Skip to content

Latest commit

 

History

History
186 lines (147 loc) · 13 KB

README.md

File metadata and controls

186 lines (147 loc) · 13 KB
Raw

Description

This module creates a compute partition that can be used as input to the schedmd-slurm-gcp-v5-controller.

The partition module is designed to work alongside the schedmd-slurm-gcp-v5-node-group module. A partition can be made up of one or more node groups, provided either through use (preferred) or defined manually in the node_groups variable.

Warning: updating a partition and running terraform apply will not cause the slurm controller to update its own configurations (slurm.conf) unless enable_reconfigure is set to true in the partition and controller modules.

Example

The following code snippet creates a partition module with:

  • 2 node groups added via use.
    • The first node group is made up of machines of type c2-standard-30.
    • The second node group is made up of machines of type c2-standard-60.
    • Both node groups have a maximum count of 200 dynamically created nodes.
  • partition name of "compute".
  • connected to the network1 module via use.
  • nodes mounted to homefs via use.
- id: node_group_1
  source: community/modules/compute/schedmd-slurm-gcp-v5-node-group
  settings:
    name: c30
    node_count_dynamic_max: 200
    machine_type: c2-standard-30

- id: node_group_2
  source: community/modules/compute/schedmd-slurm-gcp-v5-node-group
  settings:
    name: c60
    node_count_dynamic_max: 200
    machine_type: c2-standard-60

- id: compute_partition
  source: community/modules/compute/schedmd-slurm-gcp-v5-partition
  use:
  - network1
  - homefs
  - node_group_1
  - node_group_2
  settings:
    partition_name: compute

For a complete example using this module, see slurm-gcp-v5-cluster.yaml.

Compute VM Zone Policies

The Slurm on GCP partition module allows you to specify additional zones in which to create VMs through bulk creation. This is valuable when configuring partitions with popular VM families and you desire access to more compute resources across zones.

WARNING: Lenient zone policies can lead to additional egress costs when moving large amounts of data between zones in the same region. For example, traffic between VMs and traffic from VMs to shared filesystems such as Filestore. For more information on egress fees, see the Network Pricing Google Cloud documentation.

To avoid egress charges, ensure your compute nodes are created in a single zone by setting var.zone and leaving var.zones to its default value of the empty list.

NOTE: If a new zone is added to the region while the cluster is active, nodes in the partition may be created in that zone. In this case, the partition may need to be redeployed (possible via enable_reconfigure if set) to ensure the newly added zone is denied.

In the zonal example below, the partition's zone implicitly defaults to the deployment variable vars.zone:

vars:
  zone: us-central1-f

- id: zonal-partition
  source: community/modules/compute/schedmd-slurm-gcp-v5-partition

In the example below, we enable creation in additional zones:

vars:
  zone: us-central1-f

- id: multi-zonal-partition
  source: community/modules/compute/schedmd-slurm-gcp-v5-partition
  settings:
    zones:
    - us-central1-a
    - us-central1-b

Support

The HPC Toolkit team maintains the wrapper around the slurm-on-gcp terraform modules. For support with the underlying modules, see the instructions in the slurm-gcp README.

License

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. 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.

Requirements

Name Version
terraform >= 0.13.0
google >= 3.83

Providers

Name Version
google >= 3.83

Modules

Name Source Version
slurm_partition github.com/GoogleCloudPlatform/slurm-gcp.git//terraform/slurm_cluster/modules/slurm_partition 5.10.6

Resources

Name Type
google_compute_zones.available data source

Inputs

Name Description Type Default Required
deployment_name Name of the deployment. string n/a yes
enable_placement Enable placement groups. bool true no
enable_reconfigure Enables automatic Slurm reconfigure on when Slurm configuration changes (e.g.
slurm.conf.tpl, partition details). Compute instances and resource policies
(e.g. placement groups) will be destroyed to align with new configuration.

NOTE: Requires Python and Google Pub/Sub API.

