[model-gateway] reorganize metrics, logging, and otel to its own module (#14590)

sgl-model-gateway/bindings/python/src/lib.rs

@@ -659,7 +659,7 @@ impl Router {
     }
 
     fn start(&self) -> PyResult<()> {
-        use metrics::PrometheusConfig;
+        use observability::metrics::PrometheusConfig;
 
         let router_config = self.to_router_config().map_err(|e| {
             pyo3::exceptions::PyValueError::new_err(format!("Configuration error: {}", e))

sgl-model-gateway/src/core/job_queue.rs

@@ -24,7 +24,7 @@ use crate::{
         WorkflowContext, WorkflowEngine, WorkflowId, WorkflowInstanceId, WorkflowStatus,
     },
     mcp::McpConfig,
-    metrics::RouterMetrics,
+    observability::metrics::RouterMetrics,
     protocols::worker_spec::{JobStatus, WorkerConfigRequest},
 };
 

sgl-model-gateway/src/core/worker.rs

@@ -17,7 +17,7 @@ use super::{
 };
 use crate::{
     core::{BasicWorkerBuilder, CircuitState, DPAwareWorkerBuilder},
-    metrics::RouterMetrics,
+    observability::metrics::RouterMetrics,
     protocols::worker_spec::WorkerInfo,
     routers::grpc::client::GrpcClient,
 };

sgl-model-gateway/src/lib.rs

@@ -1,15 +1,12 @@
 pub mod app_context;
 pub mod config;
-pub mod logging;
-
 pub mod core;
 pub mod data_connector;
 pub mod grpc_client;
 pub mod mcp;
-pub mod metrics;
 pub mod middleware;
 pub mod multimodal;
-pub mod otel_trace;
+pub mod observability;
 pub mod policies;
 pub mod protocols;
 pub mod reasoning_parser;

sgl-model-gateway/src/main.rs

@@ -8,8 +8,10 @@ use sgl_model_gateway::{
         RouterConfig, RoutingMode, TokenizerCacheConfig, TraceConfig,
     },
     core::ConnectionMode,
-    metrics::PrometheusConfig,
+    observability::{
-    otel_trace::{is_otel_enabled, shutdown_otel},
+        metrics::PrometheusConfig,
+        otel_trace::{is_otel_enabled, shutdown_otel},
+    },
     server::{self, ServerConfig},
     service_discovery::ServiceDiscoveryConfig,
     version,

sgl-model-gateway/src/middleware.rs

@@ -22,7 +22,7 @@ use tracing::{debug, error, field::Empty, info, info_span, warn, Span};
 
 pub use crate::core::token_bucket::TokenBucket;
 use crate::{
-    metrics::RouterMetrics,
+    observability::metrics::RouterMetrics,
     server::AppState,
     wasm::{
         module::{MiddlewareAttachPoint, WasmModuleAttachPoint},

sgl-model-gateway/src/observability/events.rs

@@ -2,7 +2,7 @@
 
 use tracing::{debug, event, Level};
 
-use crate::otel_trace::is_otel_enabled;
+use crate::observability::otel_trace::is_otel_enabled;
 
 pub fn get_module_path() -> &'static str {
     module_path!()

sgl-model-gateway/src/observability/logging.rs

@@ -10,7 +10,8 @@ use tracing_subscriber::{
     fmt::time::ChronoUtc, layer::SubscriberExt, util::SubscriberInitExt, EnvFilter, Layer,
 };
 
-use crate::{config::TraceConfig, otel_trace::get_otel_layer};
+use super::otel_trace::get_otel_layer;
+use crate::config::TraceConfig;
 
 #[derive(Debug, Clone)]
 pub struct LoggingConfig {

sgl-model-gateway/src/observability/mod.rs

@@ -0,0 +1,6 @@
+//! Observability utilities for logging, metrics, and tracing.
+
+pub mod events;
+pub mod logging;
+pub mod metrics;
+pub mod otel_trace;

sgl-model-gateway/src/observability/otel_trace.rs

@@ -25,7 +25,7 @@ use tracing_subscriber::{
     Layer,
 };
 
-use crate::routers::http::events::get_module_path as http_router_get_module_path;
+use super::events::get_module_path as http_router_get_module_path;
 
 static ENABLED: AtomicBool = AtomicBool::new(false);
 

sgl-model-gateway/src/policies/cache_aware.rs

@@ -66,7 +66,7 @@ use rand::Rng;
 use tracing::debug;
 
 use super::{get_healthy_worker_indices, tree::Tree, CacheAwareConfig, LoadBalancingPolicy};
-use crate::{core::Worker, metrics::RouterMetrics};
+use crate::{core::Worker, observability::metrics::RouterMetrics};
 
