Deadline Cloud Diagnostics

AWS Deadline Cloud Diagnostics

When to use

Any AWS Deadline Cloud investigation — farm creation, fleet configuration, queue management, job submission, worker agents, worker hosts, storage profiles, job attachments, budgets, license endpoints, monitors, identity management, or custom job templates.

Investigation workflow

Step 1 — Collect and triage

aws deadline list-farms --query 'farms[*].{FarmId:farmId,DisplayName:displayName,KmsKeyArn:kmsKeyArn}'
aws deadline list-fleets --farm-id <farm-id> --query 'fleets[*].{FleetId:fleetId,DisplayName:displayName,Status:status,MaxWorkerCount:maxWorkerCount}'
aws deadline list-queues --farm-id <farm-id> --query 'queues[*].{QueueId:queueId,DisplayName:displayName,Status:status}'

Step 2 — Domain deep dive

aws deadline get-farm --farm-id <farm-id>
aws deadline get-fleet --farm-id <farm-id> --fleet-id <fleet-id>
aws deadline get-queue --farm-id <farm-id> --queue-id <queue-id>

Step 3 — Detailed investigation

aws deadline list-jobs --farm-id <farm-id> --queue-id <queue-id> --query 'jobs[*].{JobId:jobId,Name:name,LifecycleStatus:lifecycleStatus,TaskRunStatus:taskRunStatus}'
aws deadline list-workers --farm-id <farm-id> --fleet-id <fleet-id> --query 'workers[*].{WorkerId:workerId,Status:status,HostProperties:hostProperties}'
aws deadline list-sessions --farm-id <farm-id> --queue-id <queue-id> --job-id <job-id>

Read references/guardrails.md before concluding on any Deadline Cloud issue.

Tool quick reference

Gotchas: AWS Deadline Cloud

• Deadline Cloud has a hierarchy: Farm → Fleet/Queue → Job → Task → Session. Farms are the top-level container. Fleets provide compute. Queues organize jobs.
• Fleets can be customer-managed (your EC2 instances) or service-managed (AWS manages scaling). Service-managed fleets auto-scale based on job demand.
• Worker agents run on fleet instances and communicate with the Deadline Cloud service. Agent failures prevent job execution even if instances are healthy.
• Job attachments use S3 for file transfer between submitter and workers. Storage profiles define file system mappings. Misconfigured profiles cause job failures.
• Budgets control spending but do NOT stop running jobs. When a budget is exhausted, new jobs are blocked but existing jobs continue to completion.
• License endpoints connect to license servers for commercial DCC applications. Network connectivity between workers and license servers is critical.
• Custom job templates define job structure using OpenJD specification. Template errors cause job submission failures, not runtime failures.

Anti-hallucination rules

1. Always cite specific farm IDs, fleet IDs, queue IDs, or API responses as evidence.
2. Service-managed and customer-managed fleets have different troubleshooting paths. Never conflate them.
3. Worker agent issues and worker host issues are different. Agent is software; host is infrastructure.
4. Job attachments use S3, not EFS or FSx by default. Never assume shared file system without checking storage profiles.
5. Spend no more than 2 minutes on any single hypothesis. Pivot if inconclusive.

14 runbooks