phoenix-oss/llama-stack-mirror
0
0
Fork 1
mirror of https://github.com/meta-llama/llama-stack.git synced 2025-12-17 17:22:38 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

fix: handle provider registration failures gracefully

Browse source 
When a provider fails during model registration or listing, the stack
should continue initializing rather than crashing. This allows the
stack to start even if some providers are misconfigured.

- Added error handling in register_resources()
- Added unit tests to verify error handling behavior
- Improved error logging with provider context
- Removed @pytest.mark.asyncio decorators (pytest already configured with async-mode=auto)

Fixes #3769
This commit is contained in:
Akram Ben Aissi 2025-10-10 22:18:37 +02:00
parent e7d21e1ee3
commit a271e3abae
2 changed files with 235 additions and 11 deletions
Download patch file Download diff file Expand all files Collapse all files

39
llama_stack/core/stack.py
View file

@ -102,6 +102,12 @@ TEST_RECORDING_CONTEXT = None
async def register_resources(run_config: StackRunConfig, impls: dict[Api, Any]):
"""Register resources from the run config with their respective providers.
This function attempts to register each resource (models, shields, etc.) with its provider.
If a registration fails, it logs the error and continues with other resources rather than
crashing the entire stack.
"""
for rsrc, api, register_method, list_method in RESOURCES:
objects = getattr(run_config, rsrc)
if api not in impls:
@ -116,20 +122,31 @@ async def register_resources(run_config: StackRunConfig, impls: dict[Api, Any]):
continue
logger.debug(f"registering {rsrc.capitalize()} {obj} for provider {obj.provider_id}")
# we want to maintain the type information in arguments to method.
# instead of method(**obj.model_dump()), which may convert a typed attr to a dict,
# we use model_dump() to find all the attrs and then getattr to get the still typed value.
await method(**{k: getattr(obj, k) for k in obj.model_dump().keys()})
try:
# we want to maintain the type information in arguments to method.
# instead of method(**obj.model_dump()), which may convert a typed attr to a dict,
# we use model_dump() to find all the attrs and then getattr to get the still typed value.
await method(**{k: getattr(obj, k) for k in obj.model_dump().keys()})
except Exception as e:
# Log the error but continue with other resources
logger.error(
f"Failed to register {rsrc} {obj} for provider {obj.provider_id if hasattr(obj, 'provider_id') else 'unknown'}: {e}"
)
continue
method = getattr(impls[api], list_method)
response = await method()
try:
method = getattr(impls[api], list_method)
response = await method()
objects_to_process = response.data if hasattr(response, "data") else response
objects_to_process = response.data if hasattr(response, "data") else response
for obj in objects_to_process:
logger.debug(
f"{rsrc.capitalize()}: {obj.identifier} served by {obj.provider_id}",
)
for obj in objects_to_process:
logger.debug(
f"{rsrc.capitalize()}: {obj.identifier} served by {obj.provider_id}",
)
except Exception as e:
# Log the error but continue with other resource types
logger.error(f"Failed to list {rsrc}: {e}")
class EnvVarError(Exception):