 /// Cache-aware routing policy
 ///

sgl-model-gateway/src/policies/power_of_two.rs

@@ -9,7 +9,7 @@ use rand::Rng;
 use tracing::info;
 
 use super::{get_healthy_worker_indices, LoadBalancingPolicy};
-use crate::{core::Worker, metrics::RouterMetrics};
+use crate::{core::Worker, observability::metrics::RouterMetrics};
 
 /// Power-of-two choices policy
 ///

sgl-model-gateway/src/policies/random.rs

@@ -5,7 +5,7 @@ use std::sync::Arc;
 use rand::Rng;
 
 use super::{get_healthy_worker_indices, LoadBalancingPolicy};
-use crate::{core::Worker, metrics::RouterMetrics};
+use crate::{core::Worker, observability::metrics::RouterMetrics};
 
 /// Random selection policy
 ///

sgl-model-gateway/src/policies/round_robin.rs

@@ -6,7 +6,7 @@ use std::sync::{
 };
 
 use super::{get_healthy_worker_indices, LoadBalancingPolicy};
-use crate::{core::Worker, metrics::RouterMetrics};
+use crate::{core::Worker, observability::metrics::RouterMetrics};
 
 /// Round-robin selection policy
 ///

sgl-model-gateway/src/routers/http/mod.rs

@@ -1,6 +1,5 @@
 //! HTTP router implementations
 
-pub mod events;
 pub mod pd_router;
 pub mod pd_types;
 pub mod router;

sgl-model-gateway/src/routers/http/pd_router.rs

@@ -14,17 +14,17 @@ use serde_json::{json, Value};
 use tokio_stream::wrappers::UnboundedReceiverStream;
 use tracing::{debug, error, warn};
 
-use super::{
+use super::pd_types::api_path;
-    events::{self, Event},
-    pd_types::api_path,
-};
 use crate::{
     config::types::RetryConfig,
     core::{
         is_retryable_status, RetryExecutor, Worker, WorkerLoadGuard, WorkerRegistry, WorkerType,
     },
-    metrics::RouterMetrics,
+    observability::{
-    otel_trace::inject_trace_context_http,
+        events::{self, Event},
+        metrics::RouterMetrics,
+        otel_trace::inject_trace_context_http,
+    },
     policies::{LoadBalancingPolicy, PolicyRegistry},
     protocols::{
         chat::{ChatCompletionRequest, ChatMessage, MessageContent},

sgl-model-gateway/src/routers/http/router.rs

@@ -15,14 +15,16 @@ use reqwest::Client;
 use tokio_stream::wrappers::UnboundedReceiverStream;
 use tracing::{debug, error};
 
-use super::events::{self, Event};
 use crate::{
     config::types::RetryConfig,
     core::{
         is_retryable_status, ConnectionMode, RetryExecutor, Worker, WorkerRegistry, WorkerType,
     },
-    metrics::RouterMetrics,
+    observability::{
-    otel_trace::inject_trace_context_http,
+        events::{self, Event},
+        metrics::RouterMetrics,
+        otel_trace::inject_trace_context_http,
+    },
     policies::PolicyRegistry,
     protocols::{
         chat::ChatCompletionRequest,

sgl-model-gateway/src/server.rs

@@ -32,10 +32,12 @@ use crate::{
         },
         Job, JobQueue, JobQueueConfig, WorkerManager, WorkerType,
     },
-    logging::{self, LoggingConfig},
-    metrics::{self, PrometheusConfig},
     middleware::{self, AuthConfig, QueuedRequest},
-    otel_trace,
+    observability::{
+        logging::{self, LoggingConfig},
+        metrics::{self, PrometheusConfig},
+        otel_trace,
+    },
     protocols::{
         chat::ChatCompletionRequest,
         classify::ClassifyRequest,

sgl-model-gateway/tests/otel_tracing_test.rs

@@ -20,7 +20,7 @@ use serial_test::serial;
 use sgl_model_gateway::{
     config::{RouterConfig, TraceConfig},
     core::Job,
-    logging, otel_trace,
+    observability::{logging, otel_trace},
     routers::RouterFactory,
 };
 use tokio::sync::oneshot;