phoenix/litellm
1
0
Fork 0
forked from phoenix/litellm-mirror
Code Pull requests Activity

(fix) Prometheus - Log Postgres DB latency, status on prometheus (#6484)

Browse source 
* fix logging DB fails on prometheus

* unit testing log to otel wrapper

* unit testing for service logger + prometheus

* use LATENCY buckets for service logging

* fix service logging
This commit is contained in:
Ishaan Jaff 2024-10-29 12:17:35 +05:30 • committed by GitHub
parent 4f8a3fd4cf
commit 69b1bc1f1e
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
5 changed files with 229 additions and 49 deletions
Download patch file Download diff file Expand all files Collapse all files

61
litellm/proxy/utils.py
View file

@ -140,25 +140,21 @@ def safe_deep_copy(data):
def log_to_opentelemetry(func):
@wraps(func)
async def wrapper(*args, **kwargs):
start_time = datetime.now()
start_time: datetime = datetime.now()
try:
result = await func(*args, **kwargs)
end_time = datetime.now()
end_time: datetime = datetime.now()
from litellm.proxy.proxy_server import proxy_logging_obj
# Log to OTEL only if "parent_otel_span" is in kwargs and is not None
if (
"parent_otel_span" in kwargs
and kwargs["parent_otel_span"] is not None
and "proxy_logging_obj" in kwargs
and kwargs["proxy_logging_obj"] is not None
):
proxy_logging_obj = kwargs["proxy_logging_obj"]
if "PROXY" not in func.__name__:
await proxy_logging_obj.service_logging_obj.async_service_success_hook(
service=ServiceTypes.DB,
call_type=func.__name__,
parent_otel_span=kwargs["parent_otel_span"],
duration=0.0,
parent_otel_span=kwargs.get("parent_otel_span", None),
duration=(end_time - start_time).total_seconds(),
start_time=start_time,
end_time=end_time,
event_metadata={
@ -179,8 +175,6 @@ def log_to_opentelemetry(func):
kwargs=passed_kwargs
)
if parent_otel_span is not None:
from litellm.proxy.proxy_server import proxy_logging_obj
metadata = get_litellm_metadata_from_kwargs(kwargs=passed_kwargs)
await proxy_logging_obj.service_logging_obj.async_service_success_hook(
service=ServiceTypes.BATCH_WRITE_TO_DB,
@ -194,28 +188,23 @@ def log_to_opentelemetry(func):
# end of logging to otel
return result
except Exception as e:
end_time = datetime.now()
if (
"parent_otel_span" in kwargs
and kwargs["parent_otel_span"] is not None
and "proxy_logging_obj" in kwargs
and kwargs["proxy_logging_obj"] is not None
):
proxy_logging_obj = kwargs["proxy_logging_obj"]
await proxy_logging_obj.service_logging_obj.async_service_failure_hook(
error=e,
service=ServiceTypes.DB,
call_type=func.__name__,
parent_otel_span=kwargs["parent_otel_span"],
duration=0.0,
start_time=start_time,
end_time=end_time,
event_metadata={
"function_name": func.__name__,
"function_kwargs": kwargs,
"function_args": args,
},
)
from litellm.proxy.proxy_server import proxy_logging_obj
end_time: datetime = datetime.now()
await proxy_logging_obj.service_logging_obj.async_service_failure_hook(
error=e,
service=ServiceTypes.DB,
call_type=func.__name__,
parent_otel_span=kwargs.get("parent_otel_span"),
duration=(end_time - start_time).total_seconds(),
start_time=start_time,
end_time=end_time,
event_metadata={
"function_name": func.__name__,
"function_kwargs": kwargs,
"function_args": args,
},
)
raise e
return wrapper
@ -348,6 +337,7 @@ class ProxyLogging:
internal_usage_cache=self.internal_usage_cache.dual_cache,
)
self.premium_user = premium_user
self.service_logging_obj = ServiceLogging()
def startup_event(
self,
@ -422,7 +412,6 @@ class ProxyLogging:
self.internal_usage_cache.dual_cache.redis_cache = redis_cache
def _init_litellm_callbacks(self, llm_router: Optional[litellm.Router] = None):
self.service_logging_obj = ServiceLogging()
litellm.callbacks.append(self.max_parallel_request_limiter) # type: ignore
litellm.callbacks.append(self.max_budget_limiter) # type: ignore
litellm.callbacks.append(self.cache_control_check) # type: ignore