Create a public cloud Standard Edition E-HPC cluster in the console-Elastic High Performance Computing(E-HPC)-阿里云帮助中心

A public cloud Standard Edition cluster runs in the cloud and consists of components such as Elastic Compute Service (ECS) instances and shared file systems. You manage the availability of cluster services yourself. This topic describes how to create a public cloud cluster in the console.

Background information

An E-HPC public cloud Standard Edition cluster includes the following components:

Control plane node: One ECS instance that runs the scheduler and domain account service to manage job scheduling and user information.
Compute nodes: Multiple ECS instances grouped into queues for management. These nodes support scaling and run jobs.
Logon node: One ECS instance with the Login component deployed and bound to an elastic IP address (EIP) for remote cluster access.
Shared storage: Supports mounting NAS and CPFS file systems to share data such as job data and software data.

Important

When you create an E-HPC cluster, the system automatically creates resources such as ECS instances, which may incur fees. For more information, see Billing overview.
After creating an E-HPC cluster, do not adjust individual cluster nodes in the ECS console unless necessary. Use the E-HPC console instead.

For more information about E-HPC clusters, see Overview.

Prerequisites

You have created a service-linked role. The first time you log on to the E-HPC console, the system prompts you to create the E-HPC service-linked role.
You have created a virtual private cloud (VPC) and vSwitches. For more information, see Create a VPC and Create a vSwitch.
You have created storage resources. E-HPC clusters support mounting NAS and CPFS file systems. Choose as needed.
- To mount NAS: Activate the NAS service and create a NAS file system and mount target. For more information, see Create a file system and Add a mount target.
- To mount CPFS-NFS: Activate the CPFS service and create a CPFS file system, protocol service, and export directory. For more information, see Create a file system, Manage protocol services, and Manage export directories.

Create manually

Step 1: Open the Create Cluster page

Go to the Create Cluster page.

Step 2: Configure the cluster

On the Cluster Configuration page, configure the cluster network, type, scheduler, and other settings.

Basic settings

Configuration item	Description
Region	Select the region where the cluster resides.
Network and Availability Zone	Select the VPC and vSwitch for the cluster. Note Cluster nodes consume IP addresses from the selected vSwitch. Ensure the number of available IP addresses in the vSwitch exceeds the required number of nodes.
Security Group	Security groups control inbound and outbound traffic for the cluster and its nodes. Automatically created security groups include rules that enable communication between cluster nodes. Select the type of automatically created security group as needed. For differences between basic and advanced security groups, see Basic security groups vs. advanced security groups.

Cluster type

This cluster type includes one control plane node and multiple compute nodes. Select a scheduler type and configure the control plane node.

Configuration item	Description
Series	Select Standard Edition.
Deployment Mode	Select Public cloud cluster.
Cluster Type	Select a scheduler type. Supported schedulers for HPC scenarios include Slurm, OpenPBS.
Management node	The control plane node is an ECS instance that runs the scheduler and domain account service. Select an appropriate configuration based on your business scenario and cluster scale. Billing Method Select how to pay for the control plane node. For billing details, see Instance billing. Pay-as-you-go: Post-payment based on actual usage duration. Spot instances are not supported. Subscription: Prepayment billed by week, month, or year. Instance Specification Select an appropriate control plane node type. Recommended configurations for different cluster scales are as follows: If the number of compute nodes ≤ 100, use at least 16 vCPUs and 64 GiB memory. If 100 < number of compute nodes ≤ 500, use at least 32 vCPUs and 128 GiB memory. If the number of compute nodes > 500, use at least 64 vCPUs and 256 GiB memory. Image After selecting an image type, choose a specific image. Different images correspond to different operating systems. The system deploys cluster nodes based on your selected image. Note Custom images have the following limitations: Only custom images created from Alibaba Cloud official images or imported CentOS images are supported. When importing an image, select Run validation after import. Otherwise, the E-HPC console cannot recognize the image. Do not use custom images created from existing E-HPC cluster nodes. Doing so causes errors when creating compute nodes. Do not modify the yum repository configuration in the custom image. Doing so prevents cluster creation or scaling. The mount paths for NAS file systems (mounted using the mount command) in custom images must not include the `/home` or `/opt` directories. Storage Select the system disk type for the control plane node and whether to attach a data disk. For information about disk types and performance, see Disks overview. Hyper-Threading CPU hyper-threading is enabled by default. Disable it if your workload requires better performance. Note After cluster creation, the control plane node automatically binds to the instance RAM role `AliyunECSInstanceForEHPCRole`. This role enables core features such as automatic scaling. Do not unbind or replace this role in the ECS console. To extend API call permissions, see E-HPC service role.

