forked from phoenix/litellm-mirror
(feat) Allow failed DB connection requests to allow virtual keys with allow_failed_db_requests
(#6605)
* fix use helper for _handle_failed_db_connection_for_get_key_object * track ALLOW_FAILED_DB_REQUESTS on prometheus * fix allow_failed_db_requests check * fix allow_requests_on_db_unavailable * fix allow_requests_on_db_unavailable * docs allow_requests_on_db_unavailable * identify user_id as litellm_proxy_admin_name when DB is failing * test_handle_failed_db_connection * fix test_user_api_key_auth_db_unavailable * update best practices for prod doc * update best practices for prod * fix handle db failure
This commit is contained in:
parent
eb171e6d95
commit
e3519aa5ae
6 changed files with 224 additions and 3 deletions
|
@ -12,6 +12,11 @@ sys.path.insert(
|
|||
0, os.path.abspath("../..")
|
||||
) # Adds the parent directory to the system path
|
||||
import pytest, litellm
|
||||
import httpx
|
||||
from litellm.proxy.auth.auth_checks import (
|
||||
_handle_failed_db_connection_for_get_key_object,
|
||||
)
|
||||
from litellm.proxy._types import UserAPIKeyAuth
|
||||
from litellm.proxy.auth.auth_checks import get_end_user_object
|
||||
from litellm.caching.caching import DualCache
|
||||
from litellm.proxy._types import LiteLLM_EndUserTable, LiteLLM_BudgetTable
|
||||
|
@ -60,3 +65,33 @@ async def test_get_end_user_object(customer_spend, customer_budget):
|
|||
customer_spend, customer_budget, str(e)
|
||||
)
|
||||
)
|
||||
|
||||
|
||||
@pytest.mark.asyncio
|
||||
async def test_handle_failed_db_connection():
|
||||
"""
|
||||
Test cases:
|
||||
1. When allow_requests_on_db_unavailable=True -> return UserAPIKeyAuth
|
||||
2. When allow_requests_on_db_unavailable=False -> raise original error
|
||||
"""
|
||||
from litellm.proxy.proxy_server import general_settings, litellm_proxy_admin_name
|
||||
|
||||
# Test case 1: allow_requests_on_db_unavailable=True
|
||||
general_settings["allow_requests_on_db_unavailable"] = True
|
||||
mock_error = httpx.ConnectError("Failed to connect to DB")
|
||||
|
||||
result = await _handle_failed_db_connection_for_get_key_object(e=mock_error)
|
||||
|
||||
assert isinstance(result, UserAPIKeyAuth)
|
||||
assert result.key_name == "failed-to-connect-to-db"
|
||||
assert result.token == "failed-to-connect-to-db"
|
||||
assert result.user_id == litellm_proxy_admin_name
|
||||
|
||||
# Test case 2: allow_requests_on_db_unavailable=False
|
||||
general_settings["allow_requests_on_db_unavailable"] = False
|
||||
|
||||
with pytest.raises(httpx.ConnectError) as exc_info:
|
||||
await _handle_failed_db_connection_for_get_key_object(e=mock_error)
|
||||
print("_handle_failed_db_connection_for_get_key_object got exception", exc_info)
|
||||
|
||||
assert str(exc_info.value) == "Failed to connect to DB"
|
||||
|
|
|
@ -28,6 +28,7 @@ from datetime import datetime
|
|||
from dotenv import load_dotenv
|
||||
from fastapi import Request
|
||||
from fastapi.routing import APIRoute
|
||||
import httpx
|
||||
|
||||
load_dotenv()
|
||||
import io
|
||||
|
@ -51,6 +52,7 @@ from litellm.proxy.management_endpoints.internal_user_endpoints import (
|
|||
user_info,
|
||||
user_update,
|
||||
)
|
||||
from litellm.proxy.auth.auth_checks import get_key_object
|
||||
from litellm.proxy.management_endpoints.key_management_endpoints import (
|
||||
delete_key_fn,
|
||||
generate_key_fn,
|
||||
|
@ -3307,3 +3309,106 @@ async def test_service_accounts(prisma_client):
|
|||
print("response from user_api_key_auth", result)
|
||||
|
||||
setattr(litellm.proxy.proxy_server, "general_settings", {})
