> For the complete documentation index, see [llms.txt](https://docs.seemoredata.io/external-docs/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.seemoredata.io/external-docs/fundamentals/getting-set-up/setting-integrations/databricks/create-databricks-service-principal-and-privileges.md).

# Create Databricks Service Principal and Privileges

Create the service principal, SQL warehouse and grants that SeeMore Data needs to read your Databricks account's system tables.

This page walks you through creating the Databricks service principal and granting the privileges that SeeMore Data needs to read your account's system tables. For what each permission allows and what SeeMore Data reads with it, see [Permissions and Data Handling](https://docs.seemoredata.io/external-docs/fundamentals/getting-set-up/setting-integrations/databricks/permissions-and-data-handling.md) — it's the page to share with your security reviewer.

{% hint style="success" %}
**Prefer to run a script?** [Automated Setup](https://docs.seemoredata.io/external-docs/fundamentals/getting-set-up/setting-integrations/databricks/automated-setup.md) does these steps with one command per region, then verifies the result.
{% endhint %}

### What you'll set up <a href="#overview" id="overview"></a>

SeeMore Data needs a service principal, and access at three levels of your Databricks account: the account, each region and each workspace. Each level has its own admin, and one person can hold all of these roles.

| Level                 | Steps                                                                                                                                              | Who                                                          | How often                                     | What it gives SeeMore Data                                                                                                                                                                                                                       |
| --------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------ | --------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ |
| **Service principal** | [1. Create the service principal](#create-service-principal)                                                                                       | Workspace admin or account admin                             | Once                                          | Signing in. Required                                                                                                                                                                                                                             |
| **Account**           | [2. Grant account access](#account-access)                                                                                                         | Account admin                                                | Once                                          | Query text, which is optional. A [metastore admin](#metastore-admin) for each region, if one is missing                                                                                                                                          |
| **Region**            | [3. Prepare the home workspace](#home-workspace), [4. Enable the system schemas](#enable-system-schemas), [5. Grant privileges](#grant-privileges) | Metastore admin, and a workspace admin of the home workspace | Once in each region SeeMore Data should fetch | The region's query history, compute, jobs, lineage, AI activity and asset catalog. In the connected region, the region of the workspace you enter in SeeMore Data, also billing, prices and the workspace list. Required in the connected region |
| **Workspace**         | [6. Give access in each workspace](#workspace-access)                                                                                              | Workspace admin                                              | In each workspace you want these features for | Live queries, users, warehouse automation and AI Services endpoint details. Optional                                                                                                                                                             |

Then [allow SeeMore Data's IP addresses](#allowlist-ips) if your workspaces restrict inbound access, and [verify the setup](#verify).

At the end you'll have three values to enter in SeeMore Data — the **workspace URL**, the **client ID** and the **secret** — and a SQL warehouse in each region, which you choose by name during setup.

### Before you start

* Your workspaces must be attached to a **Unity Catalog metastore**. To check, run `SELECT current_metastore();` in a SQL editor — a workspace with no metastore returns an error.
* An account in several regions has **one metastore per region**. System schemas, grants and catalogs belong to a metastore, so steps 3 to 5 are done once for each region you want fetched. A region you don't set up still counts toward your total spend; only its query history, compute, jobs, lineage and AI detail are missing.

You enter one workspace in SeeMore Data. It discovers every other workspace in the account, and the region each one is in. A workspace added later is picked up on the next sync: its cost right away, and its query history and compute if it's in a region SeeMore Data fetches.

## Service principal

### 1. Create the service principal <a href="#create-service-principal" id="create-service-principal"></a>

You create the service principal once. Every region and workspace uses the same one.

A workspace admin or an account admin can create the service principal and its OAuth secret:

* **Workspace admin**: in the workspace, go to **Settings** → **Identity and access** → **Service principals** → **Manage** → **Add service principal** → **Add new**. Then select the service principal, open its **Secrets** tab and generate a secret.
* **Account admin**: in the **account console**, go to **User management** → **Service principals** → **Add service principal**. Then select the service principal, open its **Credentials & secrets** tab and generate an OAuth secret.

A service principal added from a workspace's settings is also added to that workspace. Either way, the secret signs in to every workspace the service principal is added to, in any region.

