Job configuration description

更新时间: 2026-07-22 16:42:51

This topic describes the common configuration items and configuration methods for RAY jobs on the Lindorm compute engine.

Configuration overview

Ray jobs support two categories of configurations: job-level and cluster-level:

  • jobConfig: Job-level configurations, such as the target resource group, RuntimeEnv, and TTL.

  • clusterConfig: Cluster-level configurations, such as the image, Head/Worker resource specifications, instance type preferences, and environment variables.

jobConfig items

Configuration item

Required

Default value

Description

jobConfig.computeGroup

Yes

None

Name of the target RAY resource group. Compatible with the unified compute-group configuration of the compute engine.

jobConfig.runtimeEnvJson

No

None

A Ray RuntimeEnv JSON string that specifies dependency packages, the working directory, and environment variables.

jobConfig.ttlSecondsAfterFinished

No

300

The retention time of the Ray cluster after the job completes, in seconds.

clusterConfig items

Cluster-level configurations

Configuration item

Required

Default value

Description

clusterConfig.image

No

Platform default image

The image address used to run the Ray job. If not specified, the default image is used. For information about how to use a custom image, see Run Ray jobs with a custom image.

clusterConfig.env.<key>

No

None

Cluster-level environment variables.

Head node configuration

Configuration item

Required

Default value

Description

clusterConfig.head.cpu

No

4

Number of CPU cores for the Head node.

clusterConfig.head.memoryGB

No

16

Memory of the Head node (GB).

clusterConfig.head.diskSizeGB

No

30

Disk of the Head node (GB).

clusterConfig.head.machineType

No

None

Instance type preference for the Head node.

clusterConfig.head.env.<key>

No

None

Head node environment variables.

Worker Group configuration

Worker Groups are distinguished by name. For example, to define a Worker Group named wg1, use clusterConfig.workerGroup.wg1.*.

Configuration item

Required

Default value

Description

clusterConfig.workerGroup.wg1.cpu

No

4

Number of CPU cores for the Worker.

clusterConfig.workerGroup.wg1.memoryGB

No

16

Worker memory (GB).

clusterConfig.workerGroup.wg1.diskSizeGB

No

30

Worker disk (GB).

clusterConfig.workerGroup.wg1.replicas

No

2

Number of Worker replicas.

clusterConfig.workerGroup.wg1.minReplicas

No

None

Minimum number of Worker replicas. Takes effect when auto scaling is enabled.

clusterConfig.workerGroup.wg1.maxReplicas

No

None

Maximum number of Worker replicas. Takes effect when auto scaling is enabled.

clusterConfig.workerGroup.wg1.machineType

No

None

Instance type preference for the Worker.

clusterConfig.workerGroup.wg1.env.<key>

No

None

Worker node environment variables.

Configuration method

Submit a Ray job by using POST /api/v1/lindorm/jobs/{token}. All configurations are passed in through the conf field of the request body.

{
  "owner": "data-team",
  "mainResourceKind": "ray",
  "mainResource": "python my_script.py --arg1 val1",
  "name": "my-ray-job",
  "conf": {
    "jobConfig.computeGroup": "raycgserverless",
    "jobConfig.runtimeEnvJson": "{\"pip\":[\"pandas\"],\"working_dir\":\"https://your-bucket.oss-cn-hangzhou.aliyuncs.com/ray-project.zip\"}",
    "clusterConfig.head.cpu": "4",
    "clusterConfig.head.memoryGB": "16",
    "clusterConfig.workerGroup.wg1.cpu": "8",
    "clusterConfig.workerGroup.wg1.memoryGB": "32",
    "clusterConfig.workerGroup.wg1.minReplicas": "1",
    "clusterConfig.workerGroup.wg1.maxReplicas": "4"
  }
}
上一篇: Submit jobs to a non-resident RAY resource group 下一篇: Configure job runtime environments
阿里云首页 云原生多模数据库 Lindorm 相关技术圈