What causes "Postgres DB Not connected" in BerriAI/litellm?

Set DATABASE_URL to a PostgreSQL connection string (postgresql://user:pass@host:5432/db) in the proxy environment and restart

BerriAI/litellm · error · HTTPException

Postgres DB Not connected

Error message

Postgres DB Not connected

What it means

Returned by POST /customer/block (alias /end_user/block) when the proxy's prisma_client is None, i.e. the process started without a database connection. Blocking end users requires upserting blocked=true rows in LiteLLM_EndUserTable, so without Prisma the endpoint aborts with HTTP 500 'Postgres DB Not connected'.

Source

Thrown at litellm/proxy/management_endpoints/customer_endpoints.py:180

        ```
    """
    from litellm.proxy.proxy_server import prisma_client

    try:
        records: Final = []
        if prisma_client is not None:
            for id in data.user_ids:
                record = await _typed_table(EndUserRepository(prisma_client)).upsert(
                    where={"user_id": id},
                    data={
                        "create": {"user_id": id, "blocked": True},
                        "update": {"blocked": True},
                    },
                )
                records.append(record)
            await _evict_end_user_cache_keys(_end_user_cache_keys(data.user_ids))
        else:
            raise HTTPException(
                status_code=500,
                detail={"error": "Postgres DB Not connected"},
            )

        return {"blocked_users": records}
    except Exception as e:
        verbose_proxy_logger.error("An error occurred - %s", e)
        raise HTTPException(status_code=500, detail={"error": str(e)})


@router.post(
    "/end_user/unblock",
    tags=["Customer Management"],
    dependencies=[Depends(user_api_key_auth)],
    include_in_schema=False,
)
@router.post(
    "/customer/unblock",

View on GitHub (pinned to 77b7c6c40c)

Solutions

Set DATABASE_URL to a PostgreSQL connection string (postgresql://user:pass@host:5432/db) in the proxy environment and restart
Check proxy startup logs for Prisma connect errors and fix credentials/network/SSL issues
Confirm the DB is live via another DB-backed route (e.g. GET /key/info or GET /customer/list) before retrying the block
If you do not need persistence, stop calling block endpoints - they have no in-memory mode

Example fix

# before
litellm --config config.yaml   # config.yaml has no database_url -> /customer/block 500s

# after
export DATABASE_URL=postgresql://postgres:postgres@localhost:5432/litellm
litellm --config config.yaml

Defensive patterns

Strategy: validation

Validate before calling

import os


def proxy_has_database() -> bool:
    url = os.getenv("DATABASE_URL", "")
    return url.startswith(("postgresql://", "postgres://"))


assert proxy_has_database(), "proxy needs DATABASE_URL before /customer/block can work"

Try / catch

try:
    resp = client.post("/customer/block", json={"user_ids": ids})
except HTTPException as e:
    if e.status_code == 500 and "Not connected" in str(e.detail):
        raise RuntimeError("proxy has no DATABASE_URL configured") from e
    raise

Prevention

Always configure DATABASE_URL when deploying the proxy with management endpoints
Monitor startup logs for Prisma connection failures instead of discovering them via 500s on first API call
Remember LiteLLM management endpoints are Postgres-only - sqlite/mysql URLs will not satisfy them

When it happens

Trigger: POST /customer/block (or /end_user/block) with {"user_ids": [...]} against a proxy started without DATABASE_URL, or whose Prisma connection failed at startup leaving prisma_client None.

Common situations: Running litellm via pip with a config.yaml that has no database_url; DATABASE_URL pointing at a non-Postgres engine (management endpoints need postgresql://); the DB was unreachable at boot so every DB-backed management route 500s.

Related errors

AI-assisted analysis of BerriAI/litellm@77b7c6c40c (2026-08-18). Data as JSON: /api/errors/abb055a01d136153. Report an issue: GitHub.