NOTE: This repository is no longer being used or maintained, and has been archived for historical purposes. You can find the latest changes at https://github.com/chaostoolkit-incubator/aws-az-failure-chaostoolkit
Warning: You are strongly advised to only utilize this extension in environments with non-production workloads, as the actions may cause unwanted downtime to your users. Be sure to check if there are any production workloads running in the target AWS account before running Chaos Toolkit experiments with this extension.
This project is a collection of actions, gathered as an extension to the Chaos Toolkit to simulate an Availability Zone (AZ) failure across multiple AWS services for you to test the resiliency of your hosted applications. This project is purposefully built for simulating AZ failures. If you wish to utilize other fault injection actions with Chaos Toolkit, you might want to consider looking at the Chaos Toolkit Extension for AWS for your experiments.
This package requires Python 3.7 or newer.
To be used from your experiment, this package must be installed in the Python environment where chaostoolkit already lives.
pip install -U aws-az-failure-chaostoolkit
To use the actions from this package, add the blocks of code below to your Chaos Toolkit experiment file. Replace TagKey1
and TagValue1
with the appropriate key-value pair you tagged your resources with. Replace the value of az
argument with an availability zone of your choice.
This action removes subnets belonging to the target AZ in all tagged ASGs and suspends the AZRebalance process if its running. If the ASG is only configured for a single AZ, it updates the min, max and desired capacity to 0:
- type: action
name: Simulate AZ Failure for ASG
provider:
type: python
module: azchaosaws.asg.actions
func: fail_az
arguments:
az: "ap-southeast-1a"
dry_run: True
tags:
- Key: "TagKey1"
Value: "TagValue1"
This action with failure_type
set to network
will affect tagged/filtered subnets in the target AZ by replacing the current NACL associations with a newly created blackhole NACL:
- type: action
name: Simulate AZ Failure for EC2
provider:
type: python
module: azchaosaws.ec2.actions
func: fail_az
arguments:
az: "ap-southeast-1a"
dry_run: True
failure_type: "network"
filters:
- Name: tag:TagKey1
Values:
- "TagValue1"
This action with failure_type
set to instance
will affect tagged/filtered normal/spot instances in the target AZ that are in pending/running state by stopping/terminating them depending on the instance lifecycle:
- type: action
name: Simulate AZ Failure for EC2
provider:
type: python
module: azchaosaws.ec2.actions
func: fail_az
arguments:
az: "ap-southeast-1a"
dry_run: True
failure_type: "instance"
filters:
- Name: tag:TagKey1
Values:
- "TagValue1"
This action removes target AZ subnets in application load balancers:
- type: action
name: Simulate AZ Failure for ALB
provider:
type: python
module: azchaosaws.elbv2.actions
func: fail_az
arguments:
az: "ap-southeast-1a"
dry_run: True
tags:
- Key: "TagKey1"
Value: "TagValue1"
This action detaches classic load balancers from subnets belonging to target AZ if they are in a non-default VPC and disables the target AZ from classic load balancers if they are in a default VPC:
- type: action
name: Simulate AZ Failure for CLB
provider:
type: python
module: azchaosaws.elb.actions
func: fail_az
arguments:
az: "ap-southeast-1a"
dry_run: True
tags:
- Key: "TagKey1"
Value: "TagValue1"
This action forces RDS to reboot and failover to another AZ, and/or promotes one of the Aurora Replicas (read-only instances) in the DB cluster to be the primary instance (cluster writer):
- type: action
name: Simulate AZ Failure for RDS
provider:
type: python
module: azchaosaws.rds.actions
func: fail_az
arguments:
az: "ap-southeast-1a"
dry_run: True
tags:
- Key: "TagKey1"
Value: "TagValue1"
This action forces ElastiCache (cluster mode disabled) to failover primary nodes if exists in the target AZ:
- type: action
name: Simulate AZ Failure for ElastiCache (cluster mode disabled)
provider:
type: python
module: azchaosaws.elasticache.actions
func: fail_az
arguments:
az: "ap-southeast-1a"
dry_run: True
tags:
- Key: "TagKey1"
Value: "TagValue1"
This action forces ElastiCache (cluster mode enabled) to failover the shards provided as cache cluster ids (sequential if multiple shards of same cluster) (replace ReplicationGroup1, CacheClusterId1 and CacheClusterId2 as required):
- type: action
name: Simulate AZ Failure for ElastiCache (cluster mode enabled)
provider:
type: python
module: azchaosaws.elasticache.actions
func: fail_az
arguments:
az: "ap-southeast-1a"
dry_run: True
tags:
- Key: "TagKey1"
Value: "TagValue1"
replication_groups:
- replication_group_id: ReplicationGroup1
cache_cluster_ids:
- CacheClusterId1
- CacheClusterId2
This action removes subnets belonging to the target AZ in all nodegroup ASGs that are part of the tagged EKS clusters and suspends the AZRebalance process if its running. failure_type
set to network
will affect target AZ subnets of the nodegroups by associating them with a newly created blackhole NACL. All its previous NACL associations will be replaced with the blackhole NACL:
- type: action
name: Simulate AZ Failure for EKS Clusters
provider:
type: python
module: azchaosaws.eks.actions
func: fail_az
arguments:
az: "ap-southeast-1a"
dry_run: True
failure_type: "network"
tags:
- TagKey1: "TagValue1"
This action removes subnets belonging to the target AZ in all nodegroup ASGs that are part of the tagged EKS clusters and suspends the AZRebalance process if its running.failure_type
set to instance
will affect worker nodes that are part of the managed node groups and are in a pending/running state in the target AZ by stopping/terminating normal/spot instances:
- type: action
name: Simulate AZ Failure for EKS Clusters
provider:
type: python
module: azchaosaws.eks.actions
func: fail_az
arguments:
az: "ap-southeast-1a"
dry_run: True
failure_type: "instance"
tags:
- TagKey1: "TagValue1"
This action reboots ActiveMQ brokers that have an active-standby setup:
- type: action
name: Simulate AZ Failure for Amazon MQ (ActiveMQ)
provider:
type: python
module: azchaosaws.mq.actions
func: fail_az
arguments:
az: "ap-southeast-1a"
dry_run: True
tags:
- TagKey1: "TagValue1"
- To 'rollback' the changes made by the
fail_az
action, you can userecover_az
in your experiment template. Therecover_az
action will read the state file generated and rollback if it's a service that's supported. - Do also note that by default, the
dry_run
argument for eachfail_az
action is required. Setting it toTrue
will only run read-only operations and not impact the target resources. Set it toFalse
if you want the actions to make changes your resources. It is best practice to set it on an experiment level under the configuration block and then reference it for every action. - To have granular filtering of resources, you can also provide a list of tags as part of the argument for the
fail_az
action.
Please explore the code to see existing actions and supported arguments. Alternatively, you can run chaos discover aws-az-failure-chaostoolkit
to view the list of supported actions along with their required and optional arguments for each service in the generated discovery.json
file.
If you wish to develop on this project, make sure to install the development dependencies. But first, create a virtual environment and then install those dependencies.
make install-dev
Now, you can edit the files and they will be automatically be seen by your
environment, even when running from the chaos
command locally.
To format your code execute the following:
make fmt
To check your code with a linter execute the following:
make lint
To run the tests for the project execute the following:
make test
See CONTRIBUTING for more information.
This project is licensed under the Apache-2.0 License.