Queues group compute nodes so you can run different job types on separate node pools, with independent scaling rules and resource limits for each group. This topic explains how to create, configure, and delete queues in E-HPC.
Queues are a key dimension in resource monitoring. View the load and performance of compute nodes by queue on the Monitoring page. For more information, see View the monitoring information.
Prerequisites
Before you begin, make sure that:
The cluster is in the Running state
(For queue deletion only) The queue has no compute nodes
Create a queue
Go to the Cluster Details 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 and click its cluster ID.
In the left-side navigation pane, choose Nodes and Queues > Queue.
Click Create Queue and configure the parameters on the Create Queue page. Select queue node configuration This section is required if automatic scaling is enabled or the initial node count is greater than 0.
ImportantSetting Minimum Nodes to a non-zero value retains that number of nodes during scale-in — idle nodes are not released. Set this value carefully to avoid unnecessary costs.
ImportantAdding multiple vSwitches and instance types creates fallback options when inventory is unavailable. When creating a node, the system tries instance types in the configured order, starting in the zone of the first vSwitch. The actual instance specification may vary based on availability.
Basic settings
Parameter Description Queue Name Enter a name for the queue. The name can be 1–15 characters long and contain uppercase letters (A–Z), lowercase letters (a–z), digits (0–9), and underscores (\_). Automatic queue scaling Toggle automatic scaling on or off. When enabled, select Auto Grow, Auto Shrink, or both based on your workload. The system then adds or removes compute nodes automatically based on real-time demand. Queue Compute Nodes Specify the number of compute nodes. Without automatic scaling, set an initial node count. With automatic scaling, set a minimum and maximum node count. Parameter Description Inter-node interconnection Select how nodes communicate: VPC network (over a virtual private cloud) or eRDMA network (elastic Remote Direct Memory Access, for instance types with Elastic RDMA Interfaces (ERIs)). Not all instance types support ERIs. For details, see Overview and Enable eRDMA on an enterprise-level instance. Virtual Switch Select a vSwitch. The system assigns IP addresses to compute nodes from the vSwitch CIDR block. Instance type group Click Add Instance to select an instance type. Without automatic scaling, only one instance type is allowed. With automatic scaling, multiple instance types are allowed. Auto scale
Parameter Description Scaling Policy Currently, only Supply Priority Strategy is supported. Nodes are created in the specified zones following the order of configured vSwitches. Maximum number of single expansion nodes The maximum nodes added or removed per scaling cycle. The default value 0 means no limit. Set this value to control scaling costs. Prefix of Hostnames A hostname prefix that identifies nodes belonging to this queue. Hostname Suffix A hostname suffix that identifies nodes belonging to this queue. Instance RAM role Bind a Resource Access Management (RAM) role to grant nodes access to Alibaba Cloud services. The default role AliyunECSInstanceForEHPCRole is recommended. Click Save.
Click the refresh icon to reload the queue list. The queue is created successfully when it appears in the list.
What's next: Submit a job to the new queue to verify it works end to end.
Configure a queue
To prevent impacts on ongoing business, we recommend that you configure queues when your business is idle.
Go to the Cluster Details 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 and click its cluster ID.
In the left-side navigation pane, choose Nodes and Queues > Queue.
Find the queue and click Edit in the Actions column.
On the Edit Queue page, modify the parameters as needed.
ImportantSetting Minimum Nodes to a non-zero value retains that number of nodes during scale-in — idle nodes are not released. Set this value carefully to avoid unnecessary costs.
ImportantAdding multiple vSwitches and instance types creates fallback options when inventory is unavailable. The system tries instance types in order, starting in the zone of the first vSwitch.
Basic settings
Parameter Description Automatic queue scaling Automatic queue scaling is turned off by default. Toggle automatic scaling on or off. When enabled, select Auto Grow, Auto Shrink, or both. Queue-level configurations take precedence over cluster-level configurations. Queue Compute Nodes Minimum Nodes: 0–1,000. Affects scale-in behavior. Maximum Nodes: 0–5,000. Affects scale-out behavior. The queue maximum cannot exceed the cluster maximum. Select queue node configuration
Parameter Description Inter-node interconnection VPC network or eRDMA network. See Overview and Enable eRDMA on an enterprise-level instance for eRDMA instance type requirements. Virtual Switch Select a vSwitch. The system assigns IP addresses from the vSwitch CIDR block. Instance type group Click Add Instance to select instance types. Multiple instance types are allowed when automatic scaling is enabled. Auto scale
Parameter Description Scaling Policy Currently, only Supply Priority Strategy is supported. Maximum number of single expansion nodes Nodes added or removed per scaling cycle. Default 0 means no limit. Prefix of Hostnames Hostname prefix for nodes in this queue. Hostname Suffix Hostname suffix for nodes in this queue. Instance RAM role RAM role bound to nodes. AliyunECSInstanceForEHPCRole is recommended. Click Save.
Click the refresh icon to reload the queue list and check the Auto Scaling Configurations column. The configuration is updated when the column reflects the new values.
Delete a queue
Before you delete a queue, make sure that the queue does not have compute nodes. Otherwise, you cannot delete the queue. We also recommend deleting queues during low-traffic periods to minimize disruption.
Go to the Cluster Details 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 and click its cluster ID.
In the left-side navigation pane, choose Nodes and Queues > Queue.
Delete the queue:
To delete a single queue, find the queue and click Delete in the Actions column.
To delete multiple queues at once, select the queues and click Batch Delete at the bottom of the page.
In the dialog box, confirm the queue information and click OK.
Click the refresh icon to reload the queue list. The queue is deleted successfully when it no longer appears in the list.