xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2025-12-10 04:15:01 +08:00
Code Issues Packages Projects Releases Wiki Activity

[BugFix] Prevent exporting duplicate OpenTelemetry spans (#9017)

Browse Source
This commit is contained in:
Ronen Schaffer 2024-10-22 21:11:53 +03:00 committed by GitHub
parent 434984e665
commit cd5601ac37
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
2 changed files with 36 additions and 7 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

30
tests/tracing/test_tracing.py
View File

@ -87,8 +87,19 @@ def test_traces(trace_service):
f"The fake trace service didn't receive a trace within "
f"the {timeout} seconds timeout")
attributes = decode_attributes(trace_service.request.resource_spans[0].
scope_spans[0].spans[0].attributes)
request = trace_service.request
assert len(request.resource_spans) == 1, (
f"Expected 1 resource span, "
f"but got {len(request.resource_spans)}")
assert len(request.resource_spans[0].scope_spans) == 1, (
f"Expected 1 scope span, "
f"but got {len(request.resource_spans[0].scope_spans)}")
assert len(request.resource_spans[0].scope_spans[0].spans) == 1, (
f"Expected 1 span, "
f"but got {len(request.resource_spans[0].scope_spans[0].spans)}")
attributes = decode_attributes(
request.resource_spans[0].scope_spans[0].spans[0].attributes)
assert attributes.get(SpanAttributes.LLM_RESPONSE_MODEL) == model
assert attributes.get(
SpanAttributes.LLM_REQUEST_ID) == outputs[0].request_id
@ -142,8 +153,19 @@ def test_traces_with_detailed_steps(trace_service):
f"The fake trace service didn't receive a trace within "
f"the {timeout} seconds timeout")
attributes = decode_attributes(trace_service.request.resource_spans[0].
scope_spans[0].spans[0].attributes)
request = trace_service.request
assert len(request.resource_spans) == 1, (
f"Expected 1 resource span, "
f"but got {len(request.resource_spans)}")
assert len(request.resource_spans[0].scope_spans) == 1, (
f"Expected 1 scope span, "
f"but got {len(request.resource_spans[0].scope_spans)}")
assert len(request.resource_spans[0].scope_spans[0].spans) == 1, (
f"Expected 1 span, "
f"but got {len(request.resource_spans[0].scope_spans[0].spans)}")
attributes = decode_attributes(
request.resource_spans[0].scope_spans[0].spans[0].attributes)
assert attributes.get(SpanAttributes.LLM_RESPONSE_MODEL) == model
assert attributes.get(
SpanAttributes.LLM_REQUEST_ID) == outputs[0].request_id

13
vllm/engine/llm_engine.py
View File

@ -1245,7 +1245,7 @@ class LLMEngine:
skip)
# Tracing
self.do_tracing(scheduler_outputs)
self.do_tracing(scheduler_outputs, finished_before)
return None
@ -1840,11 +1840,18 @@ class LLMEngine:
def is_tracing_enabled(self) -> bool:
return self.tracer is not None
def do_tracing(self, scheduler_outputs: SchedulerOutputs) -> None:
def do_tracing(self,
scheduler_outputs: SchedulerOutputs,
finished_before: Optional[List[int]] = None) -> None:
if self.tracer is None:
return
for scheduled_seq_group in scheduler_outputs.scheduled_seq_groups:
for idx, scheduled_seq_group in enumerate(
scheduler_outputs.scheduled_seq_groups):
# Skip double tracing when using async output proc
if finished_before and idx in finished_before:
continue
seq_group = scheduled_seq_group.seq_group
if seq_group.is_finished():
self.create_trace_span(seq_group)