Job Delivery

Procedure

1. In the example a800_tensorflow_vcjob.yaml file, the training job is deployed under the vcjob namespace, so you need to execute the following command on the management node to create a namespace for the job. If the job is created in a non-default namespace, create the namespace according to the actual situation.

   kubectl create namespace vcjob

2. Run the following command in the path of the sample YAML file on the management node to deliver a training job using the YAML file:

   kubectl apply -f XXX.yaml

   

   If the YAML file of a job is modified after the job is successfully delivered, run the kubectl delete -f XXX.yaml command to delete the original job and then deliver the job again.

   • The following is an example of configuring resource information using environment variables:

     kubectl apply -f tensorflow_standalone_acjob.yaml

     Command output:

     1	ascendjob.mindxdl.gitee.com/default-tensorflow-test created

   • The following is an example of configuring resource information using a file:

     kubectl apply -f a800_tensorflow_vcjob.yaml

     Command output:

     1 2	configmap/rings-config-mindx-dls-test created job.batch.volcano.sh/mindx-dls-test created

• If a training job is always in the Pending status after being delivered, refer to Training Job Is in the Pending State Because "nodes are unavailable" or A Job Is Pending Due to Insufficient Resources to rectify the fault.
• If the hccl.json file of a training job container is in the Initializing status after a training job is started, refer to Failed to Generate the hccl.json File to rectify the fault.