Monitor account activity with system tables

This article explains the concept of system tables in Databricks and highlights resources you can use to get the most out of your system tables data.

What are system tables?

System tables are a Databricks-hosted analytical store of your account’s operational data found in the system catalog. System tables can be used for historical observability across your account.

Note

For documentation on system.information_schema, see Information schema.

Requirements

To access system tables, your workspace must be enabled for Unity Catalog. For more information, see Enable system table schemas.

For system tables GCP project numbers and identities, see Private Service Connect (PSC) attachment URIs and project numbers.

Which system tables are available?

Currently, Databricks hosts the following system tables:

Table	Description	Location	Supports streaming	Free retention period	Includes global or regional data
Audit logs (Public Preview)	Includes records for all audit events across your Databricks account. For a list of available audit events, see Audit log reference.	`system.access.audit`	Yes	365 days	Regional for workspace-level events. Global for account-level events.
Table lineage (Public Preview)	Includes a record for each read or write event on a Unity Catalog table or path.	`system.access.table_lineage`	Yes	365 days	Regional
Column lineage (Public Preview)	Includes a record for each read or write event on a Unity Catalog column (but does not include events that do not have a source).	`system.access.column_lineage`	Yes	365 days	Regional
Billable usage	Includes records for all billable usage across your account. Each usage record is an hourly aggregate of a resource’s billable usage.	`system.billing.usage`	Yes	365 days	Global
Pricing	A historical log of SKU pricing. A record gets added each time there is a change to a SKU price.	`system.billing.list_prices`	No	N/A	Global
Clusters (Public Preview)	A slow-changing dimension table that contains the full history of compute configurations over time for any cluster.	`system.compute.clusters`	Yes	None	Regional
Node timeline (Public Preview)	Captures the utilization metrics of your all-purpose and jobs compute resources.	`system.compute.node_timeline`	Yes	30 days	Regional
Node types (Public Preview)	Captures the currently available node types with their basic hardware information.	`system.compute.node_types`	No	N/A	Regional
SQL warehouses (Public Preview)	Contains the full history of configurations over time for any SQL warehouse.	`system.compute.warehouses`	Yes	365 days	Regional
SQL warehouse events (Public Preview)	Captures events related to SQL warehouses. For example, starting, stopping, running, scaling up and down.	`system.compute.warehouse_events`	Yes	365 days	Regional
Jobs	Tracks all jobs created in the account.	`system.lakeflow.jobs`	Yes	365 days	Regional
Job tasks (Public Preview)	Tracks all job tasks that run in the account.	`system.lakeflow.job_tasks`	Yes	365 days	Regional
Job run timeline (Public Preview)	Tracks the start and end times of job runs.	`system.lakeflow.` `job_run_timeline`	Yes	365 days	Regional
Job task timeline (Public Preview)	Tracks the start and end times and compute resources used for job run tasks.	`system.lakeflow.` `job_task_run_timeline`	Yes	365 days	Regional
Marketplace funnel events (Public Preview)	Includes consumer impression and funnel data for your listings.	`system.marketplace.listing_` `funnel_events`	Yes	365 days	Regional
Marketplace listing access (Public Preview)	Includes consumer info for completed request data or get data events on your listings.	`system.marketplace.listing_` `access_events`	Yes	365 days	Regional
Databricks Assistant events	(Public Preview) Tracks user messages sent to the Databricks Assistant.	`system.access.assistant_events`	No	365 days	Regional
Query history (Public Preview)	Captures records for all queries run on SQL warehouses.	`system.query.history`	Yes	90 days	Regional
Model serving endpoint usage (Public Preview)	Captures token counts for each request to a model serving endpoint and its responses. To capture the endpoint usage in this table, you must enable usage tracking on your serving endpoint.	`system.serving.endpoint_usage`	Yes	90 days	Regional
Model serving endpoint data (Public Preview)	A slow changing dimension table that stores metadata for each served external model in a model serving endpoint.	`system.serving.served_entities`	Yes	365 days	Regional

The billable usage and pricing tables are free to use. Tables in Public Preview are also free to use during the preview but could incur a charge in the future.

Note

You may see other system tables in your account besides the ones listed above. Those tables are in Private Preview currently and are empty by default. If you are interested in using any of these tables, please reach out to your Databricks account team.

Enable system table schemas

Since system tables are governed by Unity Catalog, you need to have at least one Unity Catalog-enabled workspace in your account to enable and access system tables. System tables include data from all workspaces in your account but they can only be accessed from a Unity Catalog-enabled workspace.

System tables are enabled at the schema level. If you enable a system schema, you enable all the tables within that schema. When new schemas are released, an account admin needs to manually enable the schema.

System tables must be enabled by an account admin. You can enable system tables using the SystemSchemas API.

Note

The billing schema is enabled by default. Other schemas must be enabled manually.

Use the List system schemas API to view which schemas are available and enabled in your account. The schemas will have one of the following states:
- state: AVAILABLE: The system schema is available but has not yet been enabled.
- state: EnableCompleted: You have enabled the system schema and it is visible in Catalog Explorer.
Use the Enable system schema API to add the schema to your system catalog.
Use the Disable system schema API to remove a schema from your system catalog.

Grant access to system tables

Access to system tables is governed by Unity Catalog. No user has access to these system schemas by default. To grant access, a user that is both a metastore admin and an account admin must grant USE and SELECT permissions on the system schemas. See Manage privileges in Unity Catalog.

System tables are read-only and cannot be modified.

Note

If your account was created after March 6, 2024, you might not have a metastore admin by default. For more information, see Set up and manage Unity Catalog.

Do system tables contain data for all workspaces in your account?

System tables contain operational data for all workspaces in your account deployed within the same cloud region. Billing system tables contain account-wide data.

Even though system tables can only be accessed through a Unity Catalog workspace, the tables also include operational data for the non-Unity Catalog workspaces in your account.

Where is system table data stored?

Your account’s system table data is stored in a Databricks-hosted storage account located in the same region as your metastore. The data is securely shared with you using Delta Sharing.

Each table has a free data retention period. For information on extending the retention period, contact your Databricks account team.

Where are system tables located in Catalog Explorer?

The system tables in your account are located in a catalog called system, which is included in every Unity Catalog metastore. In the system catalog you’ll see schemas such as access and billing that contain the system tables.

Considerations for streaming system tables

Databricks uses Delta Sharing to share system table data with customers. Be aware of the following considerations when streaming with Delta Sharing:

If you are using streaming with system tables, set the skipChangeCommits option to true. This ensures the streaming job is not disrupted from deletes in the system tables. See Ignore updates and deletes.
Trigger.AvailableNow is not supported with Delta Sharing streaming. It will be converted to Trigger.Once.

If you use a trigger in your streaming job and find it isn’t catching up to the latest system table version, Databricks recommends increasing the scheduled frequency of the job.

Read incremental changes from streaming system tables

spark.readStream.option("skipChangeCommits", "true").table("system.billing.usage")

Known issues

Currently no support for real-time monitoring. Data is updated throughout the day. If you don’t see a log for a recent event, check back later.