Release 481-e STS (15 Jun 2026)#

Starburst Enterprise platform (SEP) 481-e is the follow up SEP release to 480-e. It includes all improvements from the following Trino releases:

Trino 481

This release is a short term support (STS) release.

Breaking changes#

Legacy Hadoop-based object storage support has been removed from the Delta Lake, Hive, and Iceberg connectors. Any catalog using fs.hadoop.enabled=true for cloud object storage or the legacy hive.s3.*, hive.azure.*, or hive.gcs.* configuration properties fail to start after upgrading. fs.hadoop.enabled now only activates HDFS support. You must migrate to native file system support before upgrading:
- S3: set fs.native-s3.enabled=true and rename hive.s3.* properties
- Azure: set fs.native-azure.enabled=true and rename hive.azure.* properties
- GCS: set fs.native-gcs.enabled=true and rename hive.gcs.* properties
See the migration guides for full property mapping tables.
The internal-communication.shared-secret and node.environment configuration properties are no longer supported in connector configuration files when managed statistics is enabled. Configurations containing these properties fail to initialize on startup, regardless of whether managed statistics is enabled. Remove these properties from your configurations before upgrading.
Geospatial functions now reject WKT input containing invalid geometry that was previously accepted and silently corrected. Queries passing non-conforming WKT strings now fail with a parse error. Validate and correct any WKT inputs in your queries before upgrading.
ST_Union() now returns an empty geometry collection instead of NULL when called with no input rows. Update any queries that check for a NULL result from ST_Union() on empty inputs.
ST_Union() no longer inserts additional vertices at intersection points when unioning a point with a line. Queries that depended on those extra intersection vertices in the output geometry produce different results after upgrading.
Removed the ai.agent.persona-directory-path configuration property and support for defining custom personas in JSON files. Remove the property from your coordinator configuration file or the cluster fails to start. You can now create and manage custom personas in the UI.
The Trino S3 file system implementation (io.trino.hdfs.s3.TrinoS3FileSystem) was removed in Trino 481. If you use Ranger Audit with SEP, you must update your file system to one supported directly by Hadoop. Reference Hadoop documentation for more information. For one example, if you are using Ranger Audit to S3, update the following properties in your Ranger Audit configuration:

Ranger Audit to S3 properties#
Former property	Former value	New property	New value
`xasecure.audit.destination.hdfs.config.fs.s3.impl`	`io.trino.hdfs.s3.TrinoS3FileSystem`	`xasecure.audit.destination.hdfs.config.fs.s3a.impl`	`org.apache.hadoop.fs.s3a.S3AFileSystem`
`xasecure.audit.destination.hdfs.config.fs.AbstractFileSystem.s3.impl`	`io.trino.hdfs.s3.TrinoS3FileSystem`	`xasecure.audit.destination.hdfs.config.fs.AbstractFileSystem.s3a.impl`	`org.apache.hadoop.fs.s3a.S3AFileSystem`
`xasecure.audit.destination.hdfs.config.trino.s3.endpoint`	Unchanged	`xasecure.audit.destination.hdfs.config.fs.s3a.endpoint`	Unchanged
`xasecure.audit.destination.hdfs.config.trino.s3.access-key`	Unchanged	`xasecure.audit.destination.hdfs.config.fs.s3a.access.key`	Unchanged
`xasecure.audit.destination.hdfs.config.trino.s3.secret-key`	Unchanged	`xasecure.audit.destination.hdfs.config.fs.s3a.secret.key`	Unchanged
`xasecure.audit.destination.hdfs.config.trino.s3.path-style-access`	Unchanged	`xasecure.audit.destination.hdfs.config.fs.s3a.path.style.access`	Unchanged

Additionally, xasecure.audit.destination.hdfs.dir must have the s3a:/// scheme instead of s3:///.

