Skip to content

GCP VM Instance Stop

Introduction

  • It causes power-off of a GCP VM instance by instance name or list of instance names before bringing it back to the running state after the specified chaos duration.
  • It helps to check the performance of the application/process running on the VM instance.
  • When the MANAGED_INSTANCE_GROUP is enable then the experiment will not try to start the instances post chaos, instead it will check the addition of new instances to the instance group.

Scenario: stop the gcp vm

GCP VM Instance Stop

Uses

View the uses of the experiment

coming soon

Prerequisites

Verify the prerequisites
  • Ensure that Kubernetes Version > 1.16
  • Ensure that the Litmus Chaos Operator is running by executing kubectl get pods in operator namespace (typically, litmus).If not, install from here
  • Ensure that the gcp-vm-instance-stop experiment resource is available in the cluster by executing kubectl get chaosexperiments in the desired namespace. If not, install from here
  • Ensure that you have sufficient GCP permissions to stop and start the GCP VM instances.
  • Ensure to create a Kubernetes secret having the GCP service account credentials in the default namespace. A sample secret file looks like:

    apiVersion: v1
    kind: Secret
    metadata:
      name: cloud-secret
    type: Opaque
    stringData:
      type: 
      project_id: 
      private_key_id: 
      private_key: 
      client_email: 
      client_id: 
      auth_uri: 
      token_uri: 
      auth_provider_x509_cert_url: 
      client_x509_cert_url: 
    

Default Validations

View the default validations
  • VM instance should be in healthy state.

Minimal RBAC configuration example (optional)

NOTE

If you are using this experiment as part of a litmus workflow scheduled constructed & executed from chaos-center, then you may be making use of the litmus-admin RBAC, which is pre installed in the cluster as part of the agent setup.

View the Minimal RBAC permissions

---
apiVersion: v1
kind: ServiceAccount
metadata:
  name: gcp-vm-instance-stop-sa
  namespace: default
  labels:
    name: gcp-vm-instance-stop-sa
    app.kubernetes.io/part-of: litmus
---
apiVersion: rbac.authorization.k8s.io/v1
kind: ClusterRole
metadata:
  name: gcp-vm-instance-stop-sa
  labels:
    name: gcp-vm-instance-stop-sa
    app.kubernetes.io/part-of: litmus
rules:
  # Create and monitor the experiment & helper pods
  - apiGroups: [""]
    resources: ["pods"]
    verbs: ["create","delete","get","list","patch","update", "deletecollection"]
  # Performs CRUD operations on the events inside chaosengine and chaosresult
  - apiGroups: [""]
    resources: ["events"]
    verbs: ["create","get","list","patch","update"]
  # Fetch configmaps & secrets details and mount it to the experiment pod (if specified)
  - apiGroups: [""]
    resources: ["secrets","configmaps"]
    verbs: ["get","list",]
  # Track and get the runner, experiment, and helper pods log 
  - apiGroups: [""]
    resources: ["pods/log"]
    verbs: ["get","list","watch"]  
  # for creating and managing to execute comands inside target container
  - apiGroups: [""]
    resources: ["pods/exec"]
    verbs: ["get","list","create"]
  # for configuring and monitor the experiment job by the chaos-runner pod
  - apiGroups: ["batch"]
    resources: ["jobs"]
    verbs: ["create","list","get","delete","deletecollection"]
  # for creation, status polling and deletion of litmus chaos resources used within a chaos workflow
  - apiGroups: ["litmuschaos.io"]
    resources: ["chaosengines","chaosexperiments","chaosresults"]
    verbs: ["create","list","get","patch","update","delete"]
  # for experiment to perform node status checks
  - apiGroups: [""]
    resources: ["nodes"]
    verbs: ["get","list"]
---
apiVersion: rbac.authorization.k8s.io/v1
kind: ClusterRoleBinding
metadata:
  name: gcp-vm-instance-stop-sa
  labels:
    name: gcp-vm-instance-stop-sa
    app.kubernetes.io/part-of: litmus
roleRef:
  apiGroup: rbac.authorization.k8s.io
  kind: ClusterRole
  name: gcp-vm-instance-stop-sa
subjects:
- kind: ServiceAccount
  name: gcp-vm-instance-stop-sa
  namespace: default
Use this sample RBAC manifest to create a chaosServiceAccount in the desired (app) namespace. This example consists of the minimum necessary role permissions to execute the experiment.