Custom Options

Configuration item	Description
Scheduler	Select the scheduler software to deploy based on your cluster type and the image configured for the control plane node.
Domain Account	Select the domain account service to deploy for the cluster.
Domain name resolution	Keep the default setting.
Specify a cluster post-processing script	Specify a script to process result data or perform other operations after compute tasks complete.
Maximum number of cluster nodes	Set the maximum number of nodes allowed in the cluster. This works together with maximum cores to control cluster size.
Maximum number of cores in the cluster	Set the maximum number of cores allowed in the cluster. This works together with maximum nodes to control cluster size.
Cluster Deletion Protection	Enable or disable deletion protection. If enabled, you must disable this feature before releasing the cluster to prevent accidental deletion.

Resource group

Resource groups help organize resources. For more information, see Resource groups. By default, clusters belong to the default resource group. You can change this as needed.

Step 3: Configure compute nodes and queues

On the Compute Node and Queue page, configure queues.

Queues group compute nodes for management. You can specify a queue when running jobs. The cluster includes one default queue (comp). Click Add more queues to add more queues. Configure the following for each queue:

Basic settings

Configuration item

Description

Automatic queue scaling

Enable or disable Auto Scale. If enabled, you can further enable Auto Grow and Auto Shrink as needed.

With auto scaling enabled, the system automatically adds or removes compute nodes based on configuration and real-time load.

Queue Compute Nodes

Set the number of nodes in the queue.

If auto scaling is disabled, set the initial number of compute nodes.
If auto scaling is enabled, set the minimum and maximum number of nodes.

Important
If you set the minimum node count to a non-zero value, the queue retains that number of nodes during scale-in—even idle nodes remain. Set this value carefully to avoid unnecessary resource waste and costs.

Select queue node configuration

If auto scaling is enabled or if the initial node count is greater than zero, configure the following so the system can create compute nodes.

Configuration item	Description
Inter-node interconnection	Select the network connectivity method between nodes. VPC network: Nodes communicate over the VPC network. eRDMA network: If nodes use ERI-supported instance types, they communicate over the eRDMA network. Note Only some instance types support ERI. For more information, see eRDMA overview and Enable eRDMA on enterprise-grade instances.
Use Preset Node Pool	Select a preset node pool. The system automatically assigns IP addresses and hostnames from unallocated nodes in the pool to create compute nodes. Note Using a preset node pool enables fast reuse of preallocated resources during scale-out. For more information, see Use preset node pools in clusters.
vSwitch	Select the vSwitch for the nodes. The system automatically assigns IP addresses from the available vSwitch CIDR block.
Instance type Group	Click Add an instance specification to select node specifications. If auto scaling is disabled, you can add only one instance type. If enabled, you can add multiple instance types.

Important

You can select multiple vSwitches and instance types as alternatives to avoid creation failures due to inventory shortages. During node creation, the system attempts to create nodes starting from the first vSwitch's zone and in the order of specified instance types until it meets the required node count. The final instance types may vary based on inventory availability.

Auto scaling

Configuration item	Description
Scaling Policy	Select a scaling strategy. Only Supply Priority Strategy is supported, which attempts to create compute nodes in the order of configured vSwitches and zones.
Maximum number of single expansion nodes	Set the maximum number of nodes added or removed in each scale-out or scale-in cycle. The default value is 0, which means no limit. If you have cost constraints, set this value to ensure scale-out stays within expected limits.
Hostname Prefix	Specify the starting characters for node hostnames to distinguish nodes.
Hostname Suffix	Specify the ending characters for node hostnames to distinguish nodes.
Host RAM role	Bind a RAM role to nodes so they can access Alibaba Cloud services. We recommend using the default role AliyunECSInstanceForEHPCRole created by the system.

