Test

In this section, you will try to configure and deploy a "problematic" workload to the cluster with the K8sGPT blueprint. We will then verify whether it was able to identify the issue and assist the user with a recommendation for resolution.

Step 1: Create Namespace¶

In this step, you will create a namespace for the a sample workload that will be published to the cluster. The sample workload will be used to see how K8sGPT can scan and identify issues.

Select Infrastructure -> Namespaces
Click New Namespace
Enter the name sample-workload
Select wizard for the type
Click Save

Workload Namespace

Click Save & Go to Placement

Workload Namespace

Select the cluster where the K8sGPT operator is installed
Click Save & Go to Publish

Workload Namespace

Click Publish

Workload Namespace

Click Exit

Step 2: Create Workload¶

In this step, you will create a a workload which will be used to rest the K8sGPT operator.

Save the below YAML to a file named sample-workload.yaml

Note

This service contains a selector that does not map to an existing pod. This will be picked up by K8sGPT.

apiVersion: v1
kind: Service
metadata:
  name: myapp
spec:
  type: ClusterIP
  ports:
  - name: myapp
    protocol: TCP
    port: 80
  selector:
    app: nonexistentapp

Select Applications -> Workloads
Click New Workload -> Create New Workload
Enter a name for the workload
Select K8s YAML for the package type
Select Upload files manually
Select sample-workload for the namespace
Click Continue

Workload

Click Choose File and select the previously saved sample-workload.yaml file
Click Save & Go to Placement

Workload

Select the cluster where the K8sGPT operator is installed
Click Save & Go to Publish

Workload

Click Publish

Workload

Click Exit

Step 3: View K8sGPT Results¶

In this step, you will view the results of the K8sGPT Operator.

Execute the following command

kubectl get results -A

You will see a result like the following providing details of the issue and possible resolution in simple terms.

{
    "apiVersion": "v1",
    "items": [
        {
            "apiVersion": "core.k8sgpt.ai/v1alpha1",
            "kind": "Result",
            "metadata": {
                "creationTimestamp": "2023-05-09T19:55:28Z",
                "generation": 1,
                "name": "sampleworkloadmyapp",
                "namespace": "kube-system",
                "resourceVersion": "54316",
                "uid": "76646698-a0f1-4632-9aad-2b0de48b2b44"
            },
            "spec": {
                "details": "The service is looking for endpoints with the label \"app=nonexistentapp\", but it cannot find any. \n\nThe solution is to make sure that the label \"app=nonexistentapp\" is correctly assigned to one or more pods that are running the application. This can be done by checking the labels of the pods using the \"kubectl get pods --show-labels\" command and verifying that the \"app=nonexistentapp\" label is present. If it is not present, add it using the \"kubectl label pods \u003cpod-name\u003e app=nonexistentapp\" command. Once the label is correctly assigned, the endpoints should be discovered by the service and the error should be resolved.",
                "error": [
                    {
                        "sensitive": [
                            {
                                "masked": "XUBj",
                                "unmasked": "app"
                            },
                            {
                                "masked": "azUvWklzT0pEUU1lMjA=",
                                "unmasked": "nonexistentapp"
                            }
                        ],
                        "text": "Service has no endpoints, expected label app=nonexistentapp"
                    }
                ],
                "kind": "Service",
                "name": "sample-workload/myapp",
                "parentObject": ""
            }
        }
    ],
    "kind": "List",
    "metadata": {
        "resourceVersion": ""
    }
}

Recap¶

Congratulations! You have successfully deployed the K8sGPT Operator on your managed Kubernetes cluster as an add-on in a custom cluster blueprint. You then tested the efficacy of the K8sGPT operator with a broken/misconfigured workload.