If your cluster nodes are pay-as-you-go instances, you can start the cluster when you need it and stop it when you do not need it. This helps reduce costs. This topic describes how to start and stop a cluster in the Elastic High Performance Computing (E-HPC) console.
Usage notes
Before you stop a cluster, follow these steps to ensure data security and smooth operations:
Check job status: Make sure that no job is the Running or Pending state. This prevents data loss and job interruption.
Disable auto scaling: Disable the auto scaling feature to prevent nodes from being added while the cluster is being stopped.
NoteIf you stop a cluster when a node is in an unstable state (for example, when the node is being installed or initialized), the node may be in the Abnormal state and unavailable.
Billing
In economical mode, you are not billed for computing resources (CPU and memory), image license fees, and bandwidth fees (for fixed public IP addresses), but you are billed for system disks, data disks, Elastic IP Addresses (EIPs), and shared storage fees. For more information, see Economical mode.
Startup failure may occur as a result of inventory insufficiency caused by computing resource recycling.
Prerequisites
The cluster to be stopped is in the Running or Exception state.
The cluster to be started is in the Stopped state.
Procedure
Go to the Cluster List page.
Log on to the E-HPC console.
In the left part of the top navigation bar, select a region.
In the left-side navigation pane, click Cluster.
On the Cluster List page, find the cluster that you want to manage and click the
icon in the upper-right corner. Start the cluster: Click Start Cluster.
Stop the cluster: Click Stop Cluster and select a Stop Mode as prompted.
NoteIf you stop the cluster in Economical mode, pay-as-you-go nodes in the cluster stop being billed for computing resources, but subscription nodes in the cluster continue being billed. For more information, see Economical mode.
In the dialog box that appears, click OK.