Experiment tunables

check the experiment tunables

Mandatory Fields

Variables Description Notes
GCP_PROJECT_ID GCP project ID to which the VM instances belong All the VM instances must belong to a single GCP project
VM_INSTANCE_NAMES Name of target VM instances Multiple instance names can be provided as instance1,instance2,...
ZONES The zones of the target VM instances Zone for every instance name has to be provided as zone1,zone2,... in the same order of VM_INSTANCE_NAMES

Optional Fields



Variables Description Notes
TOTAL_CHAOS_DURATION The total time duration for chaos insertion (sec) Defaults to 30s
CHAOS_INTERVAL The interval (in sec) between successive instance termination Defaults to 30s
MANAGED_INSTANCE_GROUP Set to enable if the target instance is the part of a managed instance group Defaults to disable
SEQUENCE It defines sequence of chaos execution for multiple instance Default value: parallel. Supported: serial, parallel
RAMP_TIME Period to wait before and after injection of chaos in sec

Experiment Examples

Common Experiment Tunables

Refer the common attributes to tune the common tunables for all the experiments.

Target GCP Instances

It will stop all the instances with the given VM_INSTANCE_NAMES instance names and corresponding ZONES zone names in GCP_PROJECT_ID project.

NOTE: The VM_INSTANCE_NAMES contains multiple comma-separated vm instances. The comma-separated zone names should be provided in the same order as instance names.

Use the following example to tune this:

## details of the gcp instance
apiVersion: litmuschaos.io/v1alpha1
kind: ChaosEngine
metadata:
  name: engine-nginx
spec:
  engineState: "active"
  annotationCheck: "false"
  chaosServiceAccount: gcp-vm-instance-stop-sa
  experiments:
  - name: gcp-vm-instance-stop
    spec:
      components:
        env:
        # comma separated list of vm instance names
        - name: VM_INSTANCE_NAMES
          value: 'instance-01,instance-02'
        # comma separated list of zone names corresponds to the VM_INSTANCE_NAMES
        # it should be provided in same order of VM_INSTANCE_NAMES
        - name: ZONES
          value: 'zone-01,zone-02'
        # gcp project id to which vm instance belongs
        - name: GCP_PROJECT_ID
          value: 'project-id'
        - name: TOTAL_CHAOS_DURATION
          value: '60'

Managed Instance Group

If vm instances belong to a managed instance group then provide the MANAGED_INSTANCE_GROUP as enable else provided it as disable, which is the default value.

Use the following example to tune this:

## scale up and down to maintain the available instance counts
apiVersion: litmuschaos.io/v1alpha1
kind: ChaosEngine
metadata:
  name: engine-nginx
spec:
  engineState: "active"
  annotationCheck: "false"
  chaosServiceAccount: gcp-vm-instance-stop-sa
  experiments:
  - name: gcp-vm-instance-stop
    spec:
      components:
        env:
        # tells if instances are part of managed instance group
        # supports: enable, disable. default: disable
        - name: MANAGED_INSTANCE_GROUP
          value: 'enable'
        # comma separated list of vm instance names
        - name: VM_INSTANCE_NAMES
          value: 'instance-01,instance-02'
        # comma separated list of zone names corresponds to the VM_INSTANCE_NAMES
        # it should be provided in same order of VM_INSTANCE_NAMES
        - name: ZONES
          value: 'zone-01,zone-02'
        # gcp project id to which vm instance belongs
        - name: GCP_PROJECT_ID
          value: 'project-id'
        - name: TOTAL_CHAOS_DURATION
          value: '60'

Mutiple Iterations Of Chaos

The multiple iterations of chaos can be tuned via setting CHAOS_INTERVAL ENV. Which defines the delay between each iteration of chaos.

Use the following example to tune this:

# defines delay between each successive iteration of the chaos
apiVersion: litmuschaos.io/v1alpha1
kind: ChaosEngine
metadata:
  name: engine-nginx
spec:
  engineState: "active"
  annotationCheck: "false"
  chaosServiceAccount: gcp-vm-instance-stop-sa
  experiments:
  - name: gcp-vm-instance-stop
    spec:
      components:
        env:
        # delay between each iteration of chaos
        - name: CHAOS_INTERVAL
          value: '15'
        # time duration for the chaos execution
        - name: TOTAL_CHAOS_DURATION
          value: '60'
        - name: VM_INSTANCE_NAMES
          value: 'instance-01,instance-02'
        - name: ZONES
          value: 'zone-01,zone-02'
        - name: GCP_PROJECT_ID
          value: 'project-id'