troubleshooting.md

Troubleshooting

Azure Cloud Shell is the most convenient way to troubleshoot any problems with your AKS and AGIC installation. Launch your shell from shell.azure.com or by clicking the link:

Test with a simple Kubernetes app

The steps below assume:

• You have an AKS cluster, with Advanced Networking enabled
• AGIC has been installed on the AKS cluster
• You already hav an App Gateway on a VNET shared with your AKS cluster

To verify that the App Gateway + AKS + AGIC installation is setup correctly, deploy the simplest possible app:

cat <<EOF | kubectl apply -f -
apiVersion: v1
kind: Pod
metadata:
  name: test-agic-app-pod
  labels:
    app: test-agic-app
spec:
  containers:
  - image: "mcr.microsoft.com/dotnet/core/samples:aspnetapp"
    name: aspnetapp-image
    ports:
    - containerPort: 80
      protocol: TCP
---
apiVersion: v1
kind: Service
metadata:
  name: test-agic-app-service
spec:
  selector:
    app: test-agic-app
  ports:
  - protocol: TCP
    port: 80
    targetPort: 80
---
apiVersion: extensions/v1beta1
kind: Ingress
metadata:
  name: test-agic-app-ingress
  annotations:
    kubernetes.io/ingress.class: azure/application-gateway
spec:
  rules:
    - host: test.agic.contoso.com
      http:
        paths:
          - path: /
            backend:
              serviceName: test-agic-app-service
              servicePort: 80
EOF

Copy and paste all lines at once from the script above into a Azure Cloud Shell. Please ensure the entire command is copied - starting with cat and including the last EOF.

After a successful deployment of the app above your AKS cluster will have a new Pod, Service and an Ingress.

Get the list of pods with Cloud Shell: kubectl get pods -o wide. We expect for a pod named 'test-agic-app-pod' to have been created. It will have an IP address. This address must be within the VNET of the App Gateway, which is used with AKS.

Get the list of services: kubectl get services -o wide. We expect to see a service named 'test-agic-app-service'.

Get the list of the ingresses: kubectl get ingress. We expect an Ingress resource named 'test-agic-app-ingress' to have been created. The resource will have a host name 'test.agic.contoso.com'.

One of the pods will be AGIC. kubectl get pods will show a list of pods, one of which will begin with 'ingress-azure'. Get all logs of that pod with kubectl logs <name-of-ingress-controller-pod> to verify that we have had a successful deployment. A successful deployment would have added the following lines to the log:

I0927 22:34:51.281437       1 process.go:156] Applied App Gateway config in 20.461335266s
I0927 22:34:51.281585       1 process.go:165] cache: Updated with latest applied config.
I0927 22:34:51.282342       1 process.go:171] END AppGateway deployment

Alternatively, from Cloud Shell we can retrieve only the lines indicating successful App Gateway configuration with kubectl logs <ingress-azure-....> | grep 'Applied App Gateway config in', where <ingress-azure....> should be the exact name of the AGIC pod.

App Gateway will have the following configuration applied:

• Listener:

• Routing Rule:

• Backend Pool:

  • There will be one IP address in the backend address pool and it will match the IP address of the Pod we observed earlier with kubectl get pods -o wide

Finally we can use the cURL command from within Cloud Shell to establish an HTTP connection to the newly deployed app:

1. Use kubectl get ingress to get the Public IP address of App Gateway
2. Use curl -I -H 'test.agic.contoso.com' <publitc-ip-address-from-previous-command>

A result of HTTP/1.1 200 OK indicates that the App Gateway + AKS + AGIC system is working as expected.

