BerriAI/litellm · error · HTTPException

Database not connected

Error message

Database not connected

What it means

POST /user/bulk_update requires a database; with prisma_client None it returns HTTP 500 {'error': 'Database not connected'} before touching any user. The endpoint otherwise applies user_updates (optionally across all_users) and enforces admin-only role changes with the same 'Only proxy admins can modify user roles.' error.

Source

Thrown at litellm/proxy/management_endpoints/internal_user_endpoints.py:1735

    
    Example request for all users:
    ```bash
    curl --location 'http://0.0.0.0:4000/user/bulk_update' \
    --header 'Authorization: Bearer sk-1234' \
    --header 'Content-Type: application/json' \
    --data '{
        "all_users": true,
        "user_updates": {
            "user_role": "internal_user",
            "max_budget": 50.0
        }
    }'
    ```
    """
    from litellm.proxy.proxy_server import litellm_proxy_admin_name, prisma_client

    if prisma_client is None:
        raise HTTPException(
            status_code=500,
            detail={"error": "Database not connected"},
        )

    # Only proxy admins can modify user_role in bulk updates
    _bulk_role = getattr(data.user_updates, "user_role", None) if data.user_updates else None
    if _bulk_role is None and data.users:
        _bulk_role = next((u.user_role for u in data.users if u.user_role is not None), None)
    if _bulk_role is not None and user_api_key_dict.user_role != LitellmUserRoles.PROXY_ADMIN.value:
        raise HTTPException(
            status_code=403,
            detail="Only proxy admins can modify user roles.",
        )

    # Determine the list of users to update
    users_to_update: list[UpdateUserRequest] | list[UpdateUserRequestNoUserIDorEmail] = []

    if data.all_users and data.user_updates:

View on GitHub (pinned to 77b7c6c40c)

Solutions

  1. Set general_settings.database_url (postgresql://...) or DATABASE_URL and restart the proxy
  2. Confirm the DB is reachable via GET /health/liveliness or GET /user/list
  3. Re-run the bulk job after the health check shows the database connected

Example fix

# before
POST /user/bulk_update -d '{"all_users": true, "user_updates": {"max_budget": 50}}'
# 500 Database not connected

# after: config.yaml
general_settings:
  database_url: postgresql://user:pass@host:5432/litellm
# restart proxy, retry -> 200
Defensive patterns

Strategy: validation

Validate before calling

import requests

def db_ready(base_url: str, admin_key: str) -> bool:
    r = requests.get(f"{base_url}/health/liveliness",
                     headers={"Authorization": f"Bearer {admin_key}"}, timeout=10)
    return r.ok and r.json().get("litellm_database", "") != ""

Try / catch

except requests.HTTPError as e:
    body = e.response.text if e.response is not None else ""
    if e.response is not None and e.response.status_code == 500 and "Database not connected" in body:
        raise ConfigError("bulk user updates need DATABASE_URL configured") from e
    raise

Prevention

When it happens

Trigger: POST /user/bulk_update (including the all_users: true form shown in its docstring) on a proxy started without general_settings.database_url / DATABASE_URL, or whose Prisma connection failed.

Common situations: Bulk onboarding scripts pointed at a proxy instance without Postgres config; the DATABASE_URL env var missing in the cron job's or container's environment; config.yaml only containing model_list.

Related errors

AI-assisted analysis of BerriAI/litellm@77b7c6c40c (2026-08-18). Data as JSON: /api/errors/2c6ceca64a038fed. Report an issue: GitHub.