Under **Scopes**, search for and select these seven: `sql`, `query-history`, `access-management`, `scim`, `model-serving`, `vector-search` and `unity-catalog`. They are the only Databricks APIs SeeMore Data calls, and a token issued from the secret can't reach any other. [Permissions and Data Handling](https://docs.seemoredata.io/external-docs/fundamentals/getting-set-up/setting-integrations/databricks/permissions-and-data-handling.md#oauth-scopes) lists what each one is for. **All APIs** also works, but Databricks doesn't recommend it.

{% hint style="info" %}
**On Azure**, adding a service principal asks you to choose **Databricks managed** or **Microsoft Entra ID managed**. Both work. For **Microsoft Entra ID managed**, paste the application (client) ID of your Microsoft Entra app registration. Either way, generate the secret in Databricks, as described above. SeeMore Data signs in only with a Databricks OAuth secret, so a client secret created for the app registration in the Azure portal is rejected.
{% endhint %}

Keep two values:

* **Client ID** — also shown as the application ID. You'll use it in the grants below and in the integration form.
* **Secret** — used only in the integration form.

{% hint style="danger" %} <mark style="color:$danger;">**The secret is displayed once.**</mark> Copy it before leaving the page and store it securely. If it's lost, generate a new secret rather than a new service principal.
{% endhint %}

{% hint style="warning" %}
**OAuth secrets expire.** Pick the longest lifetime your policy allows, and record the expiry date. When the secret expires, SeeMore Data stops reading until you generate a new one, with the same scopes, and update the integration.
{% endhint %}

To do this through the API, see [Databricks' OAuth guide for service principals](https://docs.databricks.com/aws/en/dev-tools/auth/oauth-m2m). The secrets call takes a `lifetime` parameter, a duration in seconds such as `63072000s` (two years), and a `scopes` list with the seven names above.

## Account

### 2. Grant account access <a href="#account-access" id="account-access"></a>

An account admin does this step once, in the **account console**.

#### Query text access <a href="#query-text-access" id="query-text-access"></a>

Databricks returns query text as `<REDACTED>` to everyone outside the account-level group **`databricks_pii_access`**. SeeMore Data needs the text for several features: similar-query grouping, query tagging and lineage, and the SQL it shows for each query, including live queries in [Real Time Monitoring](https://docs.seemoredata.io/external-docs/fundamentals/our-features/real-time-monitoring.md). Without it, cost, timing and volume still arrive, but every query has the same placeholder text.

An account admin adds the service principal to that group in the **account console** under **User management** → **Groups**. The group doesn't exist by default. Create it there with exactly that name — it's case-sensitive — and not in workspace admin settings, where the creator gets **Manage** on the group.

{% hint style="warning" %}
This shows SeeMore Data the full text of every query run in the account, and SeeMore Data stores it as it is. If your policy doesn't allow that, skip this step: cost, timing and volume still work, and the features above see placeholder text. Error messages and customer-managed keys are covered in [Query text and error messages](https://docs.seemoredata.io/external-docs/fundamentals/getting-set-up/setting-integrations/databricks/permissions-and-data-handling.md#query-text).
{% endhint %}

#### A metastore admin in each region <a href="#metastore-admin" id="metastore-admin"></a>

Step 5 needs a metastore admin in each region you set up. A metastore that Databricks created automatically starts without one. If a region's metastore has none, an account admin assigns one in the account console.

## Each region

Do steps 3 to 5 once in each region SeeMore Data should fetch. Start with the region of the workspace you'll enter in SeeMore Data, then repeat them in each other region.

### 3. Prepare the home workspace <a href="#home-workspace" id="home-workspace"></a>

In each region, choose one workspace to hold SeeMore Data's SQL warehouse: the region's **home workspace**. In the region you start with, use the workspace you'll enter in SeeMore Data, because the connection test signs in there. In each other region, SeeMore Data signs in to the home workspace to learn the region and list the SQL warehouses you can choose from.

#### Add the service principal to the workspace <a href="#add-to-workspace" id="add-to-workspace"></a>

A service principal created in the account console isn't a member of any workspace, and one created from a workspace's settings is a member of that workspace only. In the workspace, go to **Settings** → **Identity and access** → **Service principals**, and add it if it isn't listed. Do this in the home workspace now. Other workspaces follow the same steps, in [step 6](#workspace-access).

