ChaosExperiment

ChaosExperiment is the core resource. It defines a single chaos injection: what to target, what action to run, how long to run it, and how to verify the system's health before and after.

Group: chaos.chaosplane.io/v1alpha1
Kind: ChaosExperiment
Scope: Namespaced

Example

apiVersion: chaos.chaosplane.io/v1alpha1
kind: ChaosExperiment
metadata:
  name: pod-kill-test
  namespace: production
spec:
  target:
    kind: Pod
    namespace: production
    labelSelector:
      matchLabels:
        app: my-app
  action:
    type: pod-kill
    parameters:
      gracePeriodSeconds: "0"
  duration: 30s
  rollback:
    enabled: false
  execution:
    parallelism: 1
  steadyState:
    before:
      - name: pods-ready
        type: k8s
        k8s:
          resource: pods
          namespace: production
          labelSelector: app=my-app
          condition:
            minReady: 3
    after:
      - name: pods-recovered
        type: k8s
        k8s:
          resource: pods
          namespace: production
          labelSelector: app=my-app
          condition:
            minReady: 3
    recoveryTimeout: 5m
  abortConditions:
    - name: error-rate-spike
      type: prometheus
      prometheus:
        url: http://prometheus:9090
        query: 'rate(http_requests_total{status=~"5.."}[1m])'
        condition:
          operator: ">"
          threshold: 0.1
      action: abort

Spec fields

spec.target

Defines which resources to target.

Field	Type	Required	Description
kind	string	Yes	Pod or Node
namespace	string	Yes (for pods)	Target namespace
labelSelector	LabelSelector	No*	Kubernetes label selector
names	[]string	No*	Explicit resource names

*Either labelSelector or names must be specified.

spec.action

Defines the chaos action to execute.

Field	Type	Required	Description
type	string	Yes	Action type (e.g. pod-kill, network-delay)
parameters	RawExtension	No	Action-specific parameters as key-value pairs
duration	Duration	No	Override for action duration

spec.duration

How long the chaos action runs. Uses Go duration format: 30s, 5m, 1h.

spec.rollback

Field	Type	Required	Default	Description
enabled	bool	Yes	—	Whether to roll back after the experiment
timeout	Duration	No	5m	Max time to wait for rollback to complete

spec.execution

Field	Type	Required	Default	Description
parallelism	int32	No	unlimited	Max number of targets to affect concurrently

spec.steadyState

Field	Type	Required	Description
before	[]ProbeSpec	No	Probes run before chaos injection
after	[]ProbeSpec	No	Probes run after chaos injection
recoveryTimeout	Duration	No	How long to wait for after probes to pass

spec.abortConditions

List of AbortConditionSpec. Each condition is evaluated continuously during the Running phase.

Field	Type	Required	Description
name	string	Yes	Condition name
type	ProbeType	Yes	prometheus, http, or k8s
prometheus	PrometheusProbe	No	Prometheus probe config
http	HTTPProbe	No	HTTP probe config
k8s	K8sProbe	No	Kubernetes probe config
action	AbortAction	Yes	abort, pause, or rollback

ProbeSpec

Field	Type	Required	Description
name	string	Yes	Probe name
type	ProbeType	Yes	prometheus, http, or k8s
prometheus	PrometheusProbe	No	Prometheus probe config
http	HTTPProbe	No	HTTP probe config
k8s	K8sProbe	No	Kubernetes probe config

PrometheusProbe

Field	Type	Required	Description
url	string	Yes	Prometheus base URL
query	string	Yes	PromQL query
condition.operator	string	Yes	<, >, <=, >=, ==, !=
condition.threshold	float64	Yes	Numeric threshold

HTTPProbe

Field	Type	Required	Description
url	string	Yes	URL to request
method	string	No	HTTP method (default: GET)
expectedStatus	int	No	Expected status code (default: 200)
expectedBody	string	No	String that must appear in response body

K8sProbe

Field	Type	Required	Description
resource	string	Yes	Resource type: pods, deployments, nodes
namespace	string	No	Namespace to query
labelSelector	string	No	Label selector string
fieldSelector	string	No	Field selector string
condition.minReady	int	Yes	Minimum number of ready resources

Status fields

Field	Type	Description
phase	ExperimentPhase	Current phase
startTime	Time	When the experiment started
endTime	Time	When the experiment ended
recoveryStartTime	Time	When recovery phase started
conditions	[]Condition	Standard Kubernetes conditions
observedGeneration	int64	Last reconciled generation
affectedResources	[]string	Names of targeted resources
message	string	Human-readable status message

Phases

Phase	Description
Pending	Created, waiting to start
SteadyStateChecking	Running before probes
Running	Chaos is active
Completing	Duration elapsed, wrapping up
Recovering	Running after probes
Completed	All probes passed
Failed	A probe failed or error occurred
Aborted	Manually aborted or abort condition triggered

Printer columns

NAME              PHASE      TARGET   ACTION     AGE
pod-kill-test     Completed  Pod      pod-kill   5m