--- title: "Unity Catalog vs Hive Metastore: What Changes on Migration" url: "https://atlan.com/know/ai-agent/databricks/unity-catalog-vs-hive-metastore/" description: "Unity Catalog vs Hive metastore on Databricks: scope, namespace, permissions, storage, lineage, upgrade paths, and the context to inventory before cutover." author: "Emily Winks" author_role: "Data Governance Expert" published: "2026-11-05" updated: "2026-11-05T00:00:00.000Z" --- ## What is the difference between Unity Catalog and the Hive metastore? Unity Catalog is an account-level object that governs data and AI assets across Databricks workspaces with a catalog.schema.object namespace. The legacy Hive metastore is a workspace-level object that Databricks calls a legacy feature and is progressively disabling for direct access, with no removal date announced. Moving tables changes how they are registered and which group policies apply. The owners and business definitions your team recorded about them are the part to inventory before cutover. ## Key takeaways - Unity Catalog is account-level; the Hive metastore is a per-workspace store with progressive disablement underway - Deep CLONE does not migrate table history, and SYNC sets the default owner to the user who runs it - Federation keeps Hive tables running under Unity Catalog governance while you stage the cutover --- Unity Catalog is the Databricks account-level governance layer; the legacy Hive metastore is a per-workspace store whose direct access Databricks is progressively disabling, with no removal date announced. Atlan, the Context Layer for AI, reads each Databricks metastore through its own connection, so you can see what each one holds before and after cutover. ### Map Which Catalog Governs Each System Give it every system in your estate and which catalog governs each today. It returns a coverage table and the specific gap ranked by what breaks first. [Read the skill](/skills/catalog-coverage-gap-map.md). *Paste into a new chat* ``` Use the skill at https://atlan.com/skills/catalog-coverage-gap-map.md to map which catalog or control plane governs each system in our estate, including Databricks Unity Catalog and the legacy Hive metastore. Ask me for whatever it needs. ``` *Run once in a terminal* ``` curl -fsSL --create-dirs \ -o ~/.agents/skills/catalog-coverage-gap-map/SKILL.md \ https://atlan.com/skills/catalog-coverage-gap-map.md ``` *For an agent* ``` curl -fsSL https://atlan.com/skills/catalog-coverage-gap-map.md ``` Human Agent ClaudeChatGPTGeminiCursorOther Copy As of the October 23, 2026 Databricks page, a workspace that stays 30 days with the legacy metastore disabled becomes Unity Catalog-only for good. The question shifts from whether to migrate to when. The owners and definitions your team recorded about those tables are what a move can drop. The [context layer for Databricks](https://atlan.com/know/context-layer-for-databricks/) covers that layer, and the [context layer](https://atlan.com/know/what-is-context-layer/) page defines it in plain terms. | Dimension | Unity Catalog | Hive metastore (legacy, per workspace) | |---|---|---| | Scope | Spans workspaces | One workspace | | Namespace | `catalog.schema.object` | Two-level, shown as `hive_metastore` catalog | | Permission model | Account groups, `USE CATALOG`, `USE SCHEMA` | Workspace-local groups, table access controls | | Status | Recommended by Databricks | Legacy; direct access progressively disabled | --- ## What is the difference between Unity Catalog and the Hive metastore? Unity Catalog is an account-level object and the Hive metastore is a workspace-level object, according to the Databricks documentation "Work with the legacy Hive metastore alongside Unity Catalog" (September 11, 2026). Unity Catalog policies apply to account groups. Hive metastore policies apply to workspace-local groups. Every object registered in the legacy metastore appears in Unity Catalog interfaces under the `hive_metastore` catalog, per [Azure Databricks, "Database objects in the legacy Hive metastore"](https://learn.microsoft.com/en-us/azure/databricks/database-objects/hive-metastore) (May 26, 2026). Azure Databricks calls that hybrid period a useful model for transitioning. The Databricks per-workspace instance is separate from the open-source Apache Hive Metastore. Teams comparing [a data catalog and a context layer](https://atlan.com/know/data-catalog-vs-context-layer/) hit a similar confusion: where definitions are stored and the context around them are different jobs, which the [enterprise context layer](https://atlan.com/know/what-is-the-enterprise-context-layer/) is