DocsICP FilingConsole
官方文档
首页

Reserved node pools

更新时间：
复制 MD 格式
产品详情
我的收藏

A reserved node pool is a resource management feature in E-HPC that optimizes cluster node scaling. It pre-creates and binds a fixed set of IP address and hostname mappings, enabling quick reuse of these preallocated resources during dynamic scaling to improve the efficiency and stability of cluster management.

Using reserved node pools in resource provisioning

Provisioning compute node resources for an E-HPC cluster using a reserved node pool differs slightly from standard node addition, as described below.

  • Contact

    • Both approaches acquire network resources from a vSwitch to scale out.

    • The overall scaling workflow remains unchanged.

  • Differences

    • Instead of selecting a random IP from a vSwitch for scaling, you select specific preassigned IPs and hostnames from a reserved node pool.

    • Instead of being unaware of subnet usage by cluster nodes, you can clearly see which nodes in the reserved node pool are allocated or idle.

Prerequisites

The cluster must be in the Running state.

Using reserved node pools in a cluster

Create a reserved node pool

  1. Go to the Cluster Details 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.

    4. On the Cluster List page, find the cluster that you want to manage and click the Cluster Name.

  2. In the left-side navigation pane, choose Nodes > Node.

  3. Click Preset Node Pool. On the Add Preset Node Pool page, configure the following parameters.

    Parameter

    Description

    Name

    Enter a name for the reserved node pool.

    Note

    Hostnames for nodes created from a reserved node pool follow this format: <prefix>[index,3]<postfix>. The postfix parameter supports setting an AD zone name.

    vSwitch

    Select the vSwitch to which the reserved node pool belongs.

    Hostname Prefix

    Enter the starting characters for reserved node hostnames to distinguish them.

    Hostname Suffix

    Enter the ending characters for reserved node hostnames to distinguish them.

    Pool Capacity

    • A single reserved node pool can contain up to 500 nodes.

    • The number of nodes in a reserved node pool cannot exceed the available IP address count of its associated vSwitch.

  4. Click OK.

    Note

    Creating or updating a reserved node pool takes time. Check the pool status using the ListReservedNodePools API operation.

    On the Create reserved node pool tab of the Manage reserved node pools dialog box, enter a Name, select a vSwitch, enter a Hostname prefix and Hostname suffix, and set the Pool capacity (default value is 1).

Resource provisioning with reserved node pools

The following sections describe how to use reserved node pools for resource provisioning through manual node scaling and queue-based autoscaling.

Note

If the requested number of nodes exceeds the reserved node pool capacity, provisioning uses the full pool capacity.

Scale out nodes using a reserved node pool

  1. Log on to the Elastic High Performance Computing console and go to the Cluster Details page. In the navigation pane on the left, choose Nodes and Queues > Node.

  2. Click Add Node. On the Add Node page, configure the following parameters.

    • Basic settings

      Configuration item

      Description

      Destination Queue

      Select an existing queue in the cluster.

      Nodes

      Set the number of nodes to scale out.

    • Node configuration

      Configuration item

      Description

      Select Node Type

      Select only New Node.

      Inter-node interconnection

      Select the network connectivity method between nodes.

      Use Preset Node Pool

      Select an existing reserved node pool. The system automatically selects unallocated IPs and hostnames from the pool to create compute nodes.

      Instance type Group

      Click Add an instance specification to select node specifications.

      If autoscaling is disabled, you can add only one instance type. If autoscaling is enabled, you can add multiple instance types.

      Instance RAM role

      Attach a RAM role to grant nodes permission to access Alibaba Cloud services.

      We recommend using the default role AliyunECSInstanceForEHPCRole created by the system.

  3. Select I understand: By default, newly added nodes have deletion protection enabled and are not affected by queue scaling configurations, then click Confirm Add.

    Note

    New compute nodes are automatically installed with cluster software and initialized. Existing compute nodes remain unaffected.

