CREATE ENGINE

Creates an engine, which serves as a compute cluster for running queries and processing data.

Syntax

CREATE ENGINE [IF NOT EXISTS] <engine_name>
[WITH
    [AUTO_START = <true/false>]
    [AUTO_STOP = <minutes>]
    [AUTO_VACUUM = <on/off>]
    [AUTO_VACUUM_WAIT_ON_STOP = <true/false>]
    [DEFAULT_DATABASE = <database_name>]
    [INITIALLY_STOPPED = <true/false>]
    [START_IMMEDIATELY = <true/false>]
    [MIN_CLUSTERS = <clusters>]
    [MAX_CLUSTERS = <clusters>]
    [NODES = <nodes>]
    [TYPE = <type>]
    [FAMILY = <family>]
    [RELEASE = (CHANNEL = <channel)]
    [QUERY_CACHE_MEMORY_FRACTION = <fraction>]
    [VECTOR_INDEX_CACHE_MEMORY_FRACTION = <fraction>]]

Options

Parameter	Description
`<engine_name>`	The name of the engine to be created.
`AUTO_START = <true/false>`	When `true`, sending a query to a stopped engine will start the engine before processing the query. If not specified, `true` will be used as default.
`AUTO_STOP = <minutes>`	The amount of idle time (in minutes) after which the engine automatically stops. Setting the minutes to `0` indicates that `AUTO_STOP` is disabled. If not specified, `20` is used as default.
`AUTO_VACUUM = <on/off>`	When `on`, the engine will automatically assess the health of tables’ data layout after processing a DML operation, and will run the `VACUUM` command when necessary to maintain the underlying table health.
`AUTO_VACUUM_WAIT_ON_STOP = <true/false>`	When `false`, stopping the engine (explicitly or via auto-stop) does not wait for auto-vacuum tasks to complete. Default: `true`.
`DEFAULT_DATABASE = <database_name>`	The database an engine will attempt to use by default when dealing with queries that require a database. If not specified, `NULL` is used as default.
`INITIALLY_STOPPED = <true/false>`	When `false`, the newly created engine will be started as part of the `CREATE ENGINE` command. Cannot be used with `START_IMMEDIATELY`. If not specified, `false` is used as default.
`START_IMMEDIATELY = <true/false>`	When `true`, the newly created engine will be started as part of the `CREATE ENGINE` command. Cannot be used with `INITIALLY_STOPPED`. If not specified, `true` is used as default.
`CLUSTERS = <clusters>` (deprecated)	Specifies the number of clusters in an engine. Each cluster is a group of nodes, and all clusters within an engine are identical in terms of node type and number of nodes. If not specified, `1` is used as default.
`MIN_CLUSTERS = <clusters>`	Specifies a minimum number of clusters in an engine. If `MIN_CLUSTERS` is different from `MAX_CLUSTERS`, the engine will automatically change the number of clusters depending on load. If not specified, `1` is used as default.
`MAX_CLUSTERS = <clusters>`	Specifies a maximum number of clusters in an engine. If `MIN_CLUSTERS` is different from `MAX_CLUSTERS`, the engine will automatically change the number of clusters depending on load. If not specified, `1` is used as default.
`NODES = <nodes>`	Indicates the number of nodes in each cluster within an engine. This number can range from `1` to `128`. If not specified, `2` is used as default.
`TYPE =<type>`	Defines the type of node used in the engine. Options include `S`, `M`, `L`, or `XL` If not specified, `M` is used as default.
`FAMILY =<family>`	Defines the family of node used in the engine. Options include `STORAGE_OPTIMIZED` (alias `SO`) or `COMPUTE_OPTIMIZED` (alias `CO`) If not specified, `STORAGE_OPTIMIZED` is used as default.
`CHANNEL = <channel>`	Specifies the release channel to be used. Options are `DEFAULT` to use the default release process or `PREVIEW` to get upgrades earlier.
`QUERY_CACHE_MEMORY_FRACTION = <fraction>`	Sets the fraction of the engines memory that can be used to cache query (sub)results. `VECTOR_INDEX_CACHE_MEMORY_FRACTION` + `QUERY_CACHE_MEMORY_FRACTION` can be at most 0.8.
`VECTOR_INDEX_CACHE_MEMORY_FRACTION = <fraction>`	Sets the fraction of the engines memory that can be used to cache vector indexes. See Vector Search Indexes for details. `VECTOR_INDEX_CACHE_MEMORY_FRACTION` + `QUERY_CACHE_MEMORY_FRACTION` can be at most 0.8.

