scriptoriumadmin/litellm
1
0
Fork 0
mirror of https://github.com/BerriAI/litellm.git synced 2026-10-03 02:22:24 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 82
litellm/tests/test_litellm/proxy
History
Exact
Yuneng Jiang 6553bc8957
fix(proxy): serialize model reconciles so concurrent writes stop evicting each other 
A model write is a read-modify-write of the shared `llm_router` global: read the
db into a snapshot, then make the router match that snapshot. Nothing serialized
it, so two of them interleaving was not a lost update but an eviction --
_delete_deployment removes every live deployment absent from the snapshot it was
handed, so the request holding the older snapshot reconciles the newer request's
model straight back out of the router. The row survives in the db, which is what
makes it easy to miss: the pod simply stops serving a model it was told to serve
until some later reload happens to put it back.

clear_cache compounds it. It deletes every db model from the router before
reloading them, so for the width of that reload the pod serves none of them --
and any concurrent write sampling the router in that window sees the hole.

Fix is one lock (MODEL_RECONCILE_LOCK) held across both, so each reconcile reads
the db and applies it atomically and no stale snapshot can evict a newer model.
clear_cache holds it across wipe+reload and calls the already-locked
_add_deployment_locked, since asyncio.Lock is not reentrant and routing back
through the public add_deployment would deadlock the pod's whole model-write
path.

The verdict needed the same treatment. raise_if_reload_degraded_serving compared
a desired-set read during the reload against a router snapshot taken after it,
so a neighbouring reconcile's in-flight wipe was reported to the caller as
collateral damage from its own reload -- a 500 on a create that had in fact
succeeded. Reconciles now return a ReconcileOutcome carrying both the desired set
and the post-reconcile serving state, captured before the lock is released, and
the verdict judges against that. Omitting live_after keeps the old live re-read,
which stays correct for the no-reconcile-ran case.

Found by running the e2e suite with pytest-xdist at 8 workers: three unrelated
tests failed together on "Previously served model id(s) [...] are also no longer
being served by this pod", which is this. Serial runs concurrent enough to hit it
are rare, which is why 78 minutes of sequential e2e never surfaced it -- but any
customer provisioning models in parallel (terraform, CI) is in exactly this race.