    You can check the status of scaled-out nodes in the node list. When a node shows as Running, scaling is complete.

Create or configure a queue using a reserved node pool

Note

Relationships between reserved node pools and queues:

  • A single cluster can have multiple reserved node pools, but each reserved node pool belongs to only one cluster.

  • A single queue can bind to one reserved node pool, and one reserved node pool can serve multiple queues.

  • Each reserved node pool maps to exactly one vSwitch.

  1. Log on to the Elastic High Performance Computing console and go to the Cluster Details page. In the navigation pane on the left, choose Nodes and Queues > Queue.

  2. Click Create Queue. On the Create Queue page, configure the following parameters.

    • Basic settings

      Configuration item

      Description

      Queue Name

      Enter a queue name that meets these requirements:

      • Length: 1–15 characters.

      • Allowed characters: uppercase letters (A–Z), lowercase letters (a–z), digits (0–9), and underscores (_).

      Automatic queue scaling

      Choose whether to enable Auto Scale. If enabled, you can further choose whether to enable Auto Grow and Auto Shrink as needed.

      When autoscaling is enabled, the system automatically adds or removes compute nodes based on configuration and real-time workload.

      Queue Compute Nodes

      Set the number of nodes in the queue.

      • If autoscaling is disabled, set the initial number of compute nodes.

      • If autoscaling is enabled, set the minimum and maximum allowed node counts.

        Important

        If you set the minimum node count to a non-zero value, the cluster retains that many nodes during scale-in—even if they are idle. Set this value carefully to avoid unnecessary resource waste and costs.

    • Configure queue node settings

      Configuration item

      Description

      Inter-node interconnection

      Select the network connectivity method between nodes.

      Use Preset Node Pool

      Select an existing reserved node pool. The system automatically selects unallocated IPs and hostnames from the pool to create compute nodes.

      Instance type Group

      Click Add an instance specification to select node specifications.

      If autoscaling is disabled, you can add only one instance type. If autoscaling is enabled, you can add multiple instance types.

    • Autoscaling

      Configuration item

      Description

      Scaling Policy

      Select a scaling policy. Only Supply Priority Strategy is supported. This policy attempts to create compute nodes that meet your specifications 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, meaning no limit.

      If you have cost constraints, set this value to ensure scaling stays within expected limits.

      Host RAM role

      Attach a RAM role to grant nodes permission to access Alibaba Cloud services.

      We recommend using the default role AliyunECSInstanceForEHPCRole created by the system.

  3. Click Save.

    On the queue list page, click the 2024-07-31_11-49-59 icon to refresh. If the target queue appears, creation succeeded.

Modify or delete a reserved node pool

Modify a reserved node pool

  1. Go to the Cluster Details 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.

    4. On the Cluster List page, find the cluster that you want to manage and click the Cluster Name.

  2. In the left-side navigation pane, choose Nodes > Node.

  3. Click Preset Node Pool. On the Manage Preset Node Pools page, edit the following properties.

    Modify property

    Description

    Name

    Change the reserved node pool name.

    Pool Capacity

    Adjust the number of reserved node resources. Reducing capacity releases some IPs. Increasing capacity uses more available IPs, but cannot exceed the vSwitch's available IP address count.

    Remove Preset Node

    Remove reserved nodes that are unallocated or in conflict state.

    Download List

    Download a CSV file of the reserved node list for batch import into DNS or a scheduler.

  4. Click Confirm.

Delete a reserved node pool

  1. Go to the Cluster Details 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.

    4. On the Cluster List page, find the cluster that you want to manage and click the Cluster Name.

  2. In the left-side navigation pane, choose Nodes > Node.

  3. Click Preset Node Pool. On the Manage Preset Node Pools page, click Delete Node Pool to remove it.

    Note

    You can delete up to 100 reserved nodes at a time.

Previous：Configure auto scaling of nodesNext：Create an accelerated image