AWS Fault Injection Service
AWS Fault Injection Service (FIS) runs experiments that disrupt AWS workloads so teams can examine application performance, observability, and resilience.
What AWS Fault Injection Service is
AWS Fault Injection Service (FIS) runs experiments that disrupt AWS workloads so teams can examine application performance, observability, and resilience. Its Scenario Library offers AWS-created templates covering disruptions from partial failures to more severe impacts. Many experiment actions do not require agents, but instance-level CPU and memory utilization faults require the SSM agent. Teams can choose targets with tags and set CloudWatch alarms or other conditions to stop or automatically roll back an experiment. The console and APIs show progress and action details, while CloudWatch metrics support monitoring. FIS is accessible through the AWS Management Console, CLI, SDKs, CloudFormation, and HTTPS API, including for CI/CD pipelines and custom tools. IAM controls who can run experiments and which resources may be affected; FIS also generates CloudTrail logs for actions it takes. Experiments affect real AWS resources, so AWS recommends planning them and running them in pre-production before production use. Pricing starts at $0.10 per action-minute; customer-provided CloudWatch Dashboard experiment reports cost $5 per report.
Good
- AWS-created disruption scenario templates
- Can stop or roll back on specified conditions
- CloudWatch metrics and experiment progress details
- IAM controls experiment access and affected resources
Watch out
- No free plan
- Most instance-level CPU and memory faults require SSM agent
- Experiments affect real AWS resources
- Reports cost $5 per report
Who it suits
FIS suits teams that want to examine how AWS workloads respond to controlled disruptions. It is especially relevant to teams able to plan experiments carefully and use pre-production before production.
Verdict
FIS offers configurable experiments, monitoring, and safeguards for AWS workloads. Its per-action-minute charges and real-resource effects call for careful planning.
Plans and prices
From AWS Fault Injection Service's own pricing page, checked 3 Oct 2001.
| Plan | Price | What it gives you |
|---|---|---|
| AWS Fault Injection Service | $0.10, paid once | No upfront costs or minimum fees · charges depend on active action duration and target accounts · action duration rounded to nearest minute · Per action-minute; plus $0.10 per action-minute for each additional account. AWS GovCloud (US-East and US-West): $0.12 per action-minute, plus $0.12 per action-minute for each additional account. Experiment reports: $5 per report. |
The facts
Each one read from AWS Fault Injection Service's own pages, with the date it was checked.
What it does
- Access security
- FIS integrates with AWS Identity and Access Management to control who can run experiments and which resources and services can be affected.checked 3 Oct 2001
- Agent exception
- Instance-level CPU and memory utilization faults require the SSM agent.checked 3 Oct 2001
- Audit trail
- AWS states that FIS generates CloudTrail logs for actions it takes on AWS resources during an experiment.checked 3 Oct 2001
- Integrations
- FIS can be accessed through the AWS Management Console, CLI, SDKs, CloudFormation, and HTTPS API, and its APIs can be used in CI/CD pipelines and custom tooling.checked 3 Oct 2001
- Managed service
- FIS is fully managed and its experiments can run without installing agents for many action types.checked 3 Oct 2001
- Monitoring
- FIS supports Amazon CloudWatch metrics for monitoring experiments and shows experiment progress and action details through its console and APIs.checked 3 Oct 2001
- Operational scope
- For a single-account experiment, target resources must be in the experiment's AWS account; multi-account experiments can target resources in other accounts.checked 3 Oct 2001
- Purpose
- AWS FIS runs fault injection experiments on AWS workloads to help teams improve application performance, observability, and resilience.checked 3 Oct 2001
- Real resource effects
- FIS carries out real actions on real AWS resources, and AWS recommends planning experiments and running them in pre-production before production use.checked 3 Oct 2001
- Report charge
- Experiment reports summarizing actions and application response from a customer-provided CloudWatch Dashboard cost $5 per report.checked 3 Oct 2001
- Rollback
- FIS supports automatically rolling back or stopping an experiment when specified conditions are met.checked 3 Oct 2001
- Safety controls
- Experiments can use tags to select targets and CloudWatch alarms or other tools to stop an experiment when conditions are met.checked 3 Oct 2001
- Scenario library
- The FIS Scenario Library provides AWS-created experiment templates for testing disruptions ranging from partial failures to more severe impacts.checked 3 Oct 2001
- Transport security
- FIS API clients must support TLS 1.2, and AWS recommends TLS 1.3.checked 3 Oct 2001
Alternatives to AWS Fault Injection Service
- 1
FaultMesh
Free planWebAPISelf-hosted7.2From $49 a month - 2
Gremlin
WebAPILinuxSelf-hosted7.2From $3,750 a month - 3
Reliably
Free planWebLinuxmacOSSelf-hosted7.1From £40 a month - 4
Krkn
Free planWebAPILinuxSelf-hosted6.7Free plan - 5
Chaos Mesh
Free planWebAPISelf-hosted6.5Free plan - 6
Toxiproxy
Free planAPILinuxmacOSSelf-hosted6.5Free plan