[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
This commit is contained in:
@@ -31,6 +31,8 @@ from vllm.multimodal.utils import (async_get_and_parse_image,
|
||||
get_full_image_text_prompt)
|
||||
from vllm.outputs import RequestOutput
|
||||
from vllm.sequence import Logprob
|
||||
from vllm.tracing import (contains_trace_headers, extract_trace_headers,
|
||||
log_tracing_disabled_warning)
|
||||
from vllm.utils import random_uuid
|
||||
|
||||
logger = init_logger(__name__)
|
||||
@@ -267,11 +269,20 @@ class OpenAIServingChat(OpenAIServing):
|
||||
if image_data is not None:
|
||||
inputs["multi_modal_data"] = image_data
|
||||
|
||||
is_tracing_enabled = await self.engine.is_tracing_enabled()
|
||||
trace_headers = None
|
||||
if is_tracing_enabled and raw_request:
|
||||
trace_headers = extract_trace_headers(raw_request.headers)
|
||||
if not is_tracing_enabled and raw_request and contains_trace_headers(
|
||||
raw_request.headers):
|
||||
log_tracing_disabled_warning()
|
||||
|
||||
result_generator = self.engine.generate(
|
||||
inputs,
|
||||
sampling_params,
|
||||
request_id,
|
||||
lora_request,
|
||||
trace_headers=trace_headers,
|
||||
)
|
||||
# Streaming response
|
||||
if request.stream:
|
||||
|
||||
@@ -24,6 +24,8 @@ from vllm.model_executor.guided_decoding import (
|
||||
get_guided_decoding_logits_processor)
|
||||
from vllm.outputs import RequestOutput
|
||||
from vllm.sequence import Logprob
|
||||
from vllm.tracing import (contains_trace_headers, extract_trace_headers,
|
||||
log_tracing_disabled_warning)
|
||||
from vllm.utils import merge_async_iterators, random_uuid
|
||||
|
||||
logger = init_logger(__name__)
|
||||
@@ -125,6 +127,14 @@ class OpenAIServingCompletion(OpenAIServing):
|
||||
truncate_prompt_tokens)
|
||||
prompt_ids, prompt_text = prompt_formats
|
||||
|
||||
is_tracing_enabled = await self.engine.is_tracing_enabled()
|
||||
trace_headers = None
|
||||
if is_tracing_enabled:
|
||||
trace_headers = extract_trace_headers(raw_request.headers)
|
||||
if not is_tracing_enabled and contains_trace_headers(
|
||||
raw_request.headers):
|
||||
log_tracing_disabled_warning()
|
||||
|
||||
generator = self.engine.generate(
|
||||
{
|
||||
"prompt": prompt_text,
|
||||
@@ -133,6 +143,7 @@ class OpenAIServingCompletion(OpenAIServing):
|
||||
sampling_params,
|
||||
f"{request_id}-{i}",
|
||||
lora_request=lora_request,
|
||||
trace_headers=trace_headers,
|
||||
)
|
||||
|
||||
generators.append(generator)
|
||||
|
||||
Reference in New Issue
Block a user