Back to All Scenarios
PASSEDcloud / aws_ecs
ECS Task Failing to Start — Image Pull Error
ECS tasks in the production cluster cannot start because the ECR image repository authentication token has expired. All new task launches failing.
Pattern
AWS_CLOUD
Expected: AWS_ECS_FAILURE
Severity
HIGH
Confidence
68%
Remediation
Auto-Heal
Test Results
| Metric | Expected | Actual | Result |
|---|---|---|---|
| Pattern Recognition | AWS_ECS_FAILURE | AWS_CLOUD | |
| Severity Assessment | CRITICAL | HIGH | |
| Incident Correlation | N/A | None | |
| Cascade Escalation | N/A | No | |
| Remediation | — | Auto-Heal — Corax resolves autonomously |
Scenario Conditions
AWS ECS Fargate cluster 'prod'. Service 'api-service' desired count: 4, running: 0. ECR pull error: 'no basic auth credentials'. ECR login token expired.
Injected Error Messages (1)
ECS task start failure — all tasks for service 'api-service' failing to launch, ECR image pull error: 'no basic auth credentials', ECR token expired, running count 0/4 desired
Neural Engine Root Cause Analysis
AWS cloud infrastructure event detected — an EC2 instance may be unreachable or in a stopped state, an RDS database is experiencing issues, a load balancer has unhealthy targets, or a Lambda function is failing. AWS service disruptions can cascade across dependent resources and affect application availability.
Remediation Plan
1. Check the AWS Health Dashboard and Personal Health Dashboard for any active service events.
2. For EC2 issues, check instance status checks (system and instance), review CloudWatch metrics, and check VPC security group rules.
3. For RDS, verify database instance status, check storage and connection limits, and review slow query logs.
4. For ELB issues, check target group health checks and verify backend instances are responding.
5. For Lambda, review CloudWatch Logs for invocation errors and check IAM permissions and VPC connectivity.
Improvements Applied
- Pattern classified as AWS_CLOUD (expected AWS_ECS_FAILURE)
- Severity: HIGH (expected CRITICAL)
Tested: 2026-04-02Monitors: 1 | Incidents: 1Test ID: cmnhnoo60001dlig77n6uacse