Inspect Kubernetes Installation

Pods, Services, Ingress

Application Gateway Ingress Controller (AGIC) continuously monitors the folowing Kubernetes resources: Deployment or Pod, Service, Ingress

The following must be in place for AGIC to function as expected:

1. AKS must have one or more healthy pods. Verify this from Cloud Shell with kubectl get pods -o wide --show-labels If you have a Pod with an apsnetapp, your output may look like this:

   delyan@Azure:~$ kubectl get pods -o wide --show-labels
   
   NAME                   READY   STATUS    RESTARTS   AGE   IP          NODE                       NOMINATED NODE   READINESS GATES   LABELS
   aspnetapp              1/1     Running   0          17h   10.0.0.6    aks-agentpool-35064155-1   <none>           <none>            app=aspnetapp

2. One or more services, referencing the pods above via matching selector labels. Verify this from Cloud Shell with kubectl get services -o wide

   delyan@Azure:~$ kubectl get services -o wide --show-labels
   
   NAME                TYPE        CLUSTER-IP     EXTERNAL-IP   PORT(S)   AGE   SELECTOR        LABELS
   aspnetapp           ClusterIP   10.2.63.254    <none>        80/TCP    17h   app=aspnetapp   <none>     

3. Ingress, annotated with kubernetes.io/ingress.class: azure/application-gateway, referencing the service above Verify this from Cloud Shell with kubectl get ingress -o wide --show-labels

   delyan@Azure:~$ kubectl get ingress -o wide --show-labels
   
   NAME        HOSTS   ADDRESS   PORTS   AGE   LABELS
   aspnetapp   *                 80      17h   <none>

4. View annotations of the ingress above: kubectl get ingress aspnetapp -o yaml (substitute aspnetapp with the name of your ingress)

   delyan@Azure:~$ kubectl get ingress aspnetapp -o yaml
   
   apiVersion: extensions/v1beta1
   kind: Ingress
   metadata:
     annotations:
       kubernetes.io/ingress.class: azure/application-gateway
     name: aspnetapp
   spec:
     backend:
       serviceName: aspnetapp
       servicePort: 80

   The ingress resource must be annotated with kubernetes.io/ingress.class: azure/application-gateway.

Verify Observed Nampespace

• Get the existing namespaces in Kubernetes cluster. What namespace is your app running in? Is AGIC watching that namespace? Refer to the Multiple Namespace Support documentation on how to properly configure observed namespaces.

# What namespaces exist on your cluster
kubectl get namespaces

# What pods are currently running
kubectl get pods --all-namespaces -o wide

• The AGIC pod should be in the default namespace (see column NAMESPACE). A healthy pod would have Running in the STATUS column. There should be at least one AGIC pod.

# Get a list of the Application Gateway Ingress Controller pods
kubectl get pods --all-namespaces --selector app=ingress-azure

• If the AGIC pod is not healthy (STATUS column from the command above is not Running):

  • get logs to understand why: kubectl logs <pod-name>
  • for the previous instance of the pod: kubectl logs <pod-name> --previous
  • describe the pod to get more context: kubectl describe pod <pod-name>

• Do you have a Kubernetes Service and Ingress resources?

# Get all services across all namespaces
kubectl get service --all-namespaces -o wide

# Get all ingress resources across all namespaces
kubectl get ingress --all-namespaces -o wide

• Is your Ingress annotated with: kubernetes.io/ingress.class: azure/application-gateway? AGIC will only watch for Kubernetes Ingress resources that have this annotation.

# Get the YAML definition of a particular ingress resource
kubectl get ingress --namespace  <which-namespace?>  <which-ingress?>  -o yaml

• AGIC emits Kubernetes events for certain critical errors. You can view these:
  • in your terminal via kubectl get events --sort-by=.metadata.creationTimestamp
  • in your browser using the Kubernetes Web UI (Dashboard)

Logging Levels

AGIC has 3 logging levels. Level 1 is the default one and it shows minimal number of log lines. Level 5, on the other hand, would display all logs, including sanitized contents of config applied to ARM.

The Kubernetes community has established 9 levels of logging for the kubectl tool. In this repository we are utilizing 3 of these, with similar semantics:

Verbosity	Description
1	Default log level; shows startup details, warnings and errors
3	Extended information about events and changes; lists of created objects
5	Logs marshaled objects; shows sanitized JSON config applied to ARM

The verbosity levels are adjustable via the verbosityLevel variable in the helm-config.yaml file. Increase verbosity level to 5 to get the JSON config dispatched to ARM:

• add verbosityLevel: 5 on a line by itself in helm-config.yaml and re-install
• get logs with kubectl logs <pod-name>