Then open it there and check that it has the **Workspace access** and **Databricks SQL access** entitlements, and add any that are missing. Some workspaces give them to every new member and some give none. Without them, the workspace refuses SeeMore Data's calls, including the connection test. It doesn't need **Allow cluster creation** or **Allow instance pool creation**, so don't grant those.

#### Create the Seemore SQL warehouse <a href="#create-warehouse" id="create-warehouse"></a>

SeeMore Data reads through a SQL warehouse, and a region's system tables can only be read through a warehouse in that region. In the home workspace, go to **SQL Warehouses** → **Create SQL warehouse** and name it **`seemore_warehouse`**. A dedicated warehouse keeps SeeMore Data's reads off your teams' warehouses, and makes its own cost easy to see.

| Setting          | Suggested                       | Why                                                                                                |
| ---------------- | ------------------------------- | -------------------------------------------------------------------------------------------------- |
| **Cluster size** | `2X-Small`                      | The queries read system tables only                                                                |
| **Type**         | Serverless where available      | It starts in seconds, so auto-stop saves money without a cold-start wait. Pro and Classic work too |
| **Auto stop**    | On, at the lowest value offered | It only needs to run while SeeMore Data is reading                                                 |
| **Scaling**      | Min 1, max 1                    | Reads are sequential                                                                               |

The warehouse runs in your account, so its DBUs appear on your bill — a small, intermittent cost. If your policy doesn't allow a new warehouse, an existing one works: give the service principal **Can use** on it instead. You don't need its ID; during setup, you choose it by name.

#### Give the service principal Can use on it <a href="#grant-warehouse-access" id="grant-warehouse-access"></a>

Give the service principal **Can use** on `seemore_warehouse`. This is how SeeMore Data runs its reads. During setup, the SQL warehouse list shows only the warehouses the service principal has a permission on, and a check through one without **Can use** fails with *"Give the service principal CAN USE on this SQL warehouse, or choose another one."*

Warehouse and endpoint permissions aren't Unity Catalog securables, so there's no `GRANT` statement for them. Set them on each object's **Permissions** tab, with the Databricks CLI as shown below, or through `PATCH /api/2.0/permissions/warehouses/{warehouse_id}` and `PATCH /api/2.0/permissions/serving-endpoints/{serving_endpoint_id}`.

{% code overflow="wrap" %}

```bash
databricks permissions update warehouses <warehouse_id> --json '{
  "access_control_list": [
    {
      "service_principal_name": "<service_principal_client_id>",
      "permission_level": "CAN_USE"
    }
  ]
}'
```

{% endcode %}

`update` patches the existing list, so the warehouse's owner and admin entries survive.

### 4. Enable the system schemas <a href="#enable-system-schemas" id="enable-system-schemas"></a>

SeeMore Data reads nine system schemas: `access`, `billing`, `compute`, `lakeflow`, `query`, `tags`, `serving`, `ai_gateway` and `mlflow`. They're enabled per metastore, so repeat this step in a workspace of each region SeeMore Data should fetch. An **account admin or metastore admin** must make these calls; anyone else gets `403`.

If SeeMore Data's setup shows a required table's schema as *not found*, the schema is already enabled, because Databricks enables the required ones. The cause is a missing `USE CATALOG` on `system` or `USE SCHEMA` on that schema, as granted in [step 5](#grant-privileges).

First, get your metastore ID:

```sql
SELECT current_metastore() AS metastore_id;
```

It returns `<cloud>:<region>:<metastore_id>`, for example `aws:us-west-2:1a2b3c4d-0000-1111-2222-333344445555`. Use the part after the last colon. Then list the state of each schema, and enable each one that's `AVAILABLE`:

```
GET /api/2.1/unity-catalog/metastores/{metastore_id}/systemschemas
PUT /api/2.1/unity-catalog/metastores/{metastore_id}/systemschemas/{schema}
```

| State                           | What to do                                                                                         |
| ------------------------------- | -------------------------------------------------------------------------------------------------- |
| `AVAILABLE`                     | Enable it with the `PUT`                                                                           |
| `ENABLE_COMPLETED` or `MANAGED` | Nothing — it's already readable                                                                    |
| `UNAVAILABLE`                   | Not offered to your account yet. Contact Databricks, unless it's one of the optional schemas below |

If the `PUT` returns *"\<schema> system schema can only be enabled by Databricks"*, Databricks manages that schema and it's already available.

