %gcc@4.8.5 target=x86_64
+        spack load gcc@10.3.0 %gcc@4.8.5 target=x86_64
         spack compiler find --scope site
 
         if ! spack env list | grep -q openfoam; then
@@ -160,68 +162,70 @@ deployment_groups:
 
   - id: spack_builder
     source: modules/compute/vm-instance
-    use: [network, swfs, spack-startup]
+    use: [network1, swfs, spack-startup]
     settings:
       name_prefix: spack-builder
       machine_type: c2d-standard-16
+      disable_public_ips: true
       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
+        family: slurm-gcp-5-11-hpc-centos-7
         project: schedmd-slurm-public
 
-  - id: low_cost_nodeset
-    source: community/modules/compute/schedmd-slurm-gcp-v6-nodeset
-    use: [network]
+  - id: low_cost_node_group
+    source: community/modules/compute/schedmd-slurm-gcp-v5-node-group
     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]
+    source: community/modules/compute/schedmd-slurm-gcp-v5-partition
+    use:
+    - network1
+    - low_cost_node_group
     settings:
       partition_name: lowcost
+      enable_placement: false
 
-  - id: compute_nodeset
-    source: community/modules/compute/schedmd-slurm-gcp-v6-nodeset
-    use: [network]
+  - id: compute_node_group
+    source: community/modules/compute/schedmd-slurm-gcp-v5-node-group
     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
+  # 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]
+    source: community/modules/compute/schedmd-slurm-gcp-v5-partition
+    use:
+    - network1
+    - compute_node_group
     settings:
       partition_name: compute
+      enable_placement: true
       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
+    source: community/modules/scheduler/schedmd-slurm-gcp-v5-controller
     use:
-    - network
+    - network1
     - 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
+
+  - id: slurm_login
+    source: community/modules/scheduler/schedmd-slurm-gcp-v5-login
+    use:
+    - network1
+    - slurm_controller
+    - slurm_startup
+    settings:
+      # need at least 8 physical cores to run OpenFOAM test
+      machine_type: c2d-standard-16