ood-emr-adapter

README

ood-emr-adapter

OOD compute adapter for Amazon EMR Serverless. Translates Open OnDemand job lifecycle calls (submit / status / delete / info) to the EMR Serverless API.

Commands

Command	Description
submit	Read a JSON job spec from stdin and submit it as an EMR Serverless job run. Prints <application-id>/<job-run-id> on success.
status <id>	Return OOD-normalised job status as JSON (queued, running, completed, failed, cancelled, undetermined).
delete <id>	Cancel a running EMR Serverless job run.
info <id>	Print the full GetJobRun API response as JSON.

<id> must be in the form <application-id>/<job-run-id>, or just <job-run-id> when --application-id is set.

Global flags

Flag	Default	Description
--region	us-east-1	AWS region
--application-id	(empty)	EMR Serverless application ID (can also be set per-job in the spec)

Job spec (stdin for submit)

{
  "application_id": "00abc123def456gh",
  "execution_role_arn": "arn:aws:iam::123456789012:role/EMRServerlessRole",
  "entry_point": "s3://my-bucket/scripts/wordcount.py",
  "entry_point_args": ["s3://my-bucket/input/", "s3://my-bucket/output/"],
  "spark_submit_parameters": "--conf spark.executor.cores=4 --conf spark.executor.memory=8g",
  "job_name": "wordcount-run-42",
  "env": {
    "MY_ENV_VAR": "value"
  }
}

Open OnDemand cluster YAML example

# config/clusters.d/emr-serverless.yml
v2:
  metadata:
    title: "Amazon EMR Serverless"
  login:
    host: "emr-serverless.internal"
  job:
    adapter: "script"
    submit: "/usr/local/bin/ood-emr-adapter submit --region us-east-1 --application-id 00abc123def456gh"
    status: "/usr/local/bin/ood-emr-adapter status --region us-east-1"
    delete: "/usr/local/bin/ood-emr-adapter delete --region us-east-1"
    info:   "/usr/local/bin/ood-emr-adapter info   --region us-east-1"

Prerequisites

• Go 1.26+
• AWS credentials configured (environment variables, ~/.aws/credentials, or IAM role)
• An existing EMR Serverless application

Build

go build -o ood-emr-adapter .

License

MIT — see LICENSE.