Robomaker Diagnostics

AWS RoboMaker Diagnostics

When to use

Any AWS RoboMaker investigation — simulation jobs, robot applications, fleet management, ROS integration, world generation, or deployment.

Investigation workflow

Step 1 — Collect and triage

aws robomaker list-simulation-jobs --query 'simulationJobSummaries[*].{Arn:arn,Status:status,Name:name}'
version: "1.0.0"
last_updated: "2025-04-12"
aws robomaker list-robot-applications
aws robomaker list-robots --query 'robots[*].{Arn:arn,Name:name,Status:status}'
version: "1.0.0"
last_updated: "2025-04-12"
aws robomaker list-fleets

Step 2 — Domain deep dive

aws robomaker describe-simulation-job --job <job-arn>
aws robomaker describe-robot-application --application <app-arn>
aws robomaker describe-fleet --fleet <fleet-arn>
aws robomaker list-world-generation-jobs

Step 3 — Detailed investigation

aws robomaker describe-world-generation-job --job <job-arn>
aws logs filter-log-events --log-group-name /aws/robomaker/SimulationJobs --start-time <epoch-ms>
aws cloudtrail lookup-events --lookup-attributes AttributeKey=EventSource,AttributeValue=robomaker.amazonaws.com --max-results 20

Read references/guardrails.md before concluding on any RoboMaker issue.

Tool quick reference

Gotchas: AWS RoboMaker

• Simulation jobs run in isolated VPC environments. Network access must be explicitly configured.
• Robot applications must be built for the target architecture (X86_64 or ARM64).
• ROS version must match between simulation and robot applications (ROS1 Melodic/ROS2 Foxy).
• World generation uses WorldForge templates. Complex worlds take longer to generate.
• Fleet deployment uses AWS IoT Greengrass. Robots must have Greengrass Core installed.
• Simulation job duration has limits. Jobs timeout after the configured max duration.

Anti-hallucination rules

1. Always cite simulation job ARNs, robot ARNs, or API responses as evidence.
2. RoboMaker simulation runs in AWS cloud, not on physical robots.
3. ROS version compatibility is strict. Never mix ROS1 and ROS2.
4. Fleet deployment requires Greengrass. Never suggest direct deployment.
5. Spend no more than 2 minutes on any single hypothesis. Pivot if inconclusive.

14 runbooks