Kubernetes中Kibana Pod无法连接Elastic集群问题求助

Let's break down your issue and walk through the most likely fixes:

You've successfully deployed Elasticsearch 6.5.4 and Kibana 6.5.4 via Helm in the elastic namespace. From inside the Kibana pod, curl -v http://elasticsearch-client:9200 returns a healthy 200 OK with cluster info, but when accessing Kibana via your browser, it can't connect to Elasticsearch. Here's what to check:

1. Verify Kibana's External Exposure Method #

First, confirm how you're exposing Kibana to external traffic—NodePort, LoadBalancer, or Ingress?

• NodePort: Make sure you're accessing [Node-IP]:[NodePort] (not the pod's IP, which is only reachable inside the cluster). Get the NodePort with:

  kubectl get svc -n elastic kibana
  

• LoadBalancer: Check if the LoadBalancer has been assigned an external IP, and that IP is reachable from your network.
• Ingress: Validate that your Ingress rules point to the Kibana service, your Ingress controller is running, and the domain resolves correctly to the Ingress endpoint.

2. Check Kibana Service Type #

By default, the Elastic Helm chart might create a ClusterIP service for Kibana—this only allows internal cluster access. To expose it externally, update the service type:
Either via Helm upgrade:

helm upgrade kibana elastic/kibana -n elastic --set service.type=NodePort

Or edit the service directly:

kubectl edit svc kibana -n elastic

Change spec.type to NodePort or LoadBalancer, then save the changes.

3. Validate Kibana's Elasticsearch Configuration #

Even though your pod's startup command specifies --elasticsearch.url=http://elasticsearch-client:9200, Kibana might be pulling settings from its kibana.yml config file. Check the file:

kubectl exec -it <your-kibana-pod-name> -n elastic -- cat /usr/share/kibana/config/kibana.yml

Ensure elasticsearch.url matches the working internal address. If not, override it with Helm:

helm upgrade kibana elastic/kibana -n elastic --set elasticsearch.url=http://elasticsearch-client.elastic.svc.cluster.local:9200

4. Check Kibana's Host Binding #

If Kibana is only bound to localhost or its pod's internal IP, external browsers won't be able to connect. Verify the server.host setting:

kubectl exec -it <your-kibana-pod-name> -n elastic -- cat /usr/share/kibana/config/kibana.yml | grep server.host

If it's set to anything other than 0.0.0.0, update it via Helm:

helm upgrade kibana elastic/kibana -n elastic --set server.host=0.0.0.0

This ensures Kibana accepts connections from any address.

5. Inspect Kibana Logs for Clues #

Check the Kibana pod logs for runtime errors or connection issues that might not be visible from the browser:

kubectl logs <your-kibana-pod-name> -n elastic

Look for entries related to Elasticsearch connection failures or service startup problems.

Most Likely Root Causes #

In most cases, this issue boils down to:

• Kibana's service isn't exposed externally (still ClusterIP)
• server.host is restricted to internal-only access
• Rarely, misconfigured proxy settings causing the browser to try accessing the internal Elasticsearch service name directly (which it can't resolve)

Start with verifying the service type and external access method—this is the most common fix.

内容的提问来源于stack exchange，提问作者Riduidel