built to keep stable. --- ## What is Unity Catalog? Unity Catalog is an account-level object, and its data and AI assets follow the `catalog.schema.object` namespace. For managed tables it handles both governance and the file storage lifecycle, per Databricks "What is Unity Catalog?" (September 11, 2026). Managed tables are the preferred way to create tables, per Databricks "Unity Catalog managed tables" (October 30, 2026). External tables leave the storage lifecycle to you. Unity Catalog captures lineage automatically, down to column level, for queries on Databricks. The `system.access.table_lineage` and `system.access.column_lineage` system tables keep a rolling 395-day window by default, and Catalog Explorer retains lineage indefinitely for data captured after September 1, 2024, per Databricks "Lineage in Unity Catalog" (November 4, 2026). Business meaning sits one layer up. The [glossary and domains features](https://atlan.com/know/ai-agent/databricks/unity-catalog-business-glossary-domains-which-features-are-ga-preview/) carry GA and preview labels worth checking, and Databricks [metrics](https://atlan.com/know/ai-agent/databricks/unity-catalog-metrics/), the [Unity Catalog semantic layer](https://atlan.com/know/ai-agent/semantic-layer/unity-catalog-semantic-layer/), and [metric views](https://atlan.com/know/ai-agent/semantic-layer/how-to-create-databricks-metric-views/) have their own pages. Migration moves tables into this model, and the context your team attached to them needs a home that survives it. --- ## What is the legacy Hive metastore in Databricks? The legacy Hive metastore is the per-workspace metadata store descended from open-source Apache Hive. In the Apache Hive administration manual, "all the metadata for Hive tables and partitions are accessed through the Hive Metastore," persisted through an ORM layer in a backing relational database such as MySQL, PostgreSQL, Oracle, or MS SQL Server, with clients reaching the server over Thrift ([Hive Metastore Administration, Apache Hive](https://hive.apache.org/docs/latest/admin/adminmanual-metastore-administration/), December 12, 2024). The Databricks version does not use catalogs itself. Its default schema location is `/user/hive/warehouse/` on the DBFS root. Enabling a workspace for Unity Catalog does not reduce your ability to work with data already registered in the Hive metastore, per Azure Databricks. The metastore tracks where tables are. The [business context](https://atlan.com/know/business-context-layer/) behind them often lives in side documents, which is where a [context catalog](https://atlan.com/know/context-catalog/) earns its place. Who owns each table and what its columns mean is context to carry across on purpose. --- ## Is the Hive metastore deprecated in Databricks? The Hive metastore is legacy, with direct access being progressively disabled and no removal date announced. Databricks release notes use "deprecated" for credential passthrough and table access controls, not for the metastore as a whole, while Azure Databricks calls data governance using Hive metastore deprecated. | Date | What the docs say | Source | |---|---|---| | May 7, 2024 | Credential passthrough and Hive metastore table access controls deprecated on Runtime 15.0 | Databricks release note | | Mar 14, 2025 | Hive metastore federation generally available | Databricks release note | | Mar 31, 2025 | Disable your legacy Hive metastore, Public Preview | Databricks release note | | Sep 11, 2026 | "The per-workspace Hive metastore is a legacy feature"; migration recommended | "Work with the legacy Hive metastore alongside Unity Catalog" | | Oct 23, 2026 | Progressive disablement, Unity Catalog-only mode; new workspaces have no Hive metastore access | "Disable Hive metastore" | According to Databricks "Disable Hive metastore" (October 23, 2026), for workspaces with no Hive metastore activity legacy access "might be disabled automatically on your behalf starting in July 2026," and "after 30 days with the legacy metastore disabled, the workspace permanently becomes a Unity Catalog-only workspace." Per the same page, once legacy access is disabled, jobs that query Hive metastore tables fail, as do jobs on Runtime below 13.3. The release notes index through October 2026 carries no later entry for the disable setting, so the March 2025 release note remains its latest label. For the wider platform picture, read the [Data + AI Summit 2026 announcements](https://atlan.com/know/ai-agent/databricks/databricks-data-ai-summit-2026-announcements/) and the [key