|
||||
|
||||
|
||||
@pytest.mark.asyncio
|
||||
async def test_user_api_key_auth_db_unavailable():
|
||||
"""
|
||||
Test that user_api_key_auth handles DB connection failures appropriately when:
|
||||
1. DB connection fails during token validation
|
||||
2. allow_requests_on_db_unavailable=True
|
||||
"""
|
||||
litellm.set_verbose = True
|
||||
|
||||
# Mock dependencies
|
||||
class MockPrismaClient:
|
||||
async def get_data(self, *args, **kwargs):
|
||||
print("MockPrismaClient.get_data() called")
|
||||
raise httpx.ConnectError("Failed to connect to DB")
|
||||
|
||||
async def connect(self):
|
||||
print("MockPrismaClient.connect() called")
|
||||
pass
|
||||
|
||||
class MockDualCache:
|
||||
async def async_get_cache(self, *args, **kwargs):
|
||||
return None
|
||||
|
||||
async def async_set_cache(self, *args, **kwargs):
|
||||
pass
|
||||
|
||||
async def set_cache(self, *args, **kwargs):
|
||||
pass
|
||||
|
||||
# Set up test environment
|
||||
setattr(litellm.proxy.proxy_server, "prisma_client", MockPrismaClient())
|
||||
setattr(litellm.proxy.proxy_server, "user_api_key_cache", MockDualCache())
|
||||
setattr(litellm.proxy.proxy_server, "master_key", "sk-1234")
|
||||
setattr(
|
||||
litellm.proxy.proxy_server,
|
||||
"general_settings",
|
||||
{"allow_requests_on_db_unavailable": True},
|
||||
)
|
||||
|
||||
# Create test request
|
||||
request = Request(scope={"type": "http"})
|
||||
request._url = URL(url="/chat/completions")
|
||||
|
||||
# Run test with a sample API key
|
||||
result = await user_api_key_auth(
|
||||
request=request,
|
||||
api_key="Bearer sk-123456789",
|
||||
)
|
||||
|
||||
# Verify results
|
||||
assert isinstance(result, UserAPIKeyAuth)
|
||||
assert result.key_name == "failed-to-connect-to-db"
|
||||
assert result.user_id == litellm.proxy.proxy_server.litellm_proxy_admin_name
|
||||
|
||||
|
||||
@pytest.mark.asyncio
|
||||
async def test_user_api_key_auth_db_unavailable_not_allowed():
|
||||
"""
|
||||
Test that user_api_key_auth raises an exception when:
|
||||
This is default behavior
|
||||
|
||||
1. DB connection fails during token validation
|
||||
2. allow_requests_on_db_unavailable=False (default behavior)
|
||||
"""
|
||||
|
||||
# Mock dependencies
|
||||
class MockPrismaClient:
|
||||
async def get_data(self, *args, **kwargs):
|
||||
print("MockPrismaClient.get_data() called")
|
||||
raise httpx.ConnectError("Failed to connect to DB")
|
||||
|
||||
async def connect(self):
|
||||
print("MockPrismaClient.connect() called")
|
||||
pass
|
||||
|
||||
class MockDualCache:
|
||||
async def async_get_cache(self, *args, **kwargs):
|
||||
return None
|
||||
|
||||
async def async_set_cache(self, *args, **kwargs):
|
||||
pass
|
||||
|
||||
async def set_cache(self, *args, **kwargs):
|
||||
pass
|
||||
|
||||
# Set up test environment
|
||||
setattr(litellm.proxy.proxy_server, "prisma_client", MockPrismaClient())
|
||||
setattr(litellm.proxy.proxy_server, "user_api_key_cache", MockDualCache())
|
||||
setattr(litellm.proxy.proxy_server, "general_settings", {})
|
||||
setattr(litellm.proxy.proxy_server, "master_key", "sk-1234")
|
||||
|
||||
# Create test request
|
||||
request = Request(scope={"type": "http"})
|
||||
request._url = URL(url="/chat/completions")
|
||||
|
||||
# Run test with a sample API key
|
||||
with pytest.raises(litellm.proxy._types.ProxyException):
|
||||
await user_api_key_auth(
|
||||
request=request,
|
||||
api_key="Bearer sk-123456789",
|
||||
)
|
||||
|
|
Loading…
Add table
Add a link
Reference in a new issue