19 Common Kubernetes Failures and How to Fix Them

Problem 1: K8s cluster service access failure?

Root cause : The certificate cannot be recognized, often due to a custom or expired certificate.

Solution : Update the certificate.

Problem 2: K8s cluster service access failure (connection refused)

<code>curl: (7) Failed connect to 10.103.22.158:3000; Connection refused</code>

Root cause : Port mapping error – the service is running but not exposed correctly.

Solution: Delete the service (svc) and remap the port.

<code>kubectl delete svc nginx-deployment</code>

Problem 3: K8s cluster service exposure failure

<code>Error from server (AlreadyExists): services "nginx-deployment" already exists</code>

Root cause: The container has already exposed the service.

Solution: Delete the existing service and remap the port.

Problem 4: External network cannot access K8s service?

Root cause: The service type is ClusterIP, not exposed to the external network.

Solution: Change the service type to NodePort, allowing access via any cluster node.

<code>kubectl edit svc nginx-deployment</code>

Problem 5: Pod status is ErrImagePull?

<code>readiness-httpget-pod   0/1     ErrImagePull   0          10s</code>

Root cause: Image cannot be pulled.

Solution: Replace the image with a valid one.

Problem 6: Init container in pod stays abnormal

<code>NAME READY STATUS RESTARTS AGE
myapp-pod 0/1 Init:0/2 0 20s</code>

Root cause: Init container has not completed, causing the pod to remain in initializing state.

<code>Error from server (BadRequest): container "myapp-container" in pod "myapp-pod" is waiting to start: PodInitializing</code>

Solution: Create the required Service and add its name to CoreDNS so that the init container can resolve DNS.

<code>kubectl apply -f myservice.yaml</code>

Problem 7: Pod shows CrashLoopBackOff

Root cause: Image issue causing container restart failures.

Solution: Replace the faulty image.

Problem 8: Pod creation fails

<code>readiness-httpget-pod 0/1 Pending 0 0s
... (multiple lines showing CrashLoopBackOff and errors)</code>

Root cause: Image problem prevents container start.

Solution: Replace the image.

Problem 9: Pod ready state never becomes true

<code>readiness-httpget-pod 0/1 Running 0 116s</code>

Root cause: Command execution in the pod fails, unable to acquire resources.

Solution: Enter the container and create the resources defined in the YAML.

Problem 10: Pod creation fails due to YAML errors

Root cause: YAML file contains Chinese characters.

Solution: Correct the myregistrykey content.

Problem 11: kube-flannel pod status Init:0/1

Investigation command:

kubectl -n kube-system describe pod kube-flannel-ds-amd64-ndsf7

Root cause: Node kube-slave1 fails to pull the image.

Solution: Restart Docker on the node and manually pull the image; reinstall the plugin on the master if needed.

<code>kubectl create -f kube-flannel.yml; kubectl get nodes</code>

Problem 12: Service status ErrImagePull

Root cause: Incorrect image name.

Solution: Delete the faulty pod and pull the correct image.

<code>kubectl delete pod test-nginx; kubectl run test-nginx --image=10.0.0.81:5000/nginx:alpine</code>

Problem 13: Cannot enter specified container

Root cause: Duplicate "containers" field in YAML, so the pod lacks the intended container.

Solution: Remove the extra "containers" field and recreate the pod.

Problem 14: PersistentVolume creation fails

Root cause: Duplicate PV name.

Solution: Change the PV name.

Problem 15: Pod cannot mount PVC

Root cause: AccessModes of PVC do not match any available PV (requires >1G and RWO).

Solution: Adjust accessModes in the YAML or modify the PV.

Problem 16: Pod can’t access content after using PV

Root cause: NFS volume is empty or permissions are incorrect.

Solution: Create files in the NFS volume and set proper permissions.

Problem 17: Node status check fails

<code>Error from server (NotFound): the server could not find the requested resource (get services http:heapster:)</code>

Root cause: Heapster service is missing.

Solution: Install the Prometheus monitoring component.

Problem 18: Pod remains in Pending state

Root cause: Same image already deployed, leaving no node to schedule the new pod.

Solution: Delete all existing pods before deploying new ones.

Problem 19: Helm component installation fails

<code># helm install
Error: This command needs 1 argument: chart name
# helm install ./
Error: no Chart.yaml exists in directory "/root/hello-world"</code>

Root cause: Incorrect chart file name.

Solution: Rename Chart.yaml to the correct case (Chart.yaml).