takeaways](https://atlan.com/know/databricks-summit-2026-key-takeaways/). --- ## Unity Catalog vs Hive metastore: head-to-head comparison Eight dimensions change when tables move from the Hive metastore to Unity Catalog, and the last column says what you do about each. | Dimension | Unity Catalog | Hive metastore | What changes for you | |---|---|---|---| | Scope | Account-level | Workspace-level | One policy set across workspaces | | Namespace | `catalog.schema.object` | `hive_metastore.schema.table` | Repoint jobs to three-level names | | Groups | Account groups | Workspace-local groups | Map local groups to account groups | | Privileges | `USE CATALOG`, `USE SCHEMA`; views need `SELECT` on referenced tables; no `DENY`, anything not granted is denied | View owner must own referenced objects | Rewrite `DENY` rules as grants; `ANY FILE`, `ANONYMOUS FUNCTION`, `READ_METADATA` have no Unity Catalog equivalent | | Default storage | Managed by Unity Catalog | DBFS root default location | Move production data off DBFS root | | Table drop | `UNDROP` for 7 days, permanent deletion within 30 | Managed drop deletes files immediately | Recovery window now exists | | Path access | Not allowed on managed tables | Allowed on managed tables | Rewrite path-based reads | | Lineage | Automatic, column-level; tables must be registered in a Unity Catalog metastore | Not captured by Unity Catalog until registered | Register tables to start capture | Illustrative example: a team has `hive_metastore.sales.sales_raw` and a view over it. After a deep clone the table is `main.sales.sales_raw`. The view must be recreated, and its access rule becomes `SELECT` on the table instead of ownership of every referenced object. Each of those changes is a place to test a dashboard, a job, or an agent. Where the tables feed analytics, the [Unity Catalog vs cross-platform semantic layer](https://atlan.com/know/ai-agent/semantic-layer/unity-catalog-vs-cross-platform-semantic-layer/) question follows the cutover. Where agents read them, [AI agent access control](https://atlan.com/know/ai-agent-access-control/) depends on the privilege mapping above being right. --- ## Which upgrade path moves which Hive tables to Unity Catalog? Databricks documents four paths: the upgrade wizard, `SYNC`, `CREATE TABLE CLONE`, and recreating views. UCX from Databricks Labs plans around all of them. | Path | Moves | Notes from the docs | |---|---|---| | Upgrade wizard (Catalog Explorer) | Schemas and tables, as external tables | Runs SYNC, leaves Hive tables intact, can be scheduled | | `SYNC` | External Hive tables to external Unity Catalog tables; managed tables stored outside workspace storage with `AS EXTERNAL` | `DRY RUN` checks first; default owner is the current user; DBFS root managed tables, views, temporary tables, and Hive SerDe tables go by other routes | | `CREATE TABLE CLONE` | Managed Delta tables to Unity Catalog managed tables | Deep clone required; table history not migrated; CTAS is the fallback | | Views | Recreated after referenced tables upgrade | Not moved automatically | | [UCX](https://github.com/databrickslabs/ucx) | Assessment, group migration, table migration workflows | Not formally supported with SLAs; provided AS-IS | Per Databricks "Upgrade Hive tables and views to Unity Catalog" (October 6, 2026), `SYNC` also writes table properties onto source tables unless a Runtime 12.2 LTS+ setting disables it. Per the Databricks `SYNC` reference (September 11, 2026), "the default owner is the current user." The person running the migration becomes the recorded owner of every upgraded table unless the optional `SET OWNER` clause names someone else. Cloned tables are treated as new tables, so you cannot run Delta Lake time travel on their pre-migration history. Automation helps with the mechanics. [Automated SQL lineage vs manual lineage mapping](https://atlan.com/know/ai-agent/data-for-ai/automated-sql-lineage-vs-manual-lineage-mapping/) covers the lineage side of that trade, the [MCP server for Databricks](https://atlan.com/know/mcp/mcp-server-for-databricks/) covers agent access afterward, and the [AI/BI Genie](https://atlan.com/know/databricks/databricks-ai-bi-genie/) experience inherits whatever owners and definitions the move leaves behind. --- ## Hive metastore federation vs full migration: when does each fit? Federation lets Unity Catalog govern Hive-registered tables without moving them, through a foreign catalog. Per Databricks "Hive metastore federation" (September 11, 2026), internal Databricks Hive metastores support reads and writes, external Hive metastores are read-only, and