Step 4: Configure shared file storage

On the Shared File Storage page, configure storage.

By default, the control plane node mounts file systems to /home and /opt as shared storage directories. To mount file systems to other directories, click Add more storage and configure accordingly. For each directory, configure the following file system information:

Note

The /home and /opt directories do not support mounting different file system directories.

Configuration item	Description
Type	Select the file system type to mount. General-purpose NAS: Mount a General-purpose NAS file system. Extreme: Mount an Extreme NAS file system. Parallel file CPFS: Mount a CPFS file system using the NFS protocol.
File System	Select the file system ID and mount target. Ensure the file system has available mount target capacity.
File System Directory	Enter the file system directory to mount.
Mount Options	Select the mount protocol.

Step 5: Configure software and service components

On the Software and Service Component page, configure software and components.

Click Add software and select the software to install in the dialog box. E-HPC provides commonly used HPC software. Select as needed.

Click Add Service Component and select a component in the dialog box. Then configure component parameters.

Note

Only the Login component is currently supported.

Public cloud clusters include the Login component by default for public network remote access. Component parameters are as follows:

Configuration	configuration item	Description
Login component custom parameters	SSH	Set the port, protocol, and allowed IP CIDR block for SSH connections to the cluster.
	VNC	Set the port, protocol, and allowed IP CIDR block for VNC connections to the cluster.
	CLIENT	Set the port, protocol, and allowed IP CIDR block for client connections to the cluster.
Component deployment resources	EIP instance	Bind an EIP to the ECS instance that runs the Login component for public network access. You can automatically create an EIP or select an existing one.
Component deployment resources	ECS instance	Set the instance type for the ECS instance that runs the Login component.

Note

After creation, the logon node automatically binds to the instance RAM role AliyunECSInstanceForEHPCRole. This role ensures features like Web Portal work properly. Do not unbind or replace this role in the ECS console. To extend API call permissions, see E-HPC service role.

Step 6: Confirm configuration

On the Confirm configuration page, review your settings and configure the cluster name and logon credentials.

Configuration item	Description
Cluster Name	Enter a name. This name appears in the cluster list for easy identification.
Cluster Password-free	Allow root users to switch from the control plane node to compute nodes without a password. Important Enabling this feature configures one-way passwordless logon from the control plane node to all compute nodes. Passwordless logon from compute nodes to the control plane node is not supported. Proceed with caution.
Login Credentials	Select credentials for cluster logon. Only Custom Password is supported.
Set Password, Confirm Password	Enter a password for cluster logon. All cluster nodes use this password as the root user logon password by default.

After completing the configuration, read the Service Agreement, confirm the billing information, and click Create Cluster.

Create from template

E-HPC supports creating clusters quickly and in batches using templates. Templates define basic parameters for cluster creation. You can use E-HPC-provided templates or create custom templates.

Create a cluster from a public template

Go to the Cluster List page.
1. Log on to the E-HPC console.
2. In the left part of the top navigation bar, select a region.
3. In the left-side navigation pane, click Cluster.
On the Cluster List page, click Cluster Templates.
In the dialog box, select a template and click Create Cluster.

Templates are grouped as follows: The General-purpose clusters group includes SLURM and SLURM Serverless templates. The Big data clusters group includes the FastMR template. The Climate and weather clusters group includes the WRF template. Each template card includes a version selector and a Create Cluster button.
Review the configuration and enter information such as the cluster name.
- Under Configuration Summary, the default template configuration appears. To modify settings, click Edit and update the relevant parameters.
- Under Manage Settings, complete the configuration as prompted.
Read the Service Agreement, confirm the billing information, and click Create Cluster.

Create a cluster from a custom template

Create a custom template locally.

This topic uses this (template) as an example. Modify the (parameter)s according to your actual needs.

