What is AWS Fault Injection Service?
AWS Fault Injection Service (FIS) is a managed chaos engineering service that enables controlled fault injections into AWS resources. You define experiments that simulate real failure scenarios and observe how your application responds.
FIS helps identify resilience weaknesses before they occur in production. Built-in safety mechanisms like stop conditions and IAM-based blast radius control ensure safe experiments.
Core Features
- Experiment Templates: Predefined scenarios for common failure types
- Multi-Service Targets: EC2, ECS, EKS, RDS, and network faults
- Stop Conditions: Automatic termination on CloudWatch alarm triggers
- Blast Radius Control: IAM-based limitation to defined resources
- Experiment Logging: Detailed logging of all actions and results
Typical Use Cases
AZ Failure Simulation: Simulating the failure of an Availability Zone to validate that applications automatically failover to other AZs. Critical for highly available architectures.
Latency Injection: Injecting network latency between services to validate timeout configurations and circuit breakers. Ensures microservices degrade gracefully.
Gameday Exercises: Regular team resilience tests where failure scenarios are simulated and response times measured. Improves both architecture and operational processes.
Benefits
- Controlled fault injections with safety mechanisms
- Identification of resilience weaknesses before production incidents
- Predefined templates for quick start
- Native integration with AWS services
Integration with innFactory
As an AWS Reseller, innFactory supports you with AWS Fault Injection Service: chaos engineering strategy, experiment design, resilience audits, and gameday planning.
Typical Use Cases
Frequently Asked Questions
What can I test with Fault Injection Service?
You can stop EC2 instances, trigger CPU/memory stress, inject network latency, simulate AZ failures, stop ECS tasks, and trigger RDS failovers. All actions have configurable stop conditions.
How do I prevent uncontrolled impact?
Stop conditions based on CloudWatch alarms automatically terminate experiments when predefined thresholds are exceeded. IAM roles limit the blast radius to defined resources.
Do I need chaos engineering experience?
FIS provides predefined experiment templates for common scenarios. You can start with simple experiments like EC2 instance stops and gradually increase complexity.