fallback mode on external locations is "intended only for use during migration." Federation went generally available on March 14, 2025. | Start with | When | |---|---| | Full migration | Tables are moved and jobs run on Runtime 13.3 or later, the documented requirements for disabling direct access | | Federation | Hive tables must keep running while governance moves first; you can federate before or after disabling direct access | | Both | Staged cutovers in a hybrid workspace, which Azure Databricks documents as a transition model | Federation buys time for the inventory in the next section, with the limits above: external Hive metastores stay read-only and fallback mode is for migration only. Once agents read the governed tables, the [Unity AI Gateway](https://atlan.com/know/ai-agent/databricks/unity-ai-gateway/) and the choice of a [neutral context layer](https://atlan.com/know/ai-agent/databricks/databricks-omnigent-vs-neutral-context-layer/) become part of the same plan. --- ## What context has to survive a Hive-to-Unity-Catalog move? Table definitions move, but the context people attached to the tables is the thing to inventory. Context continuity means the owners, definitions, dependents, and lineage that stay true after cutover. The upgrade commands handle some of it, such as `SET OWNER`, and the Databricks docs do not promise the rest. Three documented facts bound the problem. Unity Catalog lineage capture requires Unity Catalog registration. `CREATE TABLE CLONE` does not migrate table history. `SYNC` sets the default owner to the current user. Everything else is an inventory question. | What to record | Where it lives today | Check after cutover | |---|---|---| | Owner per table | Team wiki, tickets, catalog | Owner named on the Unity Catalog table, not "current user" | | Business definition per column | Glossary, docs | Definition resolves to the new three-level name | | Downstream dependents | Dashboards, jobs | Each job and view repointed, none on Runtime below 13.3 | | Lineage | Unity Catalog system tables from registration onward | `system.access.table_lineage` shows the upgraded table; earlier history kept elsewhere | | Privilege mapping | Workspace-local groups | Account groups, `USE CATALOG`, `USE SCHEMA` | Leaving the owner at the migrating user is the easiest miss, and the cheapest to catch before cutover. The [catalog coverage gap map skill](https://atlan.com/skills/catalog-coverage-gap-map.md) maps which catalog or control plane governs each system in an estate, at what depth, and ranks the gaps by what breaks first. Running it across your metastores shows where owner and definition coverage stops. [AI-ready data lineage](https://atlan.com/know/ai-readiness/ai-ready-data-lineage/) and [data lineage for AI](https://atlan.com/know/ai-agent/data-for-ai/data-lineage-for-ai/) explain why the lineage row matters to agents. [MCP for data lineage](https://atlan.com/know/mcp/mcp-for-data-lineage/) and [lineage root cause analysis with MCP](https://atlan.com/know/mcp/data-lineage-rca-with-mcp/) show what breaks when it goes missing. Who owns that context is a question for [data and AI teams together](https://atlan.com/know/context-layer-ownership-data-vs-ai-teams/), and definitions can live in a managed [KPI and business glossary](https://atlan.com/know/ai-agent/how-to-manage-kpi-definitions-and-a-business-glossary-in-a-data-catalog/) with shared [context layer terms](https://atlan.com/know/context-layer-glossary/). --- ## How Atlan approaches context continuity across Databricks metastores Atlan is the Context Layer for AI. In a metastore move its role is continuity of context: a place to see, side by side, what each Databricks metastore holds. Atlan's setup documentation states that one Databricks connection in Atlan corresponds to one metastore. For the legacy `hive_metastore` catalog, Atlan extracts metadata over JDBC. If your workspace uses Unity Catalog, the documentation directs you to the system tables extraction method ([How to set up Databricks in Atlan, Atlan Docs](https://docs.atlan.com/apps/connectors/data-warehouses/databricks/how-tos/set-up-databricks)). Each metastore is its own connection, so you can compare what each holds and agree what must be true on the other side. Downstream agents depend on owners and definitions that survive the move. The [Genie ontology and Atlan context layer](https://atlan.com/know/ai-agent/databricks/genie-ontology-and-atlan-context-layer/) page, the [Genie context requirements](https://atlan.com/know/ai-agent/databricks/databricks-genie-context-requirements/), the steps