Limitations: For small and medium instance types:

Each engine is limited to a maximum of two clusters
Each cluster can have up to ten nodes
The total number of nodes across all clusters cannot exceed 15
When scaling a running engine either vertically or horizontally, new queries will be directed to the new cluster. Queries running on the original clusters will continue until completion. The clusters will wait up to 24 hours for these queries to finish, after which any unfinished queries may be stopped.

For large and extra-large instance types:

Each engine is limited to a maximum of one node and one cluster

If you would like to remove the following limitations, reach out to Firebolt support at support@firebolt.io to do the following:

Use more than ten nodes per cluster.
Use more than 15 nodes across all clusters.
Use more than one large or extra-large node per engine.

Examples

Create a basic engine

The following example creates an engine with one cluster, using node type ‘M’ from the storage-optimized family and 2 nodes per cluster :

CREATE ENGINE my_engine;

Create an engine with multiple nodes

The following example creates an engine with one cluster, using node type ‘S’ of Compute Optimized family and 5 nodes per cluster :

CREATE ENGINE my_engine
WITH TYPE="S"  FAMILY="CO" NODES = 5 MIN_CLUSTERS = 1 MAX_CLUSTERS = 1;

Create an engine with delayed start

The following example creates an engine with one cluster, using node type ‘M’ and 3 nodes per cluster. The engine will not be automatically started after creation because INITIALLY_STOPPED is set to true.

CREATE ENGINE my_engine
WITH TYPE="M" NODES=3 INITIALLY_STOPPED=true;

Create an engine with auto-stop

The following example creates an engine with one cluster, using node type ‘L’ in the compute-optimized family with 2 nodes per cluster. The engine will be automatically stopped after 10 minutes of idle time because AUTO_STOP is set to 10.

CREATE ENGINE my_engine
WITH TYPE="L" FAMILY="COMPUTE_OPTIMIZED" NODES=2 AUTO_STOP=10;

Create an engine with concurrency auto-scaling

The following example creates an engine that automatic concurrency scaling. The engine adjusts the number of clusters based on workload demand, maintaining between one and two clusters.

CREATE ENGINE my_engine
WITH MIN_CLUSTERS=1 MAX_CLUSTERS=2;

Create an engine in the preview release channel

The following example creates an engine and assigns it to the preview release channel. For more information, see release settings.

CREATE ENGINE myengine
WITH TYPE="S" RELEASE=(CHANNEL=PREVIEW);

Overview

Performance and Observability

Security

Self-Managed

Managed service

Guides

SQL reference

Release notes

API reference

Legal

Syntax

Options

Examples

Create a basic engine

Create an engine with multiple nodes

Create an engine with delayed start

Create an engine with auto-stop

Create an engine with concurrency auto-scaling

Create an engine in the preview release channel

​Syntax

​Options

​Examples

​Create a basic engine

​Create an engine with multiple nodes

​Create an engine with delayed start

​Create an engine with auto-stop

​Create an engine with concurrency auto-scaling

​Create an engine in the preview release channel

Syntax

Options

Examples

Create a basic engine

Create an engine with multiple nodes

Create an engine with delayed start

Create an engine with auto-stop

Create an engine with concurrency auto-scaling

Create an engine in the preview release channel