Snowpark Connect History UI

The Spark Activity UI provides a view of Snowpark Connect for Spark usage and history across your Snowflake account. This UI consolidates the relevant information from each Snowpark Connect for Spark job or interactive session to aid in monitoring and troubleshooting.

Jobs run through the snowpark-submit CLI, EXECUTE CODE BUNDLE (with type: spark), and EXECUTE NOTEBOOK PROJECT (once the notebook initializes a Spark session) all appear under the Batch Jobs tab with their respective status, duration, name, and owner. Any Spark Session connection is tracked under the Spark Sessions tab with their respective queries and owner.

Spark Activity UI showing runs of `snowpark-submit` in the account.

The following sections cover how to open the page and how each tab behaves.

Open Spark Activity

  1. Sign in to Snowsight.
  2. Select Monitoring » Spark Activity.
Navigate to the Spark Activity UI by going to the Monitoring tab from the home page.

Batch Jobs Tab

The Batch Jobs tab lists batch-style Snowpark Connect for Spark executions for your account: invocations of the snowpark-submit utility, EXECUTE CODE BUNDLE runs whose specification sets type: spark, and EXECUTE NOTEBOOK PROJECT runs whose notebook code initializes a Spark session with snowflake.snowpark_connect.init_spark_session(). For each run you can see duration, who submitted the job, and status.

For more background on installing the CLI, command-line options, and batch patterns, see Orchestration with Snowpark Connect for Spark and the snowpark-submit command reference.

Batch Job Details Page

Select a row to open the details view for that run. The details include:

  • The full snowpark-submit command line that was executed
  • Queries associated with the job
  • Log output and stderr from the run
  • An OpenTelemetry trace for the job

For log and event-table behavior for submit workloads, see Monitoring Snowpark Connect for Spark workloads.

Note

The Logs section shows log records that were ingested into your event table, which is controlled by the LOG_LEVEL parameter. Set LOG_LEVEL to any level other than OFF and the Logs section shows records at that level and more severe — for example, INFO shows INFO, WARN, ERROR, and FATAL, while WARN shows WARN and more severe. Set LOG_LEVEL at the scope that matches your event table: at the account level if you use an account-level event table, or at the database level if the job’s database uses a database-level event table. If LOG_LEVEL is OFF, the Logs section is empty even when the run succeeds. For more information, see Setting levels for logging, metrics, and tracing.

The details page for a batch job shows the status, duration, logs, and queries.

Batch Job Trace Tab

The Trace tab on the Details page shows an OpenTelemetry trace of the job execution. Each DataFrame action is represented by a span, and the associated queries for each of those DataFrame actions is shown as a child span. This can help you more easily visualize the portions of your job that are taking the longest to run, and more easily get to the Query Profile for any slow or failed DataFrame actions.

For how trace data is represented and how to view or query it in your event table, see Viewing trace data.

The Trace tab shows the end to end trace of the job.

Spark Sessions Tab

The Spark Sessions tab lists Snowpark Connect for Spark client sessions that have connected to your account. That includes sessions from local laptops, Snowflake Notebooks, Snowflake Workspaces, and other supported clients.

For each session, Snowsight shows when the session started, who started it, the session ID, and the session name.

The Spark Sessions tab lists active and recent Spark sessions for the account.

Spark Session Details Page

The details page for a Spark Session shows all of the Queries issued from that session. Clicking the Query ID for one of these queries will take you to the Query Profile, which you can use to debug long or failed queries.

The Spark Session details page lists queries issued from the session.

Setting the App Name

The app name comes from the app_name argument when you call init_spark_session. If you set app_name, that value appears as the session name in Snowsight.

from snowflake.snowpark_connect import init_spark_session

init_spark_session(app_name="my_app_name")

In this example, my_app_name is shown as the session name.

If you do not pass app_name, Snowpark Connect uses the file name where the session was created as the app_name (for example main.py). For how the Python API derives the default application name, see the app_name parameter on init_spark_session.

Limitations

  • For a scheduled Snowflake Notebook run (EXECUTE NOTEBOOK PROJECT), the run only appears in the Batch Jobs tab once the notebook code calls snowflake.snowpark_connect.init_spark_session() to start a Spark session. If your notebook initializes the session late in its execution, there can be a lag between when the notebook actually starts and when Snowpark Connect for Spark registers it as a Spark job in Batch Jobs.
  • As of May 12, 2026 it may take up to 30 minutes for the Queries card to show the queries for a Batch Job.