### Basic cluster settings
Region: cn-hangzhou                            # Cluster region. Optional. If omitted, the console auto-fills the region.
ClusterName: "TestClusterName"                 # Cluster name. Optional. If omitted, the system generates a name such as SLURM-Region-DATESTAMP.
ClusterDescription: "XXXXX"                    # Cluster description. Optional.
ClusterCategory: "Standard"                    # Cluster edition. Required. Valid values: ['Standard', 'Serverless', 'SuperComputing'].
ClusterVpcId: ""                               # Cluster VPC ID. Optional. If omitted, the console auto-selects a valid value for the region.
ClusterVSwitchId: ""                           # Control plane node vSwitch ID. Optional. If omitted, the console auto-selects a valid value for the VPC.
IsEnterpriseSecurityGroup: true                # Use advanced security group. Takes effect only if SecurityGroupId is empty. Optional. Default: false (basic security group).
SecurityGroupId: sg-bp1gje9ip78z7v6zy203       # Security group ID. If empty, the system creates one. Optional. Default: empty (auto-create).
ClusterCustomConfiguration:                    # Custom post-installation script. Optional.
  Script: oss://                               # OSS URL of the script.
  Args: arg1 arg2                              # Script arguments.
MaxCount: 1000                                 # Maximum nodes. Optional. Default: 1000.
MaxCoreCount: 100000                           # Maximum vCPUs. Optional. Default: 100000.
DeletionProtection: true                       # Deletion protection. Optional. Default: true (enabled).
ResourceGroupId: rg-acfm2xumdifd3ri            # Resource group ID. Optional. If omitted, the console auto-selects a valid value for the account.
Tags:                                          # Cluster tags. Optional.
  - Key: String
    Value: String
### Cluster control services
Manager:                                     # Control plane node module.
  Scheduler:                                 # Scheduler module.
    Type: "SLURM"                            # Scheduler type. Optional. Default: SLURM.
    Version: "22"                            # Scheduler version. Optional. Default: 22.
  DirectoryService:                          # Account service module.
    Type: "NIS"                              # Account service type. Optional. Default: NIS.
    Version: "x.x.x"                         # Domain account version. Optional.
  DNS:                                       # Domain name resolution module.
    Type: "NIS"                              # DNS service type. Optional. Default: NIS.
    Version: "x.x.x"                         # DNS service version. Optional.
  ManagerNode:                               # Control plane node instance.
    InstanceType: "ecs.c7.xlarge"            # Instance type. Required for non-managed clusters.
    ImageId: "m-xxxxxx"                      # Instance image. Required for non-managed clusters.
    InstanceChargeType: "PostPaid"           # Billing method. Optional. Valid values: PostPaid, Subscription. Default: PostPaid.
    PeriodUnit: "Month"                      # Subscription duration unit. Required only for Subscription billing.
    Period: 1                                # Subscription duration. Required only for Subscription billing.
    AutoRenew: false                         # Auto-renewal. Required only for Subscription billing.
    AutoRenewPeriod: 1                       # Auto-renewal period. Required only for Subscription billing.
    SpotStrategy: "SpotWithPriceGo"          # Spot instance strategy. Ignored for control plane nodes.
    SpotPriceLimit: 0.5                      # Spot price limit. Ignored for control plane nodes.
    Duration: 1                              # Spot instance duration. Ignored for control plane nodes.
    SystemDisk:                              # System disk. Optional.
      Category: "cloud_essd"                 # Disk category. Optional. Default: cloud_essd.
      Size: 40                               # Disk size (GiB). Optional. Default: 40.
      Level: "PL0"                           # Performance level. Optional. Default: PL0.
    DataDisks:                               # Data disks. Optional.
      - Category: "cloud_essd"               # Disk category. Optional. Default: cloud_essd.
        Size: 40                             # Disk size (GiB). Optional. Default: 40.
        Level: "PL0"                         # Performance level. Optional. Default: PL0.
        DeleteWithInstance: false            # Delete disk with instance. Optional. Default: false.
    EnableHT: false                          # Enable hyper-threading. Optional. Default: true.
### Compute queues and node configuration      # Optional. Queue configuration.
Queues:                                      # 
  - Name: workq                              # Queue name. Optional.
