Configuration parameters
Applies to: Databricks SQL
A configuration parameter is a setting which affects the behavior of Databricks SQL outside of the specified SQL syntax.
The effective value of a configuration parameter is derived from the different levels where it is set.
Configuration parameter scopes
System
The system value is the default value used by Databricks SQL if there is no override.
Global
An administrator can override the system default value for a parameter using the Configure SQL parameters or SQL Warehouse API. This value is then the initial value seen by any newly established session moving forward.
Session
A user can override a configuration parameter temporarily within a session scope by using the SET statement. The setting remains in effect for the duration of the session unless the user overrides is again with a SET statement or RESET.
Supported configuration parameters
The following is a list of the supported configuration parameters that you can set in Databricks SQL, their system defaults, and at which levels they can be set. Databricks SQL does not support setting Spark configurations that aren't listed.
Parameter name | Description | System default | Globally settable | Session settable |
---|---|---|---|---|
ANSI_MODE | Allows you to disable strict ANSI SQL behavior for certain functions and casting rules. | TRUE | Yes | Yes |
LEGACY_TIME_PARSER_POLICY | Controls how date and timestamps are parsed and formatted. | EXCEPTION | Yes | Yes |
MAX_FILE_PARTITION_BYTES | The maximum number of bytes to pack into a single partition when reading from file based sources. | 128m | No | Yes |
READ_ONLY_EXTERNAL_METASTORE | Controls whether an external metastore is treated as read-only. | FALSE | Yes | No |
STATEMENT_TIMEOUT | Sets a SQL statement timeout in seconds | 172800 seconds | Yes | Yes |
TIMEZONE | Sets the local time zone. | UTC | Yes | Yes |
USE_CACHED_RESULT | Controls whether Databricks SQL caches and reuses results whenever possible. | TRUE | No | Yes |
The ANSI_MODE
system default value is FALSE
for accounts added before Databricks SQL 2022.35.