General#

Added support for skipping header rows with the skip_header parameter when using the starburst.io.load table function.
Added support for scheduling SQL jobs as frequently as every minute. Previously, the minimum interval between job runs was 15 minutes.
Added the jobs.work.sql-job.history.retain-max-attempts-per-job configuration property to limit the number of history entries retained per SQL job. The default is 500, with a range of 1–2000.
Added support for configuring the Azure connection-pool idle time and HTTP request timeout via the azure.connection-pool-max-idle-time and azure.http-request-timeout configuration properties.
Fixed query failure when casting char values to numeric, boolean, varbinary, and temporal types, including when querying Hive views that contain such casts.
Fixed an issue in the catalog explorer where the column description field appeared empty in edit mode after a page refresh.
Fixed an issue in the catalog explorer where long column descriptions overflowed their containers.

Security#

Fixed Vault AppRole pull mode to re-authenticate on every secret read, ensuring token_num_uses and secret_id_num_uses constraints are honored.
Fixed a vulnerability in the cache service’s redirections API that could allow unauthorized data access.

Starburst AI#

Added support for Guardrails, a set of cluster-wide protections for AIDA.
Added support for creating and managing custom persona prompts through the UI.
Fixed an issue that caused prompts to OSS OpenAI models on Amazon Bedrock to return no results.

BigQuery connector#

Extended bigquery.skip-view-materialization configuration property support to include TIMESTAMP and BIGNUMERIC types.

ClickHouse connector#

Added support for reading ClickHouse Tuple columns as Trino ROW type, including nested tuples and nullable elements.
Added support for reading ClickHouse Array columns as Trino ARRAY type.

Delta Lake connector#

Added support for Everpure as a backend object storage.

Hive connector#

Added support for Everpure as a backend object storage.
Added support for configurable HTTP client timeouts for the Glue metastore, including connection TTL, maximum idle time, socket connect timeout, and socket read/write timeout. See AWS Glue catalog configuration properties for details.
Fixed an issue where some Hive queries on Apache Ozone File System (OFS) failed with a NullPointerException when zero-byte files were present in the table directory.
Fixed an issue where materialized view creation metadata did not use the configured Hive Metastore catalog namespace, preventing dropped materialized views from being recreated with the same name.

Iceberg connector#

Added support for writing to external tables when using Unity Catalog with Iceberg.
Added support for server-side scan planning when using the Databricks Unity Catalog as a metastore.
Added support for Everpure as a backend object storage.
Added support for configurable HTTP client timeouts for the Glue metastore, including connection TTL, maximum idle time, socket connect timeout, and socket read/write timeout. See AWS Glue catalog configuration properties for details.
Added support for prefixed-path storage credentials in the Iceberg REST catalog.

Kafka connector#

Added optional kafka.authentication.scope configuration property for OAuth 2.0 authentication. Use this property to specify an OAuth2 scope for finer-grained authorization control when authenticating with Kafka brokers.
Added OAuth2 authentication support for Confluent Schema Registry. Set kafka.confluent-schema-registry.authentication.type=OAUTH2 and configure the required token endpoint URL, client ID, and client secret properties. HTTPS is required.

MongoDB connector#

Added the mongodb.sampling-count and mongodb.sampling-order configuration properties for controlling document sampling during schema inference.
Added the update_schema table procedure for re-sampling documents and updating the inferred table schema.

Snowflake connector#

Added support for pushdown of IN list predicates on VARCHAR columns.
Added support for pushdown of LIKE predicates, the lower() and upper() functions, and coalesce() predicates on VARCHAR columns when the snowflake.experimental-pushdown.enabled configuration property or the corresponding experimental_pushdown_enabled session property is set to true.
Fixed correctness and performance issues pushing down predicates to Snowflake columns using non-default collation.
Fixed an issue where querying view-only schemas such as ACCOUNT_USAGE or ORGANIZATION_USAGE in the Snowflake SNOWFLAKE shared database returned empty results or failed when snowflake.database-prefix-for-schema.enabled=true was configured.
Fixed an issue where query annotations configured with query.comment-format were not visible in Snowflake query history when executing queries through the Snowflake parallel connector.

Redshift connector#

Added read-only support for Redshift spectrum external tables and table redirection to configured Hive or Iceberg catalogs.

Teradata connector#

Added support for multiple native table operator versions coexisting in a single Teradata database. The CREATE FUNCTION statement now uses version-specific function and symbol names, allowing a single Teradata instance to serve multiple SEP clusters running different versions. See Native table operator for updated installation instructions.