Skip to main content
Bob decides to run Kibana inside the Kubernetes cluster so the team can inspect logs via a browser-based UI instead of raw JSON search output. This guide follows Bob’s workflow one file at a time and shows the manifests and commands he uses to get Kibana running and reachable. A quick taste of the kind of JSON documents Kibana will display (Elasticsearch search results):

Deploy Kibana (Deployment manifest)

Kibana is deployed with a small Deployment that creates one replica in the elastic-stack namespace and selects pods by the label app: kibana. The container exposes port 5601 — this is the port Kibana serves its web UI on.
Apply the manifest and watch the pod status. Initially the pod may show ContainerCreating while the node pulls the Kibana image (the image is relatively large), then transition to Running:
The pod has an IP, but it is an internal cluster address — not reachable directly from Bob’s laptop and subject to change if the pod is recreated.
A diagram titled "An Address Inside the Cluster" showing a user labeled "Bob" with a browser trying to reach a cluster. The request is blocked at the cluster boundary while an orange "dead" server/pod labeled 10.244.1.7 is shown inside.
Pods have internal IPs by design. To expose a pod so an external browser can reach it, you must create a Kubernetes Service (or use an Ingress / LoadBalancer). NodePort gives you a port on each node that forwards to the pod.

Expose Kibana with a NodePort Service

To let a browser outside the cluster reach Kibana, Bob exposes it with a Service of type NodePort. The Service selects pods with app: kibana and maps the cluster port 5601 to a node port (for example, 30601). Any request to node:30601 on any node is forwarded to the pod port 5601.
The image, titled "A Door Number on the Node," shows a stylized server node with a blue smiling mascot labeled "5601" and a door icon labeled "30601." A browser/search window is displayed on the left.
Example Service manifest:
Apply it and verify the service:
The PORT(S) column shows 5601:30601/TCP — the cluster port and the node port. From any machine with network access to the cluster nodes, Bob can sanity-check Kibana with a simple HTTP request:
Then he opens the Kibana UI in a browser at http://node01:30601/.

Create an index pattern and explore logs

When Bob opens Kibana and clicks Discover, Kibana asks which data (index) to view. It lists available indices (for example fluentd-YYYY.MM.DD). Rather than repeatedly entering the date-stamped index name, Bob creates an index pattern using a wildcard like fluentd-* so the pattern matches current and future indices with that prefix.
A dark-themed screenshot titled "The Wildcard" showing a "Create index pattern" dialog with the pattern "fluentd-*" matching indices (fluentd-2026.10.12, fluentd-2026.10.13). A smiling circular user avatar labeled "Bob" appears to the right.
After clicking Next step, Kibana asks which field contains the timestamp. Bob selects the timestamp field (commonly @timestamp) and creates the index pattern. Kibana inspects the documents in the matched indices and displays discovered fields. Back in Discover the interface populates: a time histogram shows log activity over the selected time range, and individual log lines appear below (newest first). These logs were generated by the application pod, collected by Fluentd on the node, stored in Elasticsearch, and are now visible in Kibana. Example log lines:
To filter results, Kibana supports KQL (Kibana Query Language). Switching the search mode to KQL and searching for:
returns only log lines that match USER1:
KQL makes it quick for team members to filter logs without learning complex commands. From a filter or discover view Bob can create visualizations and pin them to a dashboard — enabling quick operational visibility. Quick reference of what Bob used to get Kibana reachable and searchable:
Caveat: Kibana depends on Elasticsearch. If Elasticsearch becomes resource-constrained (low memory, high load, or degraded cluster state), Kibana’s Discover view may not show errors — it may simply appear quiet or the histogram can flatten. Implement monitoring and alerting for Elasticsearch cluster health and node resources so the team is notified when the backend is unhealthy or searches stop returning expected results.
With the Deployment, NodePort Service, and an index pattern in Kibana, Bob has an end-to-end browser workflow to explore logs, build visualizations, and create dashboards for his team.

Watch Video

Practice Lab