Kubernetes locations installation

Run a Control Plane on Kubernetes, to set up your Private Locations and run load generators in your own Kubernetes network.

Introduction#

A control plane, installed on a kubernetes cluster with the needed roles will be able to deploy kubernetes private locations.

To better understand what roles are needed, here’s a quick overview of the mechanisms involved when deploying a private location on Kubernetes.

First, the control plane regularly polls the Gatling Enterprise API, and it is notified to deploy a number of instances in a given Kubernetes location.

Then, it creates a config map for each namespace of each Kubernetes location requested for this run. The config map includes the load generator’s start script, so these scripts don’t need to be embedded in the Docker images used for the load generators.

Finally, the control plane initiates a batch job, with the number of instances configured for the simulation run, and mounts the appropriate config map on each pod.

Roles#

For each namespace configured for a private location, the control plane needs the following roles:

apiVersion: rbac.authorization.k8s.io/v1
kind: Role
metadata:
  name: control-plane-role
  namespace: <namespace>
rules:
  - apiGroups: [""]
    resources: ["configmaps"]
    verbs: ["create", "get", "patch", "update", "delete"]
  - apiGroups: ["batch"]
    resources: ["jobs"]
    verbs: ["create", "deletecollection"]

Example#

The control plane requires a configuration under /app/conf/control-plane.conf, see control plane configuration.

kubectl create configmap control-plane-config \
        --from-file=control-plane.conf

apiVersion: v1
kind: ServiceAccount
metadata:
  name: control-plane-service-account
  namespace: gatling
---
apiVersion: rbac.authorization.k8s.io/v1
kind: Role
metadata:
  name: control-plane-role
  namespace: gatling
rules:
  - apiGroups: [""]
    resources: ["configmaps"]
    verbs: ["create", "get", "patch", "update", "delete"]
  - apiGroups: ["batch"]
    resources: ["jobs"]
    verbs: ["create", "deletecollection"]
---
apiVersion: rbac.authorization.k8s.io/v1
kind: RoleBinding
metadata:
  name: control-plane-role-binding
  namespace: gatling
roleRef:
  apiGroup: rbac.authorization.k8s.io
  kind: Role
  name: control-plane-role
subjects:
- kind: ServiceAccount
  name: control-plane-service-account
  namespace: gatling
---
apiVersion: apps/v1 
kind: Deployment 
metadata:
  name: control-plane
  namespace: gatling
spec:
  replicas: 1 
  selector: 
    matchLabels: 
      name: control-plane-template
  template: 
    metadata:
      labels:
        name: control-plane-template
    spec:
      serviceAccountName: control-plane-service-account
      containers:
        - name: control-plane
          image: gatlingcorp/control-plane:latest
          volumeMounts:
            - mountPath: /app/conf/
              name: control-plane-conf-volume
      volumes:
        - name: control-plane-conf-volume
          configMap:
            name: control-plane-config

Troubleshooting#

Run timeout while control plane logged a successful deployment#

The control plane initiated a request to the Kubernetes API to deploy load generators using Batch Job. Although the Kubernetes API responded with an OK status, the actual deployment of the underlying job may have failed.

First, inspect existing jobs and pods:

kubectl get jobs --namespace=gatling
kubectl get pods --namespace=gatling

If no resources are found, examine events related to jobs:

kubectl get events --namespace=gatling --field-selector involvedObject.kind=Job

Edit this page on GitHub