February 18-21, 2025 — 9.3 Release Notes¶

Tasks now support lower scheduling intervals — General availability¶

Tasks now support lower scheduling executions as frequent as every 10 seconds. (The previous minimum scheduling interval was 1 minute.)

Data lineage — General availability¶

Data lineage automatically tracks the flow of data between Snowflake objects in real-time, for example, from a table to a view. You can track the lineage of table-like objects, columns, and stages as well as the lineage between data objects and machine learning objects like datasets, feature views, and models. You can use lineage information to assist in impact analysis, monitoring, troubleshooting, and compliance efforts. Lineage can also help you propagate knowledge of sensitive data elements using tags.

Lineage information is available through Snowsight, SQL, and Python. For more information, see:

• Data Lineage

• GET_LINEAGE SQL function

• Snowpark ML lineage API

SEARCH function: Support for conjunctive semantics¶

With this release, the SEARCH function supports conjunctive (AND) semantics. Before this release, the function only supported disjunctive (OR) semantics. To specify the semantics for a full-text search using the function, set the new SEARCH_MODE argument to ‘AND’ or ‘OR’.

When you specify ‘AND’ for the SEARCH_MODE argument, there is a match if the tokens extracted from at least one of the columns or fields being searched match all of the tokens extracted from the search string. The matching tokens must all be in one column or field; they can’t be spread across multiple columns or fields.

Support for a wildcard character in network rule network identifiers — General availability¶

With this release, support for using an asterisk as a wildcard character when specifying a network identifier for a network rule in its VALUE_LIST parameter is generally available.

For more information, see CREATE NETWORK RULE.

Support for telemetry metrics and custom spans, with visualizations in Snowsight — General availability¶

With this release, support for collecting metrics data, implementing custom spans, and viewing log, trace, and metric data in Snowsight is generally available. Metrics data provides signs of resource consumption by using CPU and memory metrics. With custom spans, you can get finer-grained tracing within the handler for a procedure or function. Snowsight provides visualizations for collected data with which you can analyze and optimize your code. This telemetry can be found under the Monitoring tab from the Snowsight home page.

For more information, see Adding custom spans to a trace, Collecting metrics data, Viewing log messages, and Viewing trace data.

For an introduction to observability in Snowflake, see Observability in Snowflake apps.

Dynamic tables: Support for UNION ALL¶

Dynamic tables now support the following use cases for UNION ALL for both full and incremental refresh:

• UNION ALL of a table and itself or a clone of itself.

• UNION ALL between a GROUP BY or DISTINCT and another GROUP BY or DISTINCT.

Dynamic tables: Support for multiple window functions with different PARTITION BY clauses¶

Dynamic tables now support multiple window functions in the same dynamic table definition, where PARTITION BY clauses are non-identical. PARTITION BY clauses that appear in multiple sequential SELECT blocks within a query are not supported.

Cloning support for Snowflake-managed Apache Iceberg™ tables — General availability¶

With this release, support for cloning Snowflake-managed Iceberg tables is available.

For more information, see CREATE <object> … CLONE and Cloning and Apache Iceberg™ tables.