AnalyticDB:Create and manage resource groups

1. Log on to the AnalyticDB for MySQL console. In the upper-left corner of the console, select a region. In the left-side navigation pane, click Clusters. Find the cluster that you want to manage and click the cluster ID.

2. In the navigation pane on the left, choose Cluster Management > Resource Management, and then click the Resource Groups tab. In the upper-right corner of the resource group list, click Create Resource Group.

3. Enter a name for the resource group and select a Job Type.

   • For online scenarios that require high queries per second (QPS) and low response time (RT), select Interactive.

     Interactive resource groups use resident computing resources to execute queries in Massively Parallel Processing (MPP) mode. This provides fast response times, typically at the millisecond level.

   • For offline scenarios that require high throughput, select Job.

     Job resource groups start temporary computing resources and execute queries in Bulk Synchronous Parallel (BSP) mode. This results in slower response times, typically at the second or minute level. The amount of temporary computing resources started is between 0 ACU and the maximum resources of the Job resource group. The specific amount depends on the size of the running task.

   • For heterogeneous computing scenarios, select AI.

     AI resource groups support heterogeneous computing resources such as GPUs and CPUs. They support multiple deployment types, including MLSQL models and Ray-hosted computing.

   Important

   The task type cannot be modified after the resource group is created.

4. The properties to configure vary based on the selected task type. After you configure the properties, click OK.

   Interactive resource group properties

   Parameter	Description
   Engine	XIHE Engine: The resource group supports only XIHE SQL. Spark Engine: The resource group supports only Spark SQL jobs. Spark SQL jobs are processed in an interactive manner. Important The engine cannot be modified after the resource group is created.
   Automatic Stop	When an Interactive resource group is idle for a specified period, the enabled clusters in the resource group are automatically released. An idle state means a few minutes have passed since the last command was executed. Enabling auto-stop reduces resource waste and saves costs. However, when you execute a query again, resources must be restarted, which causes a delay. Important This parameter is available only when Engine is set to Spark.
   Cluster Size	If Engine is set to XIHE: The size of a single cluster in ACUs. You can enter any value. The minimum value is 16 ACU. If Engine is set to Spark: The size of a single cluster, which is the number of ACUs allocated to a Spark application. The minimum value is 24 ACU. Multiple Spark applications can run in each Spark Interactive resource group. The Minimum Clusters and Maximum Clusters parameters specify the number of Spark applications that can run in the resource group. For more information about the mapping between cluster size and the specifications of Spark Drivers and Spark Executors, see Appendix: Mapping between cluster size and Spark Driver and Spark Executor specifications.
   Minimum Clusters Maximum Clusters	Minimum Clusters: The minimum number of clusters that must run in the resource group. The minimum value is 1. Maximum Clusters: The maximum number of clusters to which the resource group can be scaled out. The maximum value is 10. If Minimum Clusters and Maximum Clusters are different, AnalyticDB for MySQL dynamically scales the number of clusters between the minimum and maximum values based on the query payload of the resource group. If Minimum Clusters and Maximum Clusters are the same, AnalyticDB for MySQL starts the specified number of clusters after the resource group is created. This lets you statically control the total computing resources of the resource group. Note If Minimum Clusters or Maximum Clusters is set to 2 or greater, the Multi-Cluster feature is enabled for the resource group. For more information about the Multi-Cluster feature, see Multi-Cluster elastic model.
   Job Resubmission Rules	Delivers queries that exceed the Query Execution Time Threshold to the Target Resource Group for execution. For more information, see Job delivery. Important This parameter is available only when Engine is set to XIHE.
   Spark Configuration	Spark application configuration parameters. These parameters apply to all Spark jobs executed in this resource group. To configure parameters for a specific Spark job, you can set them in the code when you submit the job. For more information about Spark configuration parameters, see Spark application configuration parameters. Important This parameter is available only when Engine is set to Spark.

   Job resource group properties

   Parameter	Description
   Minimum Computing Resources	The minimum value is 0 ACU. Important Min Computing Resources cannot be modified after the resource group is created.
   Maximum Computing Resources	The maximum value that can be set in the console is 1024 ACU, with a step size of 8 ACU. If you need more resources, submit a ticket to contact technical support.
   Spot Instance	Specifies whether to enable spot instances. If you enable spot instances, Spark jobs that run on the Job resource group will attempt to use spot instance resources. For more information, see Spot instances.
   Spark Configuration	Spark application configuration parameters. These parameters apply to all Spark jobs executed in this resource group. To configure parameters for a specific Spark job, you can set them in the code when you submit the job. For more information about Spark configuration parameters, see Spark application configuration parameters.

