Add job services to an app

This topic describes how to create and manage job services within a Snowflake Native App with Snowpark Container Services. For information on using services in an app, see Add job services to an app.

A Snowflake Native App with Snowpark Container Services can run a Snowpark Container Services job service.

A service created using CREATE SERVICE is long-running. An app must explicitly stop the service when it is no longer needed. In contrast, a job service created using EXECUTE JOB SERVICE is a service that terminates when the code of the service exits, similar to a stored procedure. When all containers exit, the job is done.

Job services run synchronously. The EXECUTE JOB SERVICE command completes after all containers exit.

Execute a job service in an app

To execute a job service in an app, add the EXECUTE JOB SERVICE command to the setup script.

The following example shows how to execute a job service in the context of a Snowflake Native App with Snowpark Container Services:

EXECUTE JOB SERVICE
  IN COMPUTE POOL consumer_compute_pool
  FROM SPECIFICATION_FILE = 'job_service.yml'
  NAME = 'services_schema.job_service'

GRANT MONITOR ON SERVICE services.job_service TO APPLICATION ROLE app_public;
Copy

Note

Note that the command parameters must be specified in the order shown in this example.

When called from the setup script, the EXECUTE JOB SERVICE command creates a job in a compute pool in the consumer account.

If the consumer creates the compute pool manually, they must grant the USAGE privilege on the compute pool to the app before this command will succeed. Therefore, providers must include logic in a stored procedure that tests if the correct privileges have been granted before running the EXECUTE JOB SERVICE.

The FROM SPECIFICATION_FILE = clause specifies the relative path to the service specification file on a stage. See Create the service specification file for more information.

The NAME = clause specifies the identifier for the job service. The name of this job service must be unique within the schema where it is located.

Note

Job services cannot be executed within a version schema.

The NAME = clause should use the schema and name of the job within the application. For , services_schema.job_service If the schema name is not specified the job service is created in the schema of the stored procedure or function that is executing the job service.

Monitor a job service in an app

To monitor the status of a job service within an app, use the SYSTEM$GET_SERVICE_STATUS — Deprecated command as shown in the following example:

CALL SYSTEM$GET_SERVICE_STATUS('schema.job_name')
Copy

This system function returns a JSON object that contains information about the specified job service within the app. Providers can call this system function from within the app to determine if the services has started or failed.

Consumers can also call this system function to determine the status of a service. This requires that providers grant the MONITOR privilege on the service an application role. See Execute a job service in an app for more information.

Accessing local container logs

To obtain the system logs for a job service within an app, use the SYSTEM$GET_SERVICE_LOGS system function as shown in the following example:

CALL SYSTEM$GET_SERVICE_LOGS('schema.job_name', 'instance_id', 'container_name'[, 10])
Copy

Providers can call this system function from within an app. In this context, the provider does not have to specify the app_name as part of the fully qualified job name.

Consumers can also run this system command. This requires that providers grant the MONITOR privilege on the service to an application role. See Execute a job service in an app for more information.