🖨️Job

Intro

Fiat job commands overview

Submit command

The submit sub-command is a utility to help users submit long-running computing jobs to an existing Ray computing cluster.

First, you need to write your codes and place relevant files into a working folder. Then define your job description file following the format down below -

{
    "entrypoint": "python script.py",
    "runtime_env": {
        "working_dir": "examples/",
        "pip": [
            "requests",
            "numpy",
            "torch",
            "rich"
        ]
    }
}

Now you can submit your job with Fiat job submit command -

fiat job submit \
    --cluster some-host:8265 \     # Target Fiat Computing cluster URL
    --file /path-to/job_description.json \ # Job descrption file of your project
    --timeout 10                           # submission timeout

Ray Job submission via Fiat CLI job submit command

Status command

A command to check job status -

fiat job status \
    --cluster target-cluster-url \ # host:port format
    --id      raysubmit_xxxxxxxx \ # ray submission id

Check submitted job status

Log command

This command is used to log out stdout when executing the job. It will be really helpful if you want to track the logs of a specific job.

fiat job log \
    --cluster target-cluster-url \ # host:port format
    --id      raysubmit_xxxxxxxx \ # ray submission id

Logging

Terminate command

Sometimes we may make some mistakes or errors and we would like to shutdown a submitted job forcely. The terminate command is the solution for these circumstances.

fiat job terminate \
    --cluster 10.112.67.227:8265 \ # Cluster URL
    --id 07000000                  # Job id, instead of submission_id

Terminate a running job