Skip to main content
Reading raw JSON from curl is no way for a team to investigate. Take Bob’s cluster. It’s Monday morning and support says User 1 had a rough weekend on the website. The raw document sitting in Elasticsearch looks like this:
Every line the app wrote is stored in Elasticsearch as a separate document. To get a single answer with curl, Bob must remember the search endpoint, build a long URL, add a size so results aren’t truncated, and then scan a wall of brackets looking for timestamps:
Then a teammate leans over and asks, “Can you check user3 too?” and the whole thing starts again. Without meaning to, Bob becomes the team’s search engine. That doesn’t scale for a team, and it barely works for an individual. What the team actually needs is a web tool that sits in front of Elasticsearch, accepts human-friendly questions, and shows answers as rows and charts instead of raw JSON. That tool is Kibana. Kibana is just a web application — a page in your browser that asks Elasticsearch questions for you. It doesn’t store log lines itself; every time you search, Kibana sends an HTTP request to Elasticsearch’s REST API (the same API Bob used with curl) and then renders whatever comes back.
A minimalist diagram titled "Just a Web Page" showing a user labeled "Bob" sending requests to Kibana, which then queries Elasticsearch. A small browser-style panel at the bottom-left displays sample search/log entries for USER1.
Where does Kibana run inside the cluster? It runs like any other stateless web app: in a Deployment. If a Kibana pod dies, nothing is lost — Kibana holds no primary data — and a fresh pod can start and reconnect to Elasticsearch. By contrast, Elasticsearch requires stable network identities and persistent storage, so it typically runs in a StatefulSet. In our lab, Bob’s Kibana runs in a Deployment named kibana inside the elastic-stack boundary. It’s Kibana version 7.1.0 and listens on port 5601 (Kibana’s default port).
A stylized diagram titled "Bob's Kibana Deployment" showing a kibana service inside an elastic-stack boundary. It features a blue mascot on a platform labeled "kibana" with tags "kibana:7.1.0" and "Port 5601."
A running Kibana pod still isn’t directly reachable from a developer’s laptop. To open it in a browser you expose it via a Kubernetes Service. In the lab that service is a NodePort which opens a fixed port on the node and forwards traffic to port 5601 on the Kibana pod. Bob types the node address and the NodePort into his browser: node01:30601 Now Kibana’s UI is open. But Kibana doesn’t yet know which logs Bob cares about. Elasticsearch can contain many indices — for example, Fluentd often creates indices named like fluentd-YYYY.MM.DD. Bob tells Kibana to look at every index whose name starts with fluentd- by creating an index pattern (for example: fluentd-*) and selecting the timestamp field so results can be filtered and sorted by time.
Create an index pattern such as fluentd-* and choose the appropriate timestamp field. This enables time filtering and makes searches (KQL) and Discover results behave correctly.
With an index pattern saved, Bob opens Discover. Discover is a scrolling list of documents (newest first) with a search bar on top and a time picker beside it. The search bar understands KQL (Kibana Query Language). Bob types log:USER1 and presses Update; the list shrinks to only User 1’s events:
His teammate wants User 3 — change one word and the answer updates immediately. A question that took minutes with curl now takes seconds, and anyone with access to the Kibana page can ask it. Discover is for reading individual log lines, but Kibana can also turn searches into visualizations. For example, you can create a bar chart that counts how often each user appears in the logs and then pin that chart to a dashboard the team keeps open all day. Dashboards can be exported and imported as NDJSON files so one person’s dashboard becomes everyone’s.
A dark-themed slide titled "Charts and Dashboards" showing a large bar chart labeled "Lines per User" comparing USER1–USER4 on the left and a panel of four smaller bar-chart widgets on the right. A small file icon labeled "export.ndjson" appears below the dashboard.
This example exposes Kibana via a NodePort with no login or password. Do not do this in production. Use authentication, RBAC, and network controls for any real deployment.
An infographic titled "No Login, No Password" showing three user icons and "USER1" labels on the left, with a lock symbol and an arrow pointing to an open door. On the right a smiling blue mascot sits on a platform, suggesting access without credentials.
When should you add Kibana? If you’re the only person reading logs and you know exactly which line you want, curl is fine. The moment a second person needs to ask their own question, or you need to see patterns over time rather than a single line, put Kibana in front of Elasticsearch. Bob’s team crossed that line on Monday morning — and Kibana made their investigations much faster and easier. Summary: Kibana provides a friendly web UI for Elasticsearch queries, time-series exploration, and team-shared dashboards. Use a Deployment for Kibana, a StatefulSet for Elasticsearch, expose Kibana securely (avoid unauthenticated NodePorts in production), and use index patterns + KQL to get fast, human-readable results. Resources and quick references: Links and references

Watch Video