#Integrate with JobSet

This page shows how to use Alauda Build of Kueue to manage JobSet workloads. A JobSet allows you to define a group of related Kubernetes Jobs that are managed together as a single unit.

JobSet is useful for distributed workloads that consist of multiple job components, such as a driver job and multiple worker jobs, where all components must run together.

#Prerequisites

• You have installed the Alauda Build of Kueue.
• You have installed the JobSet controller.
• The JobSet framework is enabled in the Kueue configuration.
• The Alauda Container Platform Web CLI has communication with your cluster.
• You have created a ClusterQueue, ResourceFlavor, and LocalQueue.

#Procedure

1. Create a JobSet resource with the kueue.x-k8s.io/queue-name label:

   apiVersion: jobset.x-k8s.io/v1alpha2
   kind: JobSet
   metadata:
     name: training-jobset
     namespace: team-ml
     labels:
       kueue.x-k8s.io/queue-name: team-ml-queue
   spec:
     replicatedJobs:
     - name: workers
       replicas: 1
       template:
         spec:
           parallelism: 4
           completions: 4
           backoffLimit: 1
           template:
             spec:
               containers:
               - name: worker
                 image: pytorch/pytorch:2.1.0-cuda12.1-cudnn8-runtime
                 resources:
                   requests:
                     cpu: "2"
                     memory: "4Gi"
                     nvidia.com/gpu: "1"
                 command: ["python", "train.py"]
     - name: driver
       template:
         spec:
           parallelism: 1
           completions: 1
           backoffLimit: 0
           template:
             spec:
               containers:
               - name: driver
                 image: pytorch/pytorch:2.1.0-cuda12.1-cudnn8-runtime
                 resources:
                   requests:
                     cpu: "2"
                     memory: "4Gi"
                 command: ["python", "driver.py"]
   1. kueue.x-k8s.io/queue-name: Specifies the LocalQueue that manages this JobSet. Kueue admits all replicated jobs together as a single unit.
   2. workers: A replicated job that runs 4 parallel worker pods, each requesting 1 GPU.
   3. driver: A single driver pod that coordinates the workers.

2. Apply the JobSet:

   kubectl apply -f training-jobset.yaml

3. Monitor the JobSet admission:

   kubectl get workloads -n team-ml

4. Check the JobSet status:

   kubectl get jobset training-jobset -n team-ml