`tags`, `serving`, `ai_gateway` and `mlflow` are optional: Databricks offers them per region, and `tags` is still in preview. Without them, only governed tags, per-request AI detail, the external-provider spend estimate and MLflow run activity are missing. What Databricks bills you for AI still comes from `system.billing`.

{% hint style="warning" %}
**The list can leave out schemas you already have.** On many metastores `tags`, `ai_gateway` and `mlflow` are missing from the list even though Databricks has already enabled them. Before you skip any of the nine, check the `system` catalog itself:

```sql
SHOW SCHEMAS IN system;
```

Grant every one of the nine that appears there. Skip a schema in the grants below only if it's missing from both the list and `SHOW SCHEMAS`.
{% endhint %}

### 5. Grant privileges <a href="#grant-privileges" id="grant-privileges"></a>

Run the following as a **metastore admin**, in a SQL editor attached to `seemore_warehouse`. Replace `<service_principal_client_id>` with the client ID from step 1.

Grants are made per metastore, and a grant in one region doesn't reach another. Run the grants in a workspace of each region SeeMore Data should fetch. When a check fails during setup, its card links here and says which region to run them in.

{% code overflow="wrap" lineNumbers="true" expandable="true" %}

```sql
-- Required before any schema below is reachable.
GRANT USE CATALOG ON CATALOG system TO `<service_principal_client_id>`;

-- Workspace inventory: resolves the account and the workspaces in it.
GRANT USE SCHEMA ON SCHEMA system.access  TO `<service_principal_client_id>`;
GRANT SELECT     ON SCHEMA system.access  TO `<service_principal_client_id>`;

-- Usage and list prices: the cost model reads these.
GRANT USE SCHEMA ON SCHEMA system.billing TO `<service_principal_client_id>`;
GRANT SELECT     ON SCHEMA system.billing TO `<service_principal_client_id>`;

-- Warehouses, clusters and events: what your compute was configured as,
-- when it was running, and how hard it was working — including the per-node
-- CPU and memory behind job utilization.
GRANT USE SCHEMA ON SCHEMA system.compute TO `<service_principal_client_id>`;
GRANT SELECT     ON SCHEMA system.compute TO `<service_principal_client_id>`;

-- Jobs and pipelines: what your scheduled work is, when each run happened,
-- and which task within it did the spending.
GRANT USE SCHEMA ON SCHEMA system.lakeflow TO `<service_principal_client_id>`;
GRANT SELECT     ON SCHEMA system.lakeflow TO `<service_principal_client_id>`;

-- Query history: per-query attribution and queue pressure.
GRANT USE SCHEMA ON SCHEMA system.query   TO `<service_principal_client_id>`;
GRANT SELECT     ON SCHEMA system.query   TO `<service_principal_client_id>`;

-- Governed tags: the tag keys your account governs, so spend can be grouped by them.
-- Skip these two only if SHOW SCHEMAS IN system doesn't list `tags` (step 4).
GRANT USE SCHEMA ON SCHEMA system.tags    TO `<service_principal_client_id>`;
GRANT SELECT     ON SCHEMA system.tags    TO `<service_principal_client_id>`;

-- Model serving telemetry: which model answered each request, for whom, and at what
-- token count. Skip these two only if SHOW SCHEMAS IN system doesn't list `serving` (step 4).
GRANT USE SCHEMA ON SCHEMA system.serving TO `<service_principal_client_id>`;
GRANT SELECT     ON SCHEMA system.serving TO `<service_principal_client_id>`;

-- Unity Gateway telemetry: every request through a model service, and Databricks' estimate
-- of external-provider spend. Skip these two only if SHOW SCHEMAS IN system doesn't list `ai_gateway` (step 4).
GRANT USE SCHEMA ON SCHEMA system.ai_gateway TO `<service_principal_client_id>`;
GRANT SELECT     ON SCHEMA system.ai_gateway TO `<service_principal_client_id>`;

-- MLflow activity: experiments, runs, their parameters, tags and metrics.
-- Skip these two only if SHOW SCHEMAS IN system doesn't list `mlflow` (step 4).
GRANT USE SCHEMA ON SCHEMA system.mlflow TO `<service_principal_client_id>`;
GRANT SELECT     ON SCHEMA system.mlflow TO `<service_principal_client_id>`;
```

