Use Kibana when Elasticsearch is already your data platform
Kibana's source build stopped after 1 second in our sandbox because the running Node version didn't match the required version. That is a contributor-environment finding. For a team choosing a dashboard product, the more consequential requirement is Elasticsearch: Kibana's README defines it as the interface for querying, analyzing, visualizing and managing data stored there.
Choose it when that relationship is useful. A team already putting operational data into Elasticsearch has a clear reason to evaluate its companion interface. Someone seeking charts over unrelated databases should compare other tools before adopting a search cluster just to obtain the UI. Kibana's source checkout also deserves a separate decision from using a packaged release.
What happened when we ran it
Our November 11, 2026 run at 650d1d3 installed dependencies successfully in 52 seconds, then failed the build with exit 1 after 1 second.
| Recorded step | Result | Duration |
|---|---|---|
| Install | Succeeded; 1 dependency recorded; 2,423 MB on disk | 52 seconds |
| Build | Exit 1; rejected the running Node version | 1 second |
| Tests | No script or target detected by the harness; skipped | Not run |
The build command was node scripts/build --all-platforms. Its error says: Kibana does not support the current Node.js version v24.19.0. Please use Node.js v24.22.1. The sandbox image was labelled lab-node:22, yet this exact log establishes Node v24.19.0 for that command. The image label cannot override the observed runtime.
The supported next step is specific: use the requested runtime for the measured checkout and repeat the build. There is no measured result for that repeat. Calling Kibana's source code broken on this evidence would mistake a version check for a compilation failure.
The configuration specified 3 CPUs and 8 GB RAM. The checkout contained 131,803 files, approximately 15,186,208 source lines and 1,224.3 MB before the recorded installation footprint. The summary's single recorded dependency isn't a credible inventory of every package in this workspace repository. Keep the reported field, but don't turn it into a claim that Kibana has only one dependency.
The scan found 58 CI workflow files, monorepo workspaces, no Dockerfile and no tests directory. Those are scanner observations. In particular, the harness finding no test target doesn't mean the project has no tests.
Source development and trying the product are separate routes
The README points production users and evaluators toward downloadable releases or Elastic's hosted service. It describes local source work as a route for contributing code, exploring recent features or trying an open pull request. That distinction matters with a checkout exceeding a gigabyte before the recorded installation footprint.
For a product evaluation, I'd start with a release and a compatible Elasticsearch instance. Enter the source tree when you need to change Kibana itself. If source development is the job, pin the checkout and its requested runtime together before interpreting a failed command. Our build stopped early enough that it says little about the work remaining after that check.
Elasticsearch compatibility constrains the upgrade sequence
Kibana recommends running the same version number as Elasticsearch. The README allows Kibana to be ahead on patch versions or behind on minor versions, while warning that support will first request matching versions when troubleshooting a mismatch.
Its compatibility table marks an Elasticsearch minor older than Kibana as fatal, along with the illustrated major-version mismatches. The major-version rule also mentions an exception for prevmajor.last; don't replace that qualification with a blanket claim that every unequal version fails. For an operator, the useful action is to plan the pair's upgrade sequence against the documented relationships before upgrading either component independently.
A current issue shows why the missing test target needs careful wording
Issue #305485, opened November 14, reports an Entity ML anomaly-overview API check that expected HTTP 200 and received HTTP 502. It names a Scout/Playwright specification under the security solution and was open when read. That report is about a particular upstream check, separate from our Node-version rejection.
The release page lists Kibana 9.5.6 as the latest release when fetched. The recent release and new issue show ongoing release and tracking activity. They establish neither a clean local test run nor a support-response guarantee.
Grafana and OpenSearch Dashboards fit different data choices
Grafana documents mixing different data sources in the same graph. Evaluate it when your dashboard needs to span systems rather than center on Elasticsearch. That is a capability distinction, without a measured performance winner here.
OpenSearch Dashboards is designed to work with OpenSearch. If that is your search engine, start with its matching interface. Don't select Kibana on the assumption that all search backends are interchangeable.
Read the applicable licence headers before redistributing code
Kibana's licence file says coverage depends on the applicable source header. It describes a default AGPLv3-only, SSPL or Elastic License 2.0 choice, other Apache-compatible coverage, and code solely under Elastic License 2.0 in x-pack. A single permissive licence label would conceal those differences. Check the components and distribution you intend to use.
What we have not tested yet
This run didn't produce a build under the requested Node version or launch Kibana against Elasticsearch. Dashboard behavior, authentication, migrations and performance under real data volumes remain outside the evidence. Community research was not collected. Anonymous repository-facts and latest-release API requests returned HTTP 403, so fresh API counts and a last-push timestamp weren't established; the public release page remained readable.
All four ratings are unassessed: the rejected runtime doesn't support a setup-ease score, and reading the README and one issue doesn't establish documentation completeness, support quality or production maturity. For contributors, the next useful result is a build with the requested runtime. For adopters, it's a matching release pair answering a real question against your data.

