Application

Source: src/AWS/EMRServerless/Application.ts

An Amazon EMR Serverless application — a serverless Spark or Hive environment that automatically provisions and scales workers per job, with no cluster to manage. An application in the CREATED or STOPPED state costs nothing; billing only occurs for workers while the application is started (including any pre-initialized initialCapacity).

Creating Applications

Spark Application

import * as AWS from "alchemy/AWS";

const app = yield* AWS.EMRServerless.Application("Spark", {
  releaseLabel: "emr-7.9.0",
});
// app.applicationId is passed to StartJobRun

Hive Application with Auto-Stop Tuning

const app = yield* AWS.EMRServerless.Application("Hive", {
  type: "HIVE",
  releaseLabel: "emr-7.9.0",
  autoStartConfiguration: { enabled: true },
  autoStopConfiguration: { enabled: true, idleTimeout: "5 minutes" },
});

Capacity

const app = yield* AWS.EMRServerless.Application("Warm", {
  releaseLabel: "emr-7.9.0",
  initialCapacity: {
    Driver: {
      workerCount: 1,
      workerConfiguration: { cpu: "2 vCPU", memory: "4 GB" },
    },
    Executor: {
      workerCount: 2,
      workerConfiguration: { cpu: "2 vCPU", memory: "4 GB" },
    },
  },
  maximumCapacity: { cpu: "16 vCPU", memory: "64 GB" },
});

Networking

const app = yield* AWS.EMRServerless.Application("InVpc", {
  releaseLabel: "emr-7.9.0",
  networkConfiguration: {
    subnetIds: [subnet.subnetId],
    securityGroupIds: [securityGroup.securityGroupId],
  },
});