Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[BugFix] Prevent exporting duplicate OpenTelemetry spans #9017

Merged
merged 2 commits into from
Oct 22, 2024
Merged

[BugFix] Prevent exporting duplicate OpenTelemetry spans #9017

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
cd3cfe5
Add assertions to reveal a bug
ronensc Oct 1, 2024
3b2916c
Pass finished_before to do_tracing() to avoid exporting duplicate spans
ronensc Oct 1, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
30 changes: 26 additions & 4 deletions tests/tracing/test_tracing.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -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
Expand Down Expand Up @@ -144,8 +155,19 @@ 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
Expand Down
13 changes: 10 additions & 3 deletions vllm/engine/llm_engine.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1224,7 +1224,7 @@ def update_prefill_num_computed_tokens(
skip)

# Tracing
self.do_tracing(scheduler_outputs)
self.do_tracing(scheduler_outputs, finished_before)
Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Should I also pass the skip list in addition to finished_before, and exclude the sequence groups marked in this list during tracing?


return None

Expand Down Expand Up @@ -1798,11 +1798,18 @@ def stop_profile(self) -> None:
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)
Expand Down
Loading