WARNING: Toggling this will impact the running workload. Deployed compute nodes
will be destroyed and their jobs will be requeued.		 bool false no
exclusive Exclusive job access to nodes. bool true no
is_default Sets this partition as the default partition by updating the partition_conf.
If "Default" is already set in partition_conf, this variable will have no effect.		 bool false no
network_storage An array of network attached storage mounts to be configured on the partition compute nodes. 
list(object({
    server_ip             = string,
    remote_mount          = string,
    local_mount           = string,
    fs_type               = string,
    mount_options         = string,
    client_install_runner = map(string)
    mount_runner          = map(string)
  }))
 [] no
node_groups A list of node groups associated with this partition. See
schedmd-slurm-gcp-v5-node-group for more information on defining a node
group in a blueprint.		 
list(object({
    node_count_static      = number
    node_count_dynamic_max = number
    group_name             = string
    node_conf              = map(string)
    access_config = list(object({
      nat_ip       = string
      network_tier = string
    }))
    additional_disks = list(object({
      disk_name    = string
      device_name  = string
      disk_size_gb = number
      disk_type    = string
      disk_labels  = map(string)
      auto_delete  = bool
      boot         = bool
    }))
    additional_networks = list(object({
      network            = string
      subnetwork         = string
      subnetwork_project = string
      network_ip         = string
      nic_type           = string
      stack_type         = string
      queue_count        = number
      access_config = list(object({
        nat_ip       = string
        network_tier = string
      }))
      ipv6_access_config = list(object({
        network_tier = string
      }))
      alias_ip_range = list(object({
        ip_cidr_range         = string
        subnetwork_range_name = string
      }))
    }))
    bandwidth_tier         = string
    can_ip_forward         = bool
    disable_smt            = bool
    disk_auto_delete       = bool
    disk_labels            = map(string)
    disk_size_gb           = number
    disk_type              = string
    enable_confidential_vm = bool
    enable_oslogin         = bool
    enable_shielded_vm     = bool
    enable_spot_vm         = bool
    gpu = object({
      count = number
      type  = string
    })
    instance_template    = string
    labels               = map(string)
    machine_type         = string
    maintenance_interval = string
    metadata             = map(string)
    min_cpu_platform     = string
    on_host_maintenance  = string
    preemptible          = bool
    reservation_name     = string
    service_account = object({
      email  = string
      scopes = list(string)
    })
    shielded_instance_config = object({
      enable_integrity_monitoring = bool
      enable_secure_boot          = bool
      enable_vtpm                 = bool
    })
    spot_instance_config = object({
      termination_action = string
    })
    source_image_family  = string
    source_image_project = string
    source_image         = string
    tags                 = list(string)
  }))
 [] no
partition_conf Slurm partition configuration as a map.
See https://slurm.schedmd.com/slurm.conf.html#SECTION_PARTITION-CONFIGURATION		 map(string) {} no
partition_name The name of the slurm partition. string n/a yes
partition_startup_scripts_timeout The timeout (seconds) applied to the partition startup script. If
any script exceeds this timeout, then the instance setup process is considered
failed and handled accordingly.

NOTE: When set to 0, the timeout is considered infinite and thus disabled.		 number 300 no
project_id Project in which the HPC deployment will be created. string n/a yes
region The default region for Cloud resources. string n/a yes
slurm_cluster_name Cluster name, used for resource naming and slurm accounting. If not provided it will default to the first 8 characters of the deployment name (removing any invalid characters). string null no
startup_script Startup script that will be used by the partition VMs. string "" no
subnetwork_project The project the subnetwork belongs to. string "" no
subnetwork_self_link Subnet to deploy to. string null no
zone Zone in which to create compute VMs. Additional zones in the same region can be specified in var.zones. string n/a yes
zone_target_shape Strategy for distributing VMs across zones in a region.
ANY
GCE picks zones for creating VM instances to fulfill the requested number of VMs
within present resource constraints and to maximize utilization of unused zonal
reservations.
ANY_SINGLE_ZONE (default)
GCE always selects a single zone for all the VMs, optimizing for resource quotas,
available reservations and general capacity.
BALANCED
GCE prioritizes acquisition of resources, scheduling VMs in zones where resources
are available while distributing VMs as evenly as possible across allowed zones
to minimize the impact of zonal failure.		 string "ANY_SINGLE_ZONE" no
zones Additional nodes in which to allow creation of partition nodes. Google Cloud
will find zone based on availability, quota and reservations.		 set(string) [] no

Outputs

Name Description
partition Details of a slurm partition