feat: Pass `parentSampleRate` to `tracesSampler` #15024

lforst · 2025-01-15T10:02:09Z

Ref: #14932
Ref: getsentry/team-sdks#117

This PR mainly does 2 things:

Pick up the sample_rate from the DSC and pass it to the tracesSampler as parentSampleRate
"Unfreeze" or rather set the sample_rate for a DSC of a given span whenever we applied a sample rate to the span's root span. The sample_rate is equivalent to the sample rate the root span has been sampled with. This is to ensure that downstream services can reliably use the sample_rate in the DSC to pass it to the tracesSampler - like we do in this PR (kinda closing the loop).

github-actions · 2025-01-15T10:08:37Z

size-limit report 📦

Path	Size	% Change	Change
@sentry/browser	23.01 KB	+0.14%	+31 B 🔺
@sentry/browser - with treeshaking flags	22.9 KB	+0.13%	+30 B 🔺
@sentry/browser (incl. Tracing)	35.74 KB	+0.17%	+61 B 🔺
@sentry/browser (incl. Tracing, Replay)	72.53 KB	+0.09%	+61 B 🔺
@sentry/browser (incl. Tracing, Replay) - with treeshaking flags	66.09 KB	+0.1%	+64 B 🔺
@sentry/browser (incl. Tracing, Replay with Canvas)	76.78 KB	+0.09%	+64 B 🔺
@sentry/browser (incl. Tracing, Replay, Feedback)	88.8 KB	+0.08%	+70 B 🔺
@sentry/browser (incl. Feedback)	39.22 KB	+0.08%	+32 B 🔺
@sentry/browser (incl. sendFeedback)	27.64 KB	+0.1%	+27 B 🔺
@sentry/browser (incl. FeedbackAsync)	32.4 KB	+0.1%	+33 B 🔺
@sentry/react	25.69 KB	+0.12%	+31 B 🔺
@sentry/react (incl. Tracing)	38.52 KB	+0.16%	+61 B 🔺
@sentry/vue	27.1 KB	+0.22%	+60 B 🔺
@sentry/vue (incl. Tracing)	37.49 KB	+0.15%	+56 B 🔺
@sentry/svelte	23.14 KB	+0.14%	+32 B 🔺
CDN Bundle	24.4 KB	+0.15%	+35 B 🔺
CDN Bundle (incl. Tracing)	36.04 KB	+0.14%	+49 B 🔺
CDN Bundle (incl. Tracing, Replay)	70.66 KB	+0.03%	+18 B 🔺
CDN Bundle (incl. Tracing, Replay, Feedback)	75.82 KB	+0.04%	+29 B 🔺
CDN Bundle - uncompressed	71.25 KB	+0.13%	+94 B 🔺
CDN Bundle (incl. Tracing) - uncompressed	106.95 KB	+0.13%	+137 B 🔺
CDN Bundle (incl. Tracing, Replay) - uncompressed	217.8 KB	+0.07%	+137 B 🔺
CDN Bundle (incl. Tracing, Replay, Feedback) - uncompressed	230.33 KB	+0.06%	+137 B 🔺
@sentry/nextjs (client)	38.63 KB	+0.16%	+60 B 🔺
@sentry/sveltekit (client)	36.27 KB	+0.17%	+62 B 🔺
@sentry/node	156.28 KB	+0.07%	+111 B 🔺
@sentry/node - without tracing	97.37 KB	+0.06%	+53 B 🔺
@sentry/aws-serverless	106.83 KB	+0.09%	+97 B 🔺

View base workflow run

codecov · 2025-01-15T10:26:35Z

❌ 2 Tests Failed:

Tests completed	Failed	Passed	Skipped
691	2	689	297

View the top 2 failed tests by shortest run time

propagation.test.ts Propagates trace for outgoing http requests

Stack Traces | 0.089s run time

propagation.test.ts:6:5 Propagates trace for outgoing http requests

events.test.ts Event emitter
Stack Traces | 0.175s run time
events.test.ts:4:5 Event emitter

To view more test analytics, go to the Test Analytics Dashboard
📢 Thoughts on this report? Let us know!

mydea · 2025-01-20T12:45:26Z

packages/core/src/tracing/sentrySpan.ts

    spans.forEach(span => {
-      delete span.data[SEMANTIC_ATTRIBUTE_SENTRY_CUSTOM_SPAN_NAME];
-      delete span.data[SEMANTIC_ATTRIBUTE_SENTRY_OVERRIDE_TRACE_SAMPLE_RATE];
+      span.data && delete span.data[SEMANTIC_ATTRIBUTE_SENTRY_CUSTOM_SPAN_NAME];


I think we can drop this check (span.data &&), as span data is not optional anymore? 🤔

mydea · 2025-01-20T12:47:50Z

packages/core/src/envelope.ts

+
+    // This attribute is important for internal logic but should not leak into actual data
+    // eslint-disable-next-line @typescript-eslint/no-dynamic-delete
+    delete spanJson.data[SEMANTIC_ATTRIBUTE_SENTRY_OVERRIDE_TRACE_SAMPLE_RATE];


hmmm no strong feelings, but can we move this somewhere else than here? E.g. into sentrySpan (where we already delete the name attribute) and/or the span exporter (where we already clear some data that we do not want to send)? These feel like more correct places for this than the envelope, which should rather just send what it gets, IMHO.

I put this here because there is no other really generic flow that runs for standalone spans. This is basically one-of logic. We already delete it in sentrySpan, but sentrySpan doesn't run for standalones, neither does the span exporter in browser. Do you have any other concrete ideas?

I could also just delete the line. We also seem to be sending other garbage attributes through standalones.

hmm, I see 🤔 you decide, it feels a bit "weird" to do this in envelope serialization step, but also OK to me!

I know. For the lack of better placement I'll leave it. Removing it wouldn't be great since we'd just end up forgetting to do so. It's super easy to refactor once we flesh out standalone spans.

packages/core/src/tracing/dynamicSamplingContext.ts

mydea · 2025-01-20T12:50:19Z

packages/core/src/tracing/dynamicSamplingContext.ts


  // For core implementation, we freeze the DSC onto the span as a non-enumerable property
  const frozenDsc = (rootSpan as SpanWithMaybeDsc)[FROZEN_DSC_FIELD];
  if (frozenDsc) {
-    return frozenDsc;
+    return applyTraceSampleRateOverrideToDsc(frozenDsc);


should this not be on the frozen DSC already in this scenario?

This is effectively what this change is about. The sample_rate becomes "unfrozen" in certain situations so no matter what the dsc may be, the sample rate that was applied to the root span wins.

mydea · 2025-01-20T12:50:39Z

packages/core/src/tracing/dynamicSamplingContext.ts

  const traceStateDsc = traceState?.get('sentry.dsc');

  // If the span has a DSC, we want it to take precedence
  const dscOnTraceState = traceStateDsc && baggageHeaderToDynamicSamplingContext(traceStateDsc);

  if