When should I use Fargate vs EC2 for AWS Batch?

Use Fargate for most batch workloads — it eliminates the need to manage EC2 instances, scales to zero when idle, and provisions resources per-job. Use EC2 when you need GPU instances, custom AMIs, higher vCPU/memory ratios than Fargate supports, placement groups for tightly coupled HPC workloads, or Amazon Linux instances with specific kernel configurations. EC2 also makes sense when you have sustained workloads where reserved instances reduce costs compared to Fargate's per-second pricing.

How do retry strategies work in Batch job definitions?

The retry strategy specifies the number of times Batch should retry a failed job (1-10 attempts). You can also configure evaluateOnExit rules that examine the exit code, status reason, or error message to decide whether to retry or fail immediately. For example, you might retry on exit code 1 (transient error) but fail immediately on exit code 137 (out of memory) since retrying would produce the same result. Each retry runs a fresh container instance.

AWS Batch Job Definition Builder

ComputeAWS

Build AWS Batch job definitions for Fargate and EC2 with container configs.

Last verified: May 2026

AWS Batch Configuration

Build AWS Batch job definition configs with container properties, retry strategies, and timeouts.

Required Fields

jobDefinitionNametypecontainerProperties.imagecontainerProperties.resourceRequirements

{
  "jobDefinitionName": "data-pipeline-processor",
  "type": "container",
  "platformCapabilities": ["FARGATE"],
  "containerProperties": {
    "image": "123456789012.dkr.ecr.us-east-1.amazonaws.com/data-processor:latest",
    "resourceRequirements": [
      {
        "type": "VCPU",
        "value": "4"
      },
      {
        "type": "MEMORY",
        "value": "8192"
      }
    ],
    "jobRoleArn": "arn:aws:iam::123456789012:role/BatchJobRole",
    "executionRoleArn": "arn:aws:iam::123456789012:role/ecsTaskExecutionRole",
    "command": ["python", "process.py", "--input", "Ref::inputPath"],
    "environment": [
      {
        "name": "ENV",
        "value": "production"
      },
      {
        "name": "LOG_LEVEL",
        "value": "INFO"
      }
    ],
    "secrets": [
      {
        "name": "DB_PASSWORD",
        "valueFrom": "arn:aws:secretsmanager:us-east-1:123456789012:secret:prod/db-password"
      }
    ],
    "logConfiguration": {
      "logDriver": "awslogs",
      "options": {
        "awslogs-group": "/aws/batch/job",
        "awslogs-region": "us-east-1",
        "awslogs-stream-prefix": "data-processor"
      }
    },
    "networkConfiguration": {
      "assignPublicIp": "DISABLED"
    },
    "fargatePlatformConfiguration": {
      "platformVersion": "LATEST"
    }
  },
  "retryStrategy": {
    "attempts": 3,
    "evaluateOnExit": [
      {
        "onStatusReason": "Host EC2*",
        "action": "RETRY"
      },
      {
        "onExitCode": "1",
        "action": "EXIT"
      }
    ]
  },
  "timeout": {
    "attemptDurationSeconds": 3600
  },
  "tags": {
    "Project": "data-pipeline",
    "Environment": "production"
  }
}

Generated Output

Output will appear here...

About This Tool

AWS Batch simplifies running batch computing workloads by automatically provisioning compute resources and scheduling jobs. A job definition specifies the container image, vCPU and memory requirements, environment variables, mount points, retry strategies, and timeout settings. With support for both Fargate and EC2 compute environments, you need to choose the right platform capabilities and configure resource requirements accordingly. The Batch Job Definition Builder generates complete job definition JSON with proper container properties, retry logic, and platform-specific configurations.

Real-World Scenario

Your team runs nightly ETL on a 24/7 EC2 cluster ($800/month). Most jobs run for ~2 hours. The builder generates a Fargate-based Batch job definition: 4 vCPU, 16 GB memory, retry strategy with evaluateOnExit (retry on transient errors, fail on OOM). Combined with a Fargate Spot priority job queue, the same workload runs in ~2 hours/day instead of 24/7 idle. New cost: ~$50/month. Annual savings: $9K, plus the operational simplification of not managing EC2 capacity at all.

When to Use This Tool

•Defining Fargate-based batch jobs with proper vCPU/memory combinations and ephemeral storage settings
•Creating EC2-based job definitions with GPU requirements for machine learning training workloads
•Configuring multi-node parallel job definitions for MPI-based high-performance computing workloads
•Building job definitions with retry strategies and timeout settings that handle transient failures gracefully

Pro Tips

TIP

Always use evaluateOnExit rules instead of blanket retries. Retrying on exit code 137 (OOM kill) is wasteful — the next run will OOM too unless you also bump memory. Retry on exit code 1 (transient errors), fail fast on 137 with a clear error message so the engineer knows to right-size memory.

TIP

Fargate Spot for batch is the killer cost optimization. Most batch jobs are checkpoint-able or idempotent, so Spot interruption is acceptable. Set up a job queue with Fargate Spot priority + Fargate on-demand fallback. You'll save 60-70% on compute with minimal operational impact.

TIP

Don't use Fargate for jobs needing >30 GB memory or specific GPU types — those need EC2. The vCPU/memory pairing rules on Fargate are restrictive: 1 vCPU max 8 GB, 4 vCPU max 30 GB. Beyond that, EC2 launch type with custom AMIs is the right choice.

How It Works Under the Hood

The builder constructs Batch job definition JSON with: type (container, multinode), platform capabilities (FARGATE or EC2), container properties (image, vCPU, memory, GPU resource requirements, environment, mount points, networkConfiguration, executionRoleArn, jobRoleArn), retry strategy (attempts + evaluateOnExit rules), timeout, and parameters. Output is generated as aws batch register-job-definition commands and Terraform aws_batch_job_definition resources.

Frequently Asked Questions

When should I use Fargate vs EC2 for AWS Batch?: Use Fargate for most batch workloads — it eliminates the need to manage EC2 instances, scales to zero when idle, and provisions resources per-job. Use EC2 when you need GPU instances, custom AMIs, higher vCPU/memory ratios than Fargate supports, placement groups for tightly coupled HPC workloads, or Amazon Linux instances with specific kernel configurations. EC2 also makes sense when you have sustained workloads where reserved instances reduce costs compared to Fargate's per-second pricing.
How do retry strategies work in Batch job definitions?: The retry strategy specifies the number of times Batch should retry a failed job (1-10 attempts). You can also configure evaluateOnExit rules that examine the exit code, status reason, or error message to decide whether to retry or fail immediately. For example, you might retry on exit code 1 (transient error) but fail immediately on exit code 137 (out of memory) since retrying would produce the same result. Each retry runs a fresh container instance.

Related Learning Guides

ECS vs EKS Decision Guide24 min read

Was this tool helpful?

Disclaimer: This tool runs entirely in your browser. No data is sent to our servers. Always verify outputs before using them in production. AWS, Azure, and GCP are trademarks of their respective owners.