Add compute nodes to an E-HPC cluster when the current nodes can no longer handle your workload. Scale-out adds capacity without stopping the cluster or affecting nodes and jobs already running.
Prerequisites
Before you begin, make sure that:
Your management node meets the specifications for the target cluster size. See the table below.
A vSwitch is available in the region where the new nodes will reside. To create one, see Work with vSwitches.
Sufficient Elastic Compute Service (ECS) instance quotas exist in the target region. To check or request a quota increase, see Manage ECS quotas.
Recommended management node specifications by cluster size
| Number of compute nodes | Management node specs | Job limits |
|---|---|---|
| 100 or fewer | 16+ vCPUs, 64+ GiB memory | Fewer than 5,000 queued; fewer than 10,000 uncompleted |
| 500 or fewer | 32+ vCPUs, 128+ GiB memory | Fewer than 10,000 queued; fewer than 20,000 uncompleted |
| More than 500 | 64+ vCPUs, 256+ GiB memory | Fewer than 10,000 queued; fewer than 20,000 uncompleted |
For more information on management node sizing, see Overview.
Scale out a cluster
Log on to the E-HPC console.
In the top navigation bar, select a region.
In the left-side navigation pane, click Cluster.
On the Cluster page, find the cluster to scale out, and click Resize.
In the Resize panel, select the type of nodes to add — Create Node, Existing Node, or On-premises Node — and complete the configuration.
Create Node
Configure the parameters below, then click Buy Now.
Parameter Description Zone The zone for the new compute nodes. New nodes and the cluster can be in different zones. vSwitch ID The vSwitch for the new compute nodes. Pricing Model The billing method: Subscription (weekly, monthly, or yearly), Pay-As-You-Go (hourly), or Preemptible Instance (discounted on-demand pricing). For details, see Overview. Quantity And Type of Instances to be Added The number and specifications of compute nodes to add. You can add up to 500 nodes per operation. To add more than 500 nodes, submit a ticket. Image Type Supported types: public image, custom image, shared image, Alibaba Cloud Marketplace image, and community image. Available types depend on the region and your account's image resources. Image The image to deploy on the new nodes. Note the following: - The OS of the new nodes must match that of existing compute nodes in the cluster. For example, if the cluster runs CentOS, only CentOS images are available.
- The major OS version must also match. For example, existing CentOS 7.x nodes require new nodes to run CentOS 7.x.
- If you select a custom image, it must be created from compute nodes — not from logon nodes or management nodes. Using an image from a logon or management node may cause errors during scale-out.
Assign Public IP Address Turn on to allow the new nodes to access the Internet. Then configure the bandwidth billing method and maximum outbound bandwidth. Queue The queue to assign the new nodes to. Hostname Prefix and Hostname Suffix Optional prefix and suffix for the hostnames of the new nodes, to simplify managing large numbers of nodes. System Disk The type and size of the system disk for the new nodes. Add Data Disk Click Add Data Disk to attach additional data disks. Specify the type, size, and quantity. Existing Node
Select one or more stopped ECS instances from the list, then click OK.
Note: Existing compute nodes must be stopped before you can add them. If no nodes appear, check their status in the ECS console.
On-premises Node
Select a queue, enter or import the node information below, then click OK.
Note: On-premises compute nodes cannot be moved between queues after they are added. If you are not using the default queue, create the target queue before adding nodes.
Provide the following information for each node:
Field Description Hostname The hostname of the compute node. Node ID/IP Address The IP address of the compute node. CPU The number of CPUs, as reported by running lscpuon the node.Memory The memory size of the compute node.
Verify the result
After scale-out, the new compute nodes automatically install and initialize the cluster's applications. Existing nodes and jobs running on them are not affected.
To confirm the new nodes are ready:
In the left-side navigation pane, choose Resource Management > Nodes.
From the Cluster drop-down list, select the cluster you scaled out.
From the Node Type drop-down list, select Compute Node.
Confirm that the newly added nodes appear with the status Running.