to [add context to a Genie space](https://atlan.com/know/ai-agent/databricks/add-context-to-databricks-genie-space/), and [Agent Bricks](https://atlan.com/know/ai-agent/databricks/agent-bricks/) cover where that context gets used. The build sequence is in [how to implement an enterprise context layer for AI](https://atlan.com/know/how-to-implement-enterprise-context-layer-for-ai/). --- ## Plan the cutover around what people know about the tables The metastore swap is a SQL job and a context job. Status is legacy, direct access is being progressively disabled, and no removal date is announced, so plan for when, not whether. The owners and definitions move when someone writes them down, checks them, and confirms they landed. --- ## FAQs about Unity Catalog vs Hive metastore ### 1. Is the Hive metastore deprecated in Databricks? Databricks calls the per-workspace Hive metastore a legacy feature and is progressively disabling direct access to it for workspaces with no Hive metastore activity, with no removal date announced. The word deprecated applies to credential passthrough and Hive metastore table access controls on Databricks Runtime 15.0, and Azure Databricks states that data governance using Hive metastore is deprecated. ### 2. Does Unity Catalog replace the Hive metastore? Unity Catalog is the account-level governance layer Databricks recommends. Objects registered in the legacy metastore still surface inside Unity Catalog under the `hive_metastore` catalog, so the two coexist during a hybrid period that Azure Databricks calls a useful model for transitioning. ### 3. Can you keep Hive metastore tables and still govern them from Unity Catalog? Yes. Hive metastore federation, generally available since March 14, 2025, lets Unity Catalog govern tables registered in a Hive metastore through a foreign catalog. Internal Databricks Hive metastores support reads and writes; external Hive metastores are read-only. ### 4. What happens to permissions when you move to Unity Catalog? Unity Catalog policies apply to account groups, while Hive metastore policies apply to workspace-local groups. Object operations need `USE CATALOG` and `USE SCHEMA`, views need `SELECT` on referenced tables, and `DENY`, `ANY FILE`, `ANONYMOUS FUNCTION`, and `READ_METADATA` need mapping before cutover. ### 5. Does table history move when you clone a Hive table to Unity Catalog? No. Databricks documentation states that table history is not migrated when you run `CREATE TABLE CLONE`. Deep clones are required when cloning from the legacy Hive metastore, and CTAS is the documented fallback. ### 6. What should you inventory before Hive metastore access is disabled? Record the owner and business definition of each table, its downstream dependents, its lineage, and the groups behind its privileges. Atlan reads each Databricks metastore through its own connection, one connection per metastore, so you can compare what each one holds. --- ## Sources 1. Disable Hive metastore, Databricks on AWS documentation (October 23, 2026). 2. Work with the legacy Hive metastore alongside Unity Catalog, Databricks documentation (September 11, 2026). 3. Upgrade Hive tables and views to Unity Catalog, Databricks documentation (October 6, 2026). 4. SYNC, Databricks SQL language reference (September 11, 2026). 5. Hive metastore federation, Databricks documentation (September 11, 2026). 6. Lineage in Unity Catalog, Databricks documentation (November 4, 2026). 7. What is Unity Catalog?, Databricks documentation (September 11, 2026). 8. UCX, Databricks documentation (October 6, 2026). 9. Credential passthrough and Hive metastore table access controls are deprecated, Databricks release note (May 7, 2024). 10. Hive metastore federation is generally available, Databricks release note (March 14, 2025). 11. Disable your legacy Hive metastore (Public Preview), Databricks release note (March 31, 2025). 12. [Database objects in the legacy Hive metastore, Microsoft Learn](https://learn.microsoft.com/en-us/azure/databricks/database-objects/hive-metastore) 13. [Hive Metastore Administration, Apache Hive](https://hive.apache.org/docs/latest/admin/adminmanual-metastore-administration/) 14. [UCX, Databricks Labs on GitHub](https://github.com/databrickslabs/ucx) 15. [How to set up Databricks in Atlan, Atlan Docs](https://docs.atlan.com/apps/connectors/data-warehouses/databricks/how-tos/set-up-databricks) ## Next steps - [See how Atlan's context layer works in recorded demos](https://atlan.com/context-layer-demo/) - [Book a demo or talk to Atlan's sales team](https://atlan.com/forms/talk-to-sales-contact/)