Rediģēt

Kopīgot, izmantojot


Basic troubleshooting of AKS cluster startup issues

This article outlines the basic troubleshooting methods to use if you can't start a Microsoft Azure Kubernetes Service (AKS) cluster successfully.

Prerequisites

View errors from Azure CLI

When you start clusters by using Azure CLI, errors are recorded as output if the operation fails. Here's how a command, user input, and operation output might appear in a Bash console:

$ az aks start --resource-group myResourceGroup --name MyManagedCluster

(VMExtensionProvisioningError) Unable to establish outbound connection from agents, please see https://learn.microsoft.com/en-us/troubleshoot/azure/azure-kubernetes/error-code-outboundconnfailvmextensionerror and https://aka.ms/aks-required-ports-and-addresses for more information.  
Details: instance 3 has extension error details : {vmssCSE error messages : {vmssCSE exit status=50, output=AGE_SHA=sha-16fd35

These errors often contain detailed descriptions of what went wrong in the cluster start operation, and they provide links to articles that contain more details. Additionally, you can use our troubleshooting articles as a reference based on the error that an Azure CLI operation produces.

View error details in the Azure portal

To view the details about errors in the Azure portal, examine the Azure activity log. To find the list of activity logs in the Azure portal, search on Activity log. Or, select Notifications (the bell icon), and then select More events in the activity log.

The list of logs on the Activity log page contains a line entry in which the Operation name column value is named Start Managed Cluster. The corresponding Event initiated by column value is set to the name of your work or school account. If the operation is successful, the Status column value shows Accepted.

Screenshot of an AKS cluster activity log blade showing a failed start operation.

What if an error occurred instead? In that case, the Start Managed Cluster operation Status field shows Failed. Unlike in the operations to create cluster components, here you must expand the failed operation entry to review the suboperation entries. Typical suboperation names are policy actions, such as 'audit' Policy action and 'auditIfNotExists' Policy action. Some suboperations will continue to show that they succeeded.

To further investigate, you can select one of the failed suboperations. A side pane opens so that you can review more information about the suboperation. You can troubleshoot values for fields such as Summary, JSON, and Change history. The JSON field contains the output text for the error in JSON format, and it usually provides the most helpful information.

Screenshot of a suboperation side pane of an AKS cluster activity log showing why a start operation failed.

View cluster insights

You can also generate cluster insights to help you troubleshoot via the Diagnose and solve problems blade in the Azure portal. To access this feature, follow these steps:

  1. In the Azure portal, search for and select Kubernetes services.

  2. Select the name of your AKS cluster.

  3. In the navigation pane of the AKS cluster page, select Diagnose and solve problems.

  4. On the Diagnose and solve problems page, select the Cluster insights link. The cluster insights tool analyzes your cluster, and then provides a list of its findings in the Observations and Solutions section of the Cluster Insights page.

  5. Select one of the findings to view more information about a problem and its possible solutions.

View resources in the Azure portal

In the Azure portal, you might want to view the resources that were created when the cluster was built. Typically, these resources are in a resource group that begins with MC_. The managed cluster resource group might have a name such as MC_MyResourceGroup_MyManagedCluster_<location-code>. However, the name may be different if you built the cluster by using a custom-managed cluster resource group.

To find the resource group, search for and select Resource groups in the Azure portal, and then select the resource group in which the cluster was created. The resource list is shown on the Overview page of the resource group.

Warning

We recommend that you don't modify resources in the MC_ resource group. This action might cause unwanted effects on your AKS cluster.

To review the status of a virtual machine scale set, you can select the scale set name within the resource list of the resource group. It might have a Name similar to aks-nodepool1-12345678-vmss and a Type value of Virtual machine scale set. The status of the scale set appears at the top of the node pool's Overview page, and more details are shown in the Essentials heading. If the deployment was unsuccessful, the displayed status is Failed.

For all resources, you can review details to better understand why the deployment failed. For a scale set, you can select the Failed status text to view details about the failure. The details are in a row that contains the Status, Level, and Code columns. The following example shows a row of column values.

Column Example value
Status Provisioning failed
Level Error
Code ProvisioningState/failed/VMExtensionProvisioningError

Select the row to see the Message field. This contains even more information about that failure. For example, the Message field for the example row begins with the following text:

VM has reported a failure when processing extension 'vmssCSE'. Error message: "Enable failed: failed to execute command: command terminated with exit status=50 [stdout] [stderr] 0 0 0 --: Armed with this information, you can conclude that the VMs in the scale set failed and generated exit status 50.

Use Kubectl commands

For another option to help troubleshoot errors on your cluster, enter kubectl commands to get details about the resources that were deployed in the cluster. To use kubectl, first sign in to your AKS cluster:

az aks get-credentials --resource-group MyResourceGroup --name MyManagedCluster

Depending on the type of failure and when it occurred, you might not be able to sign in to your cluster to get more details. But in general, if your cluster was created and shows up in the Azure portal, you should be able to sign in and run kubectl commands.

View cluster nodes (kubectl get nodes)

To get more details to determine the state of the nodes, view the cluster nodes by entering the kubectl get nodes command. In this example, no nodes are reporting in the cluster:

$ kubectl get nodes

No resources found

View pods in the system namespace (kubectl get pods)

Viewing the pods in the kube-system namespace is also a good way to troubleshoot your issue. This method lets you view the status of the Kubernetes system pods. In this example, we enter the kubectl get pods command:

$ kubectl get pods -n kube-system
NAME                                  READY   STATUS    RESTARTS   AGE
coredns-845757d86-7xjqb               0/1     Pending   0          78m
coredns-autoscaler-5f85dc856b-mxkrj   0/1     Pending   0          77m
konnectivity-agent-67f7f5554f-nsw2g   0/1     Pending   0          77m
konnectivity-agent-8686cb54fd-xlsgk   0/1     Pending   0          65m
metrics-server-6bc97b47f7-dfhbr       0/1     Pending   0          77m

Describe the status of a pod (kubectl describe pod)

By describing the status of the pods, you can view the configuration details and any events that have occurred on the pods. Run the kubectl describe pod command:

$ kubectl describe pod coredns-845757d86-7xjqb -n kube-system
Name:                 coredns-845757d86-7xjqb
Namespace:            kube-system
Priority:             2000001000
Priority Class Name:  system-node-critical
Node:                 <none>
Labels:               k8s-app=kube-dns
                      kubernetes.io/cluster-service=true
                      pod-template-hash=845757d86
                      version=v20
...
Events:
  Type     Reason            Age                 From               Message
  ----     ------            ----                ----               -------
  Warning  FailedScheduling  24m (x1 over 25m)   default-scheduler  no nodes available to schedule pods
  Warning  FailedScheduling  29m (x57 over 84m)  default-scheduler  no nodes available to schedule pods

In the command output, you can see that the pod can't be deployed to a node because no nodes are available.

Contact us for help

If you have questions or need help, create a support request, or ask Azure community support. You can also submit product feedback to Azure feedback community.