Hive Metastore

Overview

The Hive Metastore is the original table catalogue of the Hadoop era, and still the most widely integrated one. It is a central repository of metadata for tables and partitions, reached through the metastore service API, and clients including Hive, Impala and Spark have used it for well over a decade.

It stores databases, tables, partitions and functions. Partitions are first-class, which is the clearest difference from Iceberg-native catalogues, where partitioning is a property of table metadata instead.

Since Hive 3.0 it has been packaged so it can run without the rest of Hive, specifically so other systems can adopt it, and from 4.1.0 it ships as a standalone component in binary and Docker form.

It has also modernised towards Iceberg. Version 4.1.0 added a REST catalogue server backed by the metastore, and 4.2.0 added an Iceberg REST catalogue client, so it can both serve and consume the same protocol the newer catalogues speak. Iceberg features such as compaction, deletion vectors and Z-ordering have arrived alongside.

Two caveats matter for planning. In standalone mode the compactor for ACID tables cannot run, and replication is untested outside Hive. And the older release lines are all end of life: 1.x, 2.x and 3.x were declared EOL during 2024, leaving 4.x as the only supported line. The project publishes no statement for or against using it in new systems; what it does publish is active maintenance, with 4.2.1 released in August 2026 as a security fix.

Fine-grained authorisation is delegated: the metastore does database and table-level authorisation and leaves policy to Apache Ranger, with Apache Atlas for governance metadata.

Key features and capabilities

The same headings are used for every table & technical catalogues entry, so two tools can be read side by side.

What it catalogues

• Databases, tables, partitions and functions
• Partitions are first-class, unlike in Iceberg-native catalogues
• Catalog properties support added in 4.2.0
• No volumes, models or other AI asset types

Table formats

• Classic Hive formats through SerDes: text, ORC, Parquet, Avro
• Apache Iceberg actively supported: compaction, deletion vectors, Z-ordering, column defaults
• ACID transactional Hive tables are read-write even standalone
• Delta Lake and Hudi register into it from their own side rather than being Hive features

Engines that can use it

• Hive, Impala and Spark through the metastore service API
• Hive's own execution runs on Tez from 4.x
• Iceberg-native engines can connect over REST instead of Thrift
• Widely consumed by others: AWS Glue, Apache Polaris and Unity Catalog all interoperate with it

Access control and auditing

• Database-level and table-level authorisation
• REST endpoint authentication: OAuth 2, JWT, simple or none; Kerberos elsewhere
• Fine-grained policy is delegated to Apache Ranger
• Row and column security and credential vending are not metastore features

Interoperability

• The Thrift metastore API, the long-standing integration point for the ecosystem
• Serves an Iceberg REST catalogue from 4.1.0, enabled by setting a servlet port
• Acts as an Iceberg REST client from 4.2.0
• Documented interoperability with Apache Gravitino and Apache Polaris

Table maintenance and operations

• Standalone limitation: the ACID compactor cannot run without Hive
• Replication is untested outside Hive
• Iceberg table compaction, and auto compaction from 4.2.0
• Schema tool and metatool for maintenance; scalability depends on the backing database

How it runs

• Self-hosted only; no managed edition from the foundation
• Backing database: SQL Server, MySQL, MariaDB, Oracle or PostgreSQL over JDBC
• Standalone metastore ships as binaries and Docker images from 4.1.0
• Only the 4.x line is supported; 1.x, 2.x and 3.x reached end of life in 2024

Pricing

Open source

Free under the Apache 2.0 licence. No managed or commercial edition is published by the project.

Vendor pricing page →

About The Apache Software Foundation

The Hive Metastore is a module of Apache Hive, a top-level Apache Software Foundation project licensed under Apache 2.0, with Hive's inception given as 2010 and HCatalog merged into it in 2013. The metastore was split out as a separately releasable package in Hive 3.0 so that non-Hive systems could adopt it. It is maintained by the Hive community on an active cadence, with 4.0.0 in March 2024 and 4.2.1 in August 2026. There is no managed edition from the foundation, though many vendors host it.

hive.apache.org