{% endcode %}

{% hint style="warning" %}
`USE CATALOG` on `system` is not optional. Without it, a `SELECT` grant alone silently produces `TABLE_OR_VIEW_NOT_FOUND`, which looks like the system schemas were never enabled rather than a missing grant.
{% endhint %}

Grant at **schema level**, not per table. Databricks adds tables to the system schemas over time, and a per-table list goes stale without any error. If your policy requires table-level grants, [Tables read](https://docs.seemoredata.io/external-docs/fundamentals/getting-set-up/setting-integrations/databricks/permissions-and-data-handling.md#optional-tables) lists every table SeeMore Data reads.

#### Browse on your own catalogs — for the asset catalog <a href="#browse-grant" id="browse-grant"></a>

SeeMore Data's **asset catalog** — the inventory of your catalogs, schemas and tables that lineage and unused-table insights are built on — reads `system.information_schema`. It returns only the objects the reader may see, so a service principal with no access to a catalog gets **no rows for it and no error**.

`BROWSE` is the privilege to grant. It works at catalog level with no `USE CATALOG` or `USE SCHEMA` beside it, and it is metadata only: a principal holding `BROWSE` and nothing else can't run `SELECT` on your data. `USE CATALOG` is not a substitute; it shows the catalog but none of its tables.

{% hint style="info" %}
**`BROWSE` exposes more than names**, including column names and types and the SQL of your views. SeeMore Data reads names, owners, comments, timestamps, tags and view SQL — not the columns. See [what `BROWSE` exposes](https://docs.seemoredata.io/external-docs/fundamentals/getting-set-up/setting-integrations/databricks/permissions-and-data-handling.md#browse) before you grant it.
{% endhint %}

Unity Catalog has no wildcard for this, so list your catalogs and repeat the grant once per catalog:

{% code overflow="wrap" lineNumbers="true" %}

```sql
SHOW CATALOGS;

GRANT BROWSE ON CATALOG `<catalog_name>` TO `<service_principal_client_id>`;
```

{% endcode %}

A catalog belongs to one metastore, so repeat this in a workspace of each region SeeMore Data fetches. Skip any catalog you'd rather SeeMore Data didn't inventory. Nothing else depends on this grant — costing, query history and warehouse automation all work without it.

## Each workspace

### 6. Give access in each workspace <a href="#workspace-access" id="workspace-access"></a>

[Real Time Monitoring](https://docs.seemoredata.io/external-docs/fundamentals/our-features/real-time-monitoring.md), warehouse automation and settings changes, the user list and AI Services endpoint details call each workspace's own API, so they need the service principal there. Do this step in each workspace you want them for, including the home workspaces. Skip the workspaces you don't.

Query history doesn't need this step: after a daily import has found the workspace, its query history is read hourly through the SQL warehouse of its region. A workspace the service principal isn't a member of still has its usage counted, because billing is account-scoped. If the workspace is in a region SeeMore Data fetches, its query history, compute, jobs and lineage are imported too, through that region's SQL warehouse.

#### Add the service principal to the workspace

Add it under **Settings** → **Identity and access** → **Service principals**, with the same entitlements as in [step 3](#add-to-workspace). A home workspace already has it.

The permissions below are set like **Can use**: on each object's **Permissions** tab, with the Databricks CLI or through the permissions API, as described in [step 3](#grant-warehouse-access).

#### Can monitor on the warehouses you want to watch live — for Real Time Monitoring

Give the service principal **Can monitor** on each SQL warehouse you want in [Real Time Monitoring](https://docs.seemoredata.io/external-docs/fundamentals/our-features/real-time-monitoring.md). The live view reads each warehouse directly and sees only the ones the service principal is permitted on. **Without this grant a warehouse is missing from the live view, with no error**; its cost and query history still arrive.