test_reconciles_serialize_so_no_stale_snapshot_can_evict fails with 5 == 1
without the lock.
2026-08-12 10:43:29 -07:00
..
_experimental/mcp_server perf: build log messages lazily so filtered-out log records cost nothing (#35703) 2026-08-04 04:34:52 +00:00
a2a fix(a2a): reject malformed protocolVersion suffixes while keeping semver prereleases 2026-07-21 13:24:58 -07:00
agent_endpoints fix(proxy): deny agent access when key and team grants resolve to nothing (#36221) 2026-08-07 20:44:11 +00:00
analytics_endpoints feat(ui): split failed requests into their own series on the cache dashboard (#34862) 2026-07-29 09:48:17 -07:00
anthropic_endpoints fix(types): correct annotations that were false about their runtime values 2026-08-06 04:37:48 +00:00
auth fix(otel): mark v2 server spans as failed for pre-call errors (#34546) 2026-08-08 12:40:00 -07:00
batches_endpoints fix(proxy): authenticate managed ids before routing 2026-08-08 02:42:34 -07:00
client feat(cli): read base_url from persistent config file (#35015) 2026-07-29 18:25:05 -07:00
common_utils fix(proxy): invalidate cached project object on project update and delete (#36028) 2026-08-07 15:19:00 +00:00
config_resolvers feat(proxy): add SAML 2.0 SSO for the admin UI (#31429) 2026-07-24 12:51:28 -07:00
credential_endpoints fix(proxy): return the real status code when a credential update is rejected (#36166) 2026-08-07 08:02:52 +00:00
db feat(auto-router): track turns per complexity tier (LIT-5302) (#36209) 2026-08-07 17:03:34 -07:00
discovery_endpoints chore: litellm oss staging160626 (#30527) 2026-06-16 18:23:13 -07:00
enterprise_billing feat(proxy): push-based OTLP billable-request metering for enterprise deployments (#31592) 2026-07-15 12:12:52 -07:00
experimental/mcp_server
fine_tuning_endpoints fix(proxy): authenticate managed ids before routing 2026-08-08 02:42:34 -07:00
google_endpoints chore: litellm oss staging160626 (#30527) 2026-06-16 18:23:13 -07:00
guardrails fix(guardrails): chunk oversized Bedrock ApplyGuardrail requests instead of failing (#36119) 2026-08-07 19:44:24 -07:00
health_endpoints fix(health): drop a stored-credential reference along with the credentials it names 2026-08-05 16:07:04 -07:00
hooks feat(proxy): add apply_user_budget_to_team_keys opt-in (#36102) 2026-08-07 15:40:13 +00:00
image_endpoints fix(proxy)!: share one destination check between body and path-supplied model 2026-08-05 15:17:43 -07:00
logging_endpoints feat(proxy): add POST /v1/callbacks/logs to replay logging payloads through callbacks (#31134) 2026-06-24 15:25:10 -07:00
management_endpoints fix(proxy): serialize model reconciles so concurrent writes stop evicting each other 2026-08-12 10:43:29 -07:00
management_helpers feat(team): custom metadata validation hook for team create and update (#33353) 2026-08-03 18:37:45 -07:00
memory fix(proxy): give proxy_admin_viewer read parity with proxy_admin (#35851) 2026-08-05 18:33:55 +00:00
middleware feat(sgr): make the gateway middleware the source of truth for successful requests (#35717) 2026-08-05 12:40:47 -07:00
openai_files_endpoint fix(proxy): authenticate managed ids before routing 2026-08-08 02:42:34 -07:00
pass_through_endpoints fix(proxy): resolve pass-through credentials live from router deployments 2026-08-04 23:01:37 -07:00
policy_engine fix(policy_engine): warn that the config-defined policy reactivates when all DB versions are deleted 2026-07-30 19:48:35 -07:00
prompts fix(proxy): give proxy_admin_viewer read parity with proxy_admin (#35851) 2026-08-05 18:33:55 +00:00
proxy_server fix(a2a): align agent list annotation and test with the tuple return type 2026-08-07 14:08:58 -07:00
public_endpoints fix(proxy): derive config agent ids from agent_name so grants survive secret rotation (#36020) 2026-08-07 19:19:49 +00:00
rag_endpoints fix(rag): track LLM completion usage and spend for /v1/rag/query (#32438) 2026-07-17 17:45:27 +00:00
realtime_endpoints chore: litellm oss 170626 (#30637) 2026-06-17 21:11:12 -07:00
response_api_endpoints fix(proxy): enforce per-model budgets against resolved cursor model variants 2026-08-04 14:36:04 -07:00
shutdown feat(proxy): native /health/drain preStop hook for graceful shutdown (#29439) 2026-06-02 16:30:44 -07:00
spend_tracking feat(spend): rebuild the auto-router benchmarks backend as a per-session rollup (#35910) 2026-08-05 20:06:32 +00:00
test_configs
types_utils fix(proxy): resolve router_settings.plugins dotted paths and load plugins from installed packages (#33644) 2026-07-17 11:23:18 -07:00
ui_crud_endpoints feat(ui): add admin-configurable user banner (#35729) 2026-08-04 09:24:29 -07:00
utils fix(proxy): resolve entity access groups in the model listing endpoints (#36230) 2026-08-07 17:45:30 -07:00
vector_store_endpoints test(vector_store): update credential resolution assertion for team_id kwarg 2026-07-18 18:41:12 +00:00
vector_store_files_endpoints fix(proxy): authenticate managed ids before routing 2026-08-08 02:42:34 -07:00
video_endpoints test(videos): add 1:1 test file scaffold for videos component paths (#30631) 2026-06-29 09:12:51 +05:30
__init__.py
conftest.py fix(proxy): avoid DB outage during planned RDS IAM rotation (#34749) 2026-07-28 13:27:50 -07:00
test_aiohttp_cleanup_closed.py
test_aiohttp_session_recovery.py
test_api_key_masking_in_errors.py
test_audio_speech_prometheus_hooks.py fix(proxy): extend response headers hook to streaming, TTS, image gen, and pass-through (#24232) 2026-06-09 22:10:23 +02:00
test_batch_expiry.py fix(proxy): strip LiteLLM policy tracking from OpenAI batch metadata (#28425) 2026-05-26 11:35:42 -07:00
test_batch_metadata_none_fix.py
test_batch_retrieve_bedrock.py
