OpenMetadata ‒ An introduction

OpenMetadata is one of the fastest growing open-source metadata platforms with, as of 2026, having over 3,000 enterprise deployments, 13,000+ open-source community members, and 430+ code contributors. With more than 130 data connectors, OpenMetadata makes it easy to import metadata from every component of your end-to-end data pipeline.

That said, importing metadata is only the means. The real goal is governance: knowing what data you have, where it came from, whether you can trust it, and who is accountable for it. OpenMetadata gets you there through three connected capabilities: Discovery, Lineage, and Observability. These feed directly into governance rather than standing apart from it. In this post, we'll walk through how these pieces fit together.

Ingest Your Metadata

Before diving in, let's look at how metadata gets into OpenMetadata in the first place. Ingestion is a straightforward two-step process.

The first step is connecting to your data source using one of the 130+ built-in turnkey connectors, covering sources such as:

• Databases (MySQL, PostgreSQL, Oracle, Delta Lake, ...)

• Pipelines (Airflow, Glue, Databricks, ...)

• Storage (GCS, S3 Storage)

• Dashboards (Power BI, Tableau, ...)

Once the connection is made, metadata ingestion can be configured using metadata agents. These agents can be scheduled to ingest metadata, determine lineage, import data usage, profile data, and run data quality checks.

Metadata agents used for ingesting metadata into the platform. These agents can be scheduled for systematic, ongoing ingestion.

This ingestion layer is what makes everything downstream possible: every governance decision in OpenMetadata (who owns what, which data can be trusted, how a dataset may be used) is only as good as the metadata feeding it. With that foundation in place, the platform's capabilities can be put to work.

Data Discovery: Knowing What You Have

Governance starts with data visibility as you cannot govern data you cannot find. The first pillar, Data Discovery, lets users navigate complex, large-scale datasets spread across many source platforms. Discovery happens through OpenMetadata's search capabilities, letting users query the entire metadata catalogue, filter results, and preview the content of any entity.

Data Discovery using the Search interface, showing a comprehensive listing of found data assets. The interface allows filtering on metadata features, sorting by category, and previewing the underlying data.

In a governance context, discovery is what gives ownership and policy any teeth. An access policy or a glossary term is meaningless if nobody can find the asset it applies to. Discovery is what connects governance rules to the actual data they're meant to govern.

Lineage: Tracing Accountability Through the Pipeline

If discovery tells you what you have, lineage tells you where it came from and where it's going, which is essential once you're trying to hold anyone accountable for data quality or compliance. Lineage is found in the lineage tab on a data asset's page, displaying the full upstream and downstream relationships relative to that asset.

Lineage can be viewed at both the table and column level (where supported), enabling origin traceability and detailed impact analysis. It's supported across databases, pipelines, and dashboards.

Lineage view showcasing table- and column-level lineage. The connection between PostgreSQL and Databricks assets, linked by an Airflow pipeline edge, shows that this transfer step is orchestrated using Apache Airflow.

Lineage matters for governance in a very concrete way: when a regulator, auditor, or internal stakeholder asks "where did this number come from?" or "what breaks if we change this column?", lineage is the mechanism that answers it. Without it, governance policies exist on paper but can't actually be enforced or verified against the real data flow.

Observability: Making Governance Continuous

Discovery and lineage give you a map. Observability is what tells you whether the data on that map is actually trustworthy right now, turning governance from a one-time exercise into an ongoing practice.

The observability pillar covers data quality, data profiling, incident management, and alerts.

Deploying the data profiling agent lets OpenMetadata pull sample data from the source and apply profiling aggregations (row count, min, max, average, and so on). This enriches the metadata with real sample data, supporting both quality checks and quick previews.

Data quality tests can be built directly in the OpenMetadata UI or imported from the source asset.

The alerting and notification system closes the loop: users can configure custom alerts for events like column renames, failing test cases, or new announcements on a data asset. This is what lets governance scale, instead of manually re-checking that policies still hold, teams get notified the moment something drifts out of compliance.

The Sample Data tab lets users view (and mask, where sensitive) a portion of an asset's data. The Data Observability tab surfaces profiling results, quality test outcomes, and incident management for failed or aborted tests.

Governance: Where It All Comes Together

Discovery, lineage, and observability all converge here, in the fourth and final pillar: Governance. OpenMetadata supports fine-grained team structures with dedicated access control roles and policies, so that the visibility and trust built up by the other three pillars can actually be acted on.

Glossary Terms centralize definitions, terminology, and concepts so they stay consistent across the organization. Each glossary term has assigned owners and reviewers responsible for approving changes, meaning the shared vocabulary itself is subject to the same accountability model as the data.

OpenMetadata also supports data domains and data products, which group data assets, glossary terms, and teams around a shared business context (a finance domain, a marketing domain, and so on). Data products group assets that serve a common purpose, answering a particular business question, for instance. Like glossary terms, both domains and data products are assigned an owner, stewards, or experts, enabling decentralized governance: accountability sits with the people closest to the data, rather than being bottlenecked through a single central team.

Bringing It Together

None of these four pillars works in isolation. Discovery surfaces what exists; lineage traces where it came from and what depends on it; observability confirms it can still be trusted; and governance ties all of that into ownership, access control, and shared definitions that the whole organization can rely on. That's the real value of OpenMetadata: not four separate features, but a single connected path from raw metadata to organization-wide data governance.