cae-slurm.yaml

# 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.
---

#
#           ****************
#######  CAE Solution Blueprint #######
#           ****************
#
# This blueprint features a reference design suited for CAE applications on GCP.
# It sets up the following infrastructure:
#   * Google's H3 VMs, ideally suited for CAE workloads
#   * Google's C3-highmem VM, suited for workloads with 16GB/core requirement
#   * Google's Filestore NFS-based shared storage
#   * Google's Chrome Remote Desktop
#   * SLURM workload scheduler
#
blueprint_name: cae-slurm-v6
vars:
  project_id: ## Set GCP Project ID Here ##
  deployment_name: cae-slurm-v6
  # check here for other regions with H3 deployments: https://cloud.google.com/compute/docs/regions-zones
  # For example
  # region: europe-west4
  # zone: europe-west4-b
  region: us-central1
  zone: us-central1-a
  # Visit https://github.com/GoogleCloudPlatform/slurm-gcp/blob/master/docs/images.md#published-image-family
  # for a list of valid family options with Slurm; note: the image types for the compute nodes
  # and the Chrome Remote Desktop (CRD) need to have the same Slurm base.
  instance_image:
    family: slurm-gcp-6-6-hpc-rocky-linux-8
    project: schedmd-slurm-public

# Documentation for each of the modules used below can be found at
# https://github.com/GoogleCloudPlatform/hpc-toolkit/blob/main/modules/README.md
deployment_groups:

# ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
#
#
# Deployment Group: Setup
#
# Sets up VPC network, persistent NFS shares, dashboard
# ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
- group: setup
  modules:

  ####### Virtual Private Cloud Setup #######
  # This creates a virtual private network for your cloud setup
  - id: network
    source: modules/network/vpc
    settings:
      network_name: cae-slurm-net
      subnetwork_name: primary-subnet

  ####### User Home Storage #######
  # This block creates an NFS file share for /home
  - id: homefs
    source: modules/file-system/filestore
    use: [network]
    settings:
      filestore_tier: BASIC_SSD
      size_gb: 2560
      filestore_share_name: homeshare
      local_mount: /home

  ####### Shared Software Storage #######
  # This block creates NFS file share for shared software installations
  - id: appsfs
    source: modules/file-system/filestore
    use: [network]
    settings:
      filestore_tier: BASIC_SSD
      size_gb: 2560
      filestore_share_name: appsshare
      local_mount: /apps

  ####### Dashboard #######
  # This module activates integration with a dashboard on the Google Cloud Console
  - id: hpc_dash
    source: modules/monitoring/dashboard
    outputs: [instructions]

# ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
#
#
# Deployment Group: Software Installation
#
# This deployment group is a stub for installing software before
# bringing up the actual cluster.
# See the README.md for useful software deployment patterns.
#
# ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
# - group: software_installation
#   modules:

# ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
#
#
# Deployment Group: Cluster
#
# Provisions the actual CAE cluster with compute partitions,
# remote desktop partition and connects to the previously set up
# NFS shares.
# ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
- group: cluster
  modules:

  #######  Scratch #######
  # This block creates an NFS file share for scratch. If you experience an IO bottleneck,
  # consider to use the more performant version HIGH_SCALE_SSD with the following settings:
  - id: scratchfs
    source: modules/file-system/filestore
    use: [network]
    settings:
      filestore_tier: HIGH_SCALE_SSD
      size_gb: 10240    # smallest size for HIGH_SSD_SCALE
      filestore_share_name: scratchshare
      local_mount: /scratch

  # If you require maximum IO performance, you can consider to bring up a dedicated parallel
  # file system, e.g. DDN Exascaler Lustre, Sycomp GPFS, or Intel DAOS.
  # Note: Those solutions may have associated license cost.
  #
  # Please visit here for more information
  # - DDN Exascaler Lustre: https://github.com/GoogleCloudPlatform/hpc-toolkit/blob/main/community/modules/file-system/DDN-EXAScaler/README.md
  # - Sycomp IBM Spectrum Scale: https://console.developers.google.com/marketplace/product/sycomp/sycomp-storage-fueled-by-ibm-spectrum-scale
  # - Intel DAOS: https://github.com/GoogleCloudPlatform/hpc-toolkit/blob/main/community/modules/file-system/Intel-DAOS/README.md

  ######## Remote Desktop(s) #######
  # This block creates chrome remote desktop.
  # For more info: https://github.com/GoogleCloudPlatform/hpc-toolkit/tree/main/community/modules/remote-desktop/chrome-remote-desktop
  - id: remotedesktop
    source: community/modules/remote-desktop/chrome-remote-desktop
    use: [network]
    settings:
      install_nvidia_driver: true
      add_deployment_name_before_prefix: true
      name_prefix: chrome-remote-desktop
      guest_accelerator:
      - type: nvidia-tesla-t4-vws
        count: 1

  ####### Balanced partition #######
  # this block creates a partition uses GCP H3-standard VM for regular jobs with 4GB/core
  - id: h3_nodeset
    source: community/modules/compute/schedmd-slurm-gcp-v6-nodeset
    use: [network]
    settings:
      node_count_dynamic_max: 10
      machine_type: h3-standard-88
      disk_type: 'pd-balanced'
      bandwidth_tier: gvnic_enabled
      allow_automatic_updates: false

  - id: h3_partition
    source: community/modules/compute/schedmd-slurm-gcp-v6-partition
    use: [h3_nodeset]
    settings:
      partition_name: balance
      is_default: true

  ####### High-Mem partition #######
  #  this block creates partition uses GCP C3-highmem VM for jobs with 16GB/core requirement
  - id: c3_nodeset
    source: community/modules/compute/schedmd-slurm-gcp-v6-nodeset
    use: [network]
    settings:
      node_count_dynamic_max: 10
      machine_type: c3-highmem-176
      disk_type: 'pd-balanced'
      bandwidth_tier: tier_1_enabled
      allow_automatic_updates: false

  - id: c3_partition
    source: community/modules/compute/schedmd-slurm-gcp-v6-partition
    use: [c3_nodeset]
    settings:
      partition_name: highmem

  ####### Scheduler: SLURM #######
  # This block creates a SLURM login node
  - id: slurm_login
    source: community/modules/scheduler/schedmd-slurm-gcp-v6-login
    use: [network]
    settings:
      machine_type: n2-standard-8

  ####### Scheduler: SLURM #######
  # This block creates a SLURM controller
  - id: slurm_controller
    source: community/modules/scheduler/schedmd-slurm-gcp-v6-controller
    use:
    - network
    - homefs
    - appsfs
    - scratchfs
    - h3_partition
    - c3_partition
    - slurm_login
    settings:
      machine_type: n2-standard-16
      compute_startup_scripts_timeout: 900
      cloud_parameters:
        resume_rate: 0
        resume_timeout: 900
        suspend_rate: 0
        suspend_timeout: 300
        no_comma_params: false