Runs in Kubernetes, kills random Kubernetes pods / nodes.

• Dry Run mode
• Randomise mode
• Num Pods to Delete
• make config filename configurable
• make grace period configurable
• Excluding namespaces for pod deletion
• Namespace inclusion option
• Dockerfile
• Missing tests for nodes
• CI in gitlab
• Reenable node deletion (setup test cluster to autoheal first)
  • May need some threading?
• Num Nodes to Delete
• assertions for test_find_no_pods & test_find_no_nodes

The agent will run against the current Kubernetes context. In other words, it'll start finding and deleting pods/nodes against your locally authenticated cluster.

It is also designed to run within the cluster itself - acting on other resources accordingly. See ./go build and ./go deploy for pointers on how this works.

/// add note about RBAC here

You can specify a config file to set a number of options to change the agent's behaviour:

• Defaults are set in class Config in chaos_agent/utils.py
• When using ./go run, the local-config.yaml will be loaded ...
• ... or you can run the python directly through pipenv and override it with export CFG_FILE=/path/to/yaml.
• In k8s, a ConfigMap is used

I like to use wrappers through a ./go bash script to save my typing, see:

1. ./go init to create your pip venv
2. ./go test to run pytest locally
3. ./go watch-tests to run pytest continuously for faster feedback
4. You can also ./go run to run the tool locally - this will connect to your current kubectl context and try to do stuff!
   • by default the config uses dry-run mode (prints what it would do, rather than delete stuff), but if you've changed it ... ;)