7.failure-toleration.yml

---
apiVersion: frisbee.io/v1alpha1
kind: TestPlan
metadata:
  name: failure-toleration
spec:
  actions:
    # Create an iperf server
    - action: Service
      name: server
      service:
        templateRef: iperf2.server
        inputs:
          - { "cpu": "2", memory: "4Gi" }


    # Create a cluster of iperf clients
    - action: Cluster
      name: clients
      depends: { running: [ server ] }
      cluster:
        templateRef: iperf2.client
        instances: 10
        inputs:
          - { server: .service.server.one, seconds: "600" }
        tolerate: # Without this field, the cluster will fail immediately after the killing
          failedServices: 3


    # Periodically kill some nodes. Beware that in this case we use multiple inputs.
    # Notice that we use fixed names, because any selecting macro may choose the same target twice, and stall the execution.
    - action: Cascade
      name: killer
      depends: { running: [ clients ] }
      cascade:
        templateRef: system.chaos.pod.kill
        inputs:
          - { target: clients-4 }
          - { target: clients-7 }
          - { target: clients-3 }
        schedule:
          cron: "@every 1m"