**Can use** is not enough here: it shows the warehouse but not other users' queries, so a busy warehouse appears with zero running queries. Databricks also lets **Can monitor** start the warehouse and run queries on it. SeeMore Data does neither.

{% code overflow="wrap" %}

```bash
databricks permissions update warehouses <warehouse_id> --json '{
  "access_control_list": [
    {
      "service_principal_name": "<service_principal_client_id>",
      "permission_level": "CAN_MONITOR"
    }
  ]
}'
```

{% endcode %}

A warehouse created later needs its own grant. To avoid revisiting this, give a group **Can monitor** on new warehouses through your warehouse policy, and add the service principal to the group once.

#### Can manage on the warehouses you want SeeMore Data to change — optional <a href="#grant-can-manage" id="grant-can-manage"></a>

Skip this unless you want to change SQL warehouse settings from SeeMore Data. Give the service principal **Can manage** only on those warehouses.

A user with the [Editor](https://docs.seemoredata.io/external-docs/fundamentals/settings/user-roles.md) role or higher can change a warehouse's size, minimum and maximum clusters, and auto stop on its **Configurations** tab. Auto stop can be as low as 1 minute on Serverless and 10 minutes on Pro and Classic. You can't turn auto stop off from SeeMore Data.

With **Can manage**, SeeMore Data changes a warehouse only when an Editor saves a change. It doesn't start, stop or delete warehouses, or change their permissions. A new size can restart a running warehouse, so SeeMore Data asks you to confirm before it applies one.

In Databricks, **Can manage** also allows starting, stopping, editing and deleting the warehouse, and changing its permissions. Grant it only where you need it. It includes **Can monitor**, so a warehouse you grant it on needs no **Can monitor** grant.

Without the grant, the **Edit** button on the warehouse's **Configurations** tab is disabled, with a link to the setup guide. To add the grant, open **SQL Warehouses** → the warehouse → **Permissions** in Databricks, add the service principal and give it **Can manage**. Or use the CLI:

{% code overflow="wrap" %}

```bash
databricks permissions update warehouses <warehouse_id> --json '{
  "access_control_list": [
    {
      "service_principal_name": "<service_principal_client_id>",
      "permission_level": "CAN_MANAGE"
    }
  ]
}'
```

{% endcode %}

#### Can view on your model serving endpoints — for AI Services <a href="#grant-serving-endpoint-access" id="grant-serving-endpoint-access"></a>

Give the service principal **Can view** on each model serving endpoint you run, so AI Services can show its served model, compute size, scale-to-zero setting and state. **Can view** can't query or change the endpoint. Without it, the endpoint's cost and requests still appear, but its configuration shows as **Hidden — needs CAN\_VIEW**.

{% code overflow="wrap" %}

```bash
databricks serving-endpoints update-permissions <serving_endpoint_id> --json '{
  "access_control_list": [
    {
      "service_principal_name": "<service_principal_client_id>",
      "permission_level": "CAN_VIEW"
    }
  ]
}'
```

{% endcode %}

The command takes the endpoint's ID: `databricks serving-endpoints get <endpoint_name>` prints it. An endpoint created later needs its own grant. The pay-per-token `databricks-*` Foundation Model endpoints need none.

## Finish

### 7. Allow SeeMore Data's IP addresses <a href="#allowlist-ips" id="allowlist-ips"></a>

Skip this step unless your workspaces restrict inbound access. If they do, SeeMore Data's connection is refused until its addresses are allowed. SeeMore Data connects from two fixed egress addresses, the same for every SeeMore Data connection:

```
3.78.110.132
3.73.171.134
```

Allow them in each control you use. Databricks enforces both together, so a request must pass each:

* **Context-based ingress**, configured for the account: an account admin allows the two addresses in the policy that covers your workspaces.
* **Workspace IP access lists**, which need the Enterprise tier: a workspace admin adds the addresses to each workspace that has a list, with the Databricks CLI or the IP Access Lists API on that workspace's own URL. The account-level IP access list API controls the account console only, not your workspaces.