    EnableScaleOut: false                     # Enable scale-out. Optional. Default: false.
    EnableScaleIn: false                      # Enable scale-in. Optional. Default: false.
    MinCount: 0                              # Minimum nodes. Optional.
    MaxCount: 500                            # Maximum nodes. Optional.
    InitialCount: 0                          # Initial nodes. Optional.
    InterConnect: erdma                      # Interconnect type. Optional. Valid values: VPC, eRDMA.
    VSwitchIds:                              # vSwitch IDs. Optional.
      - "vsw-xxxxxxx"
      - "vsw-xxxxxxx"        
    ComputeNodes:                            # Compute node configuration. Optional.
      - InstanceType: "ecs.c7.xlarge"            # Instance type. Required for non-managed clusters.
        ImageId: "m-xxxxxx"                      # Instance image. Required for non-managed clusters.
        InstanceChargeType: "PostPaid"           # Billing method. Optional. Valid values: PostPaid, Subscription. Default: PostPaid.
        PeriodUnit: "Month"                      # Subscription duration unit. Required only for Subscription billing.
        Period: 1                                # Subscription duration. Required only for Subscription billing.
        AutoRenew: false                         # Auto-renewal. Required only for Subscription billing.
        AutoRenewPeriod: 1                       # Auto-renewal period. Required only for Subscription billing.
        SpotStrategy: "SpotWithPriceGo"          # Spot instance strategy. Ignored for control plane nodes.
        SpotPriceLimit: 0.5                      # Spot price limit. Ignored for control plane nodes.
        Duration: 1                              # Spot instance duration. Ignored for control plane nodes.
        SystemDisk:                              # System disk. Optional.
          Category: "cloud_essd"                 # Disk category. Optional. Default: cloud_essd.
          Size: 40                               # Disk size (GiB). Optional. Default: 40.
          Level: "PL0"                           # Performance level. Optional. Default: PL0.
        DataDisks:                               # Data disks. Optional.
          - Category: "cloud_essd"               # Disk category. Optional. Default: cloud_essd.
            Size: 40                             # Disk size (GiB). Optional. Default: 40.
            Level: "PL0"                         # Performance level. Optional. Default: PL0.
            DeleteWithInstance: false            # Delete disk with instance. Optional. Default: false.
        EnableHT: false                          # Enable hyper-threading. Optional. Default: true.   
    AllocationStrategy: "PriorityInstanceType"   # Scaling strategy. Optional. Valid values: supply priority, cost priority.
    RamRole: "xxxxxx"                            # RAM role name for nodes. Optional.
    HostNamePrefix: "xxxxx"                        # Hostname prefix. Optional.
    HostNameSuffix: "xxxxx"                        # Hostname suffix. Optional.
    KeepAliveNodes:                                # Exception node list. Optional.
      - compute000
      - compute001
      - compute002
### Shared storage
SharedStorage:
  - MountDirectory: "/home"                    # Mount directory. Optional.
    FileSystemId: "xxxx"                       # File system ID. Optional.
    NASDirectory: "/"                          # File system directory. Optional.
    MountTargetDomain: "xxxxxx"                # Mount target. Optional.
    ProtocolType: "NFS"                        # Protocol. Optional.
    MountOptions: "xxxxx"                      # Mount options. Optional.
### Software
AdditionalPackages:                            # Software list. Optional.
  - Name: "LAMMPS"                             # Software name.
    Version: "xxxx"                            # Software version.
  - Name: "Gromacs"
    Version: "xxx"
### Components
Addons:
  - Name: "LoginNode"                          # Component name.    
    Version: "xxxxxx"                          # Component version.
    ServicesSpec: "JSON String"                # Custom service parameters.
    ResourcesSpec: "JSON String"               # Custom resource parameters.

Go to the Cluster List page.
1. Log on to the E-HPC console.
2. In the left part of the top navigation bar, select a region.
3. In the left-side navigation pane, click Cluster.
On the Cluster List page, click Cluster Templates.
In the dialog box, click Import local template and upload your custom template file.
In the Edit Cluster Template dialog box, verify your custom template and click Confirm and Create.
On the Create Cluster page, verify the configuration and click Create Cluster.

References

After you create a cluster, you need to create a user to submit jobs. For detailed steps, see User Management and Job Overview.