[Misc] Add OpenTelemetry support (#4687)

This PR adds basic support for OpenTelemetry distributed tracing.
It includes changes to enable tracing functionality and improve monitoring capabilities.

I've also added a markdown with print-screens to guide users how to use this feature. You can find it here

vllm/engine/arg_utils.py

@@ -7,8 +7,9 @@ from typing import List, Optional, Tuple, Union
 
 from vllm.config import (CacheConfig, DecodingConfig, DeviceConfig,
                          EngineConfig, LoadConfig, LoRAConfig, ModelConfig,
-                         ParallelConfig, SchedulerConfig, SpeculativeConfig,
-                         TokenizerPoolConfig, VisionLanguageConfig)
+                         ObservabilityConfig, ParallelConfig, SchedulerConfig,
+                         SpeculativeConfig, TokenizerPoolConfig,
+                         VisionLanguageConfig)
 from vllm.model_executor.layers.quantization import QUANTIZATION_METHODS
 from vllm.utils import str_to_int_tuple
 
@@ -101,6 +102,8 @@ class EngineArgs:
 
     qlora_adapter_name_or_path: Optional[str] = None
 
+    otlp_traces_endpoint: Optional[str] = None
+
     def __post_init__(self):
         if self.tokenizer is None:
             self.tokenizer = self.model
@@ -599,6 +602,13 @@ class EngineArgs:
                             type=str,
                             default=None,
                             help='Name or path of the QLoRA adapter.')
+
+        parser.add_argument(
+            '--otlp-traces-endpoint',
+            type=str,
+            default=None,
+            help='Target URL to which OpenTelemetry traces will be sent.')
+
         return parser
 
     @classmethod
@@ -757,6 +767,9 @@ class EngineArgs:
         decoding_config = DecodingConfig(
             guided_decoding_backend=self.guided_decoding_backend)
 
+        observability_config = ObservabilityConfig(
+            otlp_traces_endpoint=self.otlp_traces_endpoint)
+
         if (model_config.get_sliding_window() is not None
                 and scheduler_config.chunked_prefill_enabled
                 and not scheduler_config.use_v2_block_manager):
@@ -764,16 +777,19 @@ class EngineArgs:
                 "Chunked prefill is not supported with sliding window. "
                 "Set --disable-sliding-window to disable sliding window.")
 
-        return EngineConfig(model_config=model_config,
-                            cache_config=cache_config,
-                            parallel_config=parallel_config,
-                            scheduler_config=scheduler_config,
-                            device_config=device_config,
-                            lora_config=lora_config,
-                            vision_language_config=vision_language_config,
-                            speculative_config=speculative_config,
-                            load_config=load_config,
-                            decoding_config=decoding_config)
+        return EngineConfig(
+            model_config=model_config,
+            cache_config=cache_config,
+            parallel_config=parallel_config,
+            scheduler_config=scheduler_config,
+            device_config=device_config,
+            lora_config=lora_config,
+            vision_language_config=vision_language_config,
+            speculative_config=speculative_config,
+            load_config=load_config,
+            decoding_config=decoding_config,
+            observability_config=observability_config,
+        )
 
 
 @dataclass