   AI resource group

   Parameter	Description
   Deployment Mode	Select RayCluster.
   Head Resource Specifications	The Head node manages Ray metadata, runs the Global Control Store (GCS) service, and participates in task scheduling, but does not execute tasks. The Head resource specification is the number of CPU cores. You can select specifications such as small, m.xlarge, and m.2xlarge. The number of CPU cores for each specification is the same as for Spark resource specifications. For more information, see Spark resource specification list. Important The Head node is mainly responsible for job scheduling. Select a Head specification based on the overall size of the Ray Cluster.
   Worker Group Name	A custom name for the Worker Group. You can configure multiple Worker Groups with different names in one AI resource group.
   Worker Resource Type	Supports CPU and GPU. If your business involves daily computing tasks, multitasking, or complex logical operations, select CPU. If your business involves large-scale data parallel processing, machine learning, or deep learning training, select GPU.
   Worker Resource Specifications	If Worker Resource Type is set to CPU, you can select specifications such as small, m.xlarge, and m.2xlarge for the Worker resource specification. The number of CPU cores for each specification is the same as for Spark resource specifications. For more information, see Spark resource specification list. If Worker Resource Type is set to GPU, due to issues related to GPU models and inventory, submit a ticket to contact technical support for assistance with model selection.
   Worker Disk Storage	The disk space is mainly used to store Ray logs, temporary data, and overflow data from Ray distributed object storage. Unit: GB. Value range: [30, 2000]. Default value: 100 GB. Important The disk is for temporary storage only. Do not use it for long-term data storage.
   Minimum Workers Maximum Workers	Minimum Workers: The minimum number of Workers that must run in a Worker Group. The minimum value is 1. Maximum Workers: The maximum number of Workers to which a Worker Group can be scaled out. The maximum value is 8. Worker Groups support automatic scaling, and each Worker Group can scale independently. If Min Workers and Max Workers are different, AnalyticDB for MySQL dynamically scales the number of Workers between the minimum and maximum values based on the number of tasks. If multiple Worker Groups exist, they are automatically matched to avoid overloading or idling a single Worker Group.
   Allocation Unit	The number of GPUs to allocate on a single Worker node. For example, an allocation unit of 1/3 means that 1/3 of a GPU is configured for each Worker node. Important This parameter is required only when Worker Resource Type is set to GPU.

Modifiable properties

• For custom resource groups (resource groups that you create), you can modify the following properties:

  • For Interactive resource groups: Auto Stop, Cluster Size, Min Clusters, Max Clusters, Job Delivery Rule, and Spark Configuration.

  • For Job resource groups: Max Computing Resources, Spot Instance, and Spark Configuration.

  • For AI resource groups (Ray Cluster deployment type): Head Resource Specification, Worker Resource Type, Worker Resource Specification, Worker Disk Space, Min Workers, and Max Workers.

  Other properties cannot be modified, including Resource Group Name, Task Type, the Engine of Interactive resource groups, the Min Computing Resources of Job resource groups, and the Deployment Type and Worker Group Name of AI resource groups.

• For default resource groups (resource groups named user_default and serverless):

  • For the user_default resource group in Enterprise Edition and Basic Edition clusters, you can modify only the Job Delivery Rule. You cannot modify any properties of the serverless resource group.

  • For the user_default resource group in Data Lakehouse Edition clusters, you can modify Reserved Computing Resources and Job Delivery Rule. You cannot modify any properties of the serverless resource group.

Procedure

1. On the Resource Groups page, find the target resource group and click Modify in the Actions column.

2. In the Modify Resource Group panel that appears, modify the property values. Then, click OK.

   The modification takes effect when the resource group status changes to Running.

Modifiable properties

After a resource group is created, you can modify its query type or resource amount.

• For the default resource group (named user_default), you can modify only the query type. You cannot manually modify the resource amount.

  Note

  The resource amount of the default resource group is calculated by subtracting the resources occupied by other resource groups in the cluster from the total cluster resources.

• For custom resource groups (resource groups that you create), you can modify the query type and resource amount.

Procedure

1. On the Resource Group Management page, find the target resource group and click Modify in the Actions column.

2. Modify the Query Type or Resource Amount as needed.

3. After you finish the modifications, click OK.

   Modifications to the resource amount of an AnalyticDB for MySQL resource group take effect in real time.