{% code overflow="wrap" %}

```bash
# Lists apply only when this returns true.
databricks workspace-conf get-status enableIpAccessLists

databricks ip-access-lists create --json '{
  "label": "seemore",
  "list_type": "ALLOW",
  "ip_addresses": ["3.78.110.132", "3.73.171.134"]
}'
```

{% endcode %}

{% hint style="warning" %}
If a workspace had no allow list before, adding one blocks every address that isn't on a list. Check that your other allow lists still cover your users first.
{% endhint %}

### 8. Verify the setup <a href="#verify" id="verify"></a>

When you select **Check** in SeeMore Data's setup, it runs these same checks as the service principal. Running them here first is a quick way to catch a missing grant, with Databricks' own error message.

Confirm the grants landed. Run this in each region you set up:

{% code overflow="wrap" lineNumbers="true" %}

```sql
SHOW GRANTS `<service_principal_client_id>` ON CATALOG system;
SHOW GRANTS `<service_principal_client_id>` ON SCHEMA system.access;
SHOW GRANTS `<service_principal_client_id>` ON SCHEMA system.billing;
SHOW GRANTS `<service_principal_client_id>` ON SCHEMA system.compute;
SHOW GRANTS `<service_principal_client_id>` ON SCHEMA system.lakeflow;
SHOW GRANTS `<service_principal_client_id>` ON SCHEMA system.query;
SHOW GRANTS `<service_principal_client_id>` ON SCHEMA system.tags;
SHOW GRANTS `<service_principal_client_id>` ON SCHEMA system.serving;
SHOW GRANTS `<service_principal_client_id>` ON SCHEMA system.ai_gateway;
SHOW GRANTS `<service_principal_client_id>` ON SCHEMA system.mlflow;
```

{% endcode %}

Then run the five required checks. The first three are account-wide, so setup checks them once, through the connected region's SQL warehouse. The last two are checked in every region SeeMore Data fetches:

{% code overflow="wrap" lineNumbers="true" %}

```sql
-- Account: checked once, through the connected region's SQL warehouse.
SELECT 1 AS access_ok       FROM system.access.workspaces_latest LIMIT 1;
SELECT 1 AS billing_use_ok  FROM system.billing.usage            LIMIT 1;
SELECT 1 AS billing_rate_ok FROM system.billing.list_prices      LIMIT 1;

-- Region: checked through each region's own SQL warehouse.
SELECT 1 AS query_ok        FROM system.query.history            LIMIT 1;
SELECT 1 AS compute_ok      FROM system.compute.warehouse_events LIMIT 1;
```

{% endcode %}

{% hint style="warning" %}
Run these **as the service principal**, not as yourself. A metastore admin can read these tables whether or not the grants worked, so running them as an admin proves nothing.
{% endhint %}

{% hint style="danger" %} <mark style="color:$danger;">**All five must return a row.**</mark> The two billing tables are separate grants — `usage` is consumption and `list_prices` is the rate behind it, and granting one does not grant the other. **Finish setup** stays disabled until all five pass in the connected region, and another region can't be fetched until its two regional checks pass.
{% endhint %}

The check also reads 23 optional tables. If it can't read one, the region shows **Ready with gaps** and a **Missing** line, and setup isn't blocked. They're listed by feature in [Tables read](https://docs.seemoredata.io/external-docs/fundamentals/getting-set-up/setting-integrations/databricks/permissions-and-data-handling.md#optional-tables).

If you granted `BROWSE`, check the asset catalog in each region. This is the query SeeMore Data's asset fetch runs, so what it returns is what will be imported:

{% code overflow="wrap" lineNumbers="true" %}

```sql
SELECT catalog_name
FROM system.information_schema.catalogs
WHERE catalog_name NOT IN ('system', 'samples', '__databricks_internal')
ORDER BY catalog_name;
```

{% endcode %}

Every catalog you granted `BROWSE` on should be listed. The default `workspace` catalog is usually visible to all users without a grant, so seeing it listed doesn't prove your grants landed. An empty result doesn't block setup; SeeMore Data then imports no assets.

### Next step

{% content-ref url="/pages/tavYfi9vb01RuziE5oFG" %}
[Integration with Seemore](https://docs.seemoredata.io/external-docs/fundamentals/getting-set-up/setting-integrations/databricks/integration-with-seemore.md)
{% endcontent-ref %}