test_batch_x_litellm_model_encoding.py test(batches): move orphan tests into tests/test_litellm for CI coverage (#30510) 2026-06-16 10:20:59 -07:00
test_blocked_response_usage.py feat: litellm oss staging (#31935) 2026-07-03 09:27:31 +05:30
test_budget_reservation.py feat(proxy): add apply_user_budget_to_team_keys opt-in (#36102) 2026-08-07 15:40:13 +00:00
test_caching_routes.py Litellm oss staging 040626 (#29671) 2026-06-04 11:07:20 -07:00
test_chat_completion_metadata.py
test_common_request_processing.py feat(complexity_router): report LLM classifier cost per request via routing_decision and x-litellm-classifier-cost header (#36015) 2026-08-05 16:27:32 -07:00
test_component_allowlists.py fix(gateway): keep the Prometheus /metrics Mount in the gateway route trim (#32317) 2026-07-07 18:36:38 +03:00
test_cors_config.py
test_custom_proxy.py
test_dynamic_mcp_route.py Litellm oss staging 040626 (#29671) 2026-06-04 11:07:20 -07:00
test_empty_model_list.py
test_enforce_user_param.py
test_fallback_management_endpoints.py
test_fastapi_offline_routes.py
test_filter_models_by_team_access_group.py fix: resolve access group names in _filter_models_by_team_id (#25224) 2026-05-01 11:55:36 +05:30
test_health_check_functions.py feat(proxy): skip disable_background_health_check models on GET /health when flag set (#27716) 2026-05-13 09:49:05 -07:00
test_health_check_max_tokens.py fix: skip health check for semantic auto_router deployments (#31668) 2026-06-29 19:47:20 -07:00
test_langfuse_passthrough_security.py
test_lazy_openapi_snapshot.py [Fix] Tests: Move Misplaced Import in Lazy OpenAPI Snapshot Test 2026-05-01 14:24:53 -07:00
test_litellm_pre_call_utils.py feat(router): independent, default-on deployment affinity for the auto-router (#36146) 2026-08-08 13:02:29 -07:00
test_max_budget_env_var.py
test_mcp_asgi_response.py fix(mcp): surface upstream 401 for token-forwarding MCP servers (#27847) 2026-05-13 12:03:36 -07:00
test_model_based_routing_files_batches.py test(batches): move orphan tests into tests/test_litellm for CI coverage (#30510) 2026-06-16 10:20:59 -07:00
test_model_dump_with_preserved_fields.py
test_model_id_header_propagation.py feat(proxy): add x-litellm-model-name response header with deployment model string (#33698) 2026-07-17 20:29:42 -07:00
test_model_info_default_limits.py fix(proxy): align /v1/model/info with router deployments (#30025) 2026-06-10 19:38:21 -07:00
test_model_level_guardrails.py fix(proxy): apply key/team router_settings.model_group_alias (#35486) 2026-08-03 22:09:47 +00:00
test_model_list_healthy_only.py feat: litellm oss 110626 (#30202) 2026-06-11 22:30:26 -07:00
test_modify_response_streaming_passthrough.py fix(proxy): use e.request_data for logging_obj in ModifyResponseException streaming passthrough (#30800) 2026-06-18 23:29:08 -07:00
test_openapi_schema_validation.py fix(proxy): preserve HTTP operations when injecting WebSocket stubs into OpenAPI schema 2026-05-06 00:28:42 +02:00
test_plugin_routes.py feat: litellm plugin architecture v2 (#30688) 2026-06-20 20:37:22 -07:00
test_pricing_field_strip.py chore: litellm oss staging160626 (#30527) 2026-06-16 18:23:13 -07:00
test_prometheus_cleanup.py
test_provider_url_destination_guard.py fix(proxy)!: share one destination check between body and path-supplied model 2026-08-05 15:17:43 -07:00
test_proxy_cli.py feat(proxy): bound DB statement and lock time via general_settings (#35496) 2026-08-01 13:54:17 -07:00
test_proxy_logging_hook_detection.py fix(proxy): recognize inherited apply_guardrail overrides and keep masking guardrails on their own stream hook 2026-07-30 15:33:49 -07:00
test_proxy_server.py feat(proxy): add apply_user_budget_to_team_keys opt-in (#36102) 2026-08-07 15:40:13 +00:00
test_proxy_types.py fix(proxy): allow /key/update to identify the key by key_alias (#34851) 2026-07-29 09:48:08 -07:00
test_proxy_utils.py fix(proxy): stop alerting on health probes that lose the planned engine-restart race (#36141) 2026-08-07 09:57:30 -07:00
test_pyroscope.py Implement normalize_nonempty_secret_str function to trim whitespace from secrets and treat empty values as unset. Update proxy_server to use this function for Grafana credentials. Enhance tests to validate the new normalization behavior. 2026-05-04 18:17:31 +00:00
test_read_model_list.py feat: add minimal rust router + axum ai-gateway calling router.realtime (2/2) (#31135) 2026-06-23 19:16:34 -07:00
test_redis_auth_cache_flag.py fix(proxy): share CLI SSO login sessions across workers without enable_redis_auth_cache (#33261) 2026-07-22 10:29:34 -07:00
test_response_model_sanitization.py perf: eliminate per-request callback scanning on proxy hot path (#27858) 2026-05-14 09:28:31 -07:00
test_route_a2a_models.py
test_route_llm_request.py fix(ci): let the E2E proxy accept the mock testing params its suite sends 2026-08-01 14:57:42 -07:00
test_sensitive_route_auth.py
test_shared_health_check.py feat(proxy): skip disable_background_health_check models on GET /health when flag set (#27716) 2026-05-13 09:49:05 -07:00
test_spend_log_cleanup.py feat(spend): rebuild the auto-router benchmarks backend as a per-session rollup (#35910) 2026-08-05 20:06:32 +00:00
test_swagger_chat_completions.py fix(proxy): avoid misleading multi-method operation ids 2026-04-30 20:44:14 -07:00
test_team_member_update.py feat(ui): add budget duration to edit team member form (#29717) 2026-06-06 17:24:55 -07:00
test_team_org_move.py
test_tools_allowlist_enforcement.py fix(responses-bridge): custom tool round-trip and allowlist preservation for Codex CLI (#32258) 2026-07-06 17:34:27 -07:00
test_update_llm_router_resilience.py fix(proxy): stop model writes 500ing on another pod's delete (#35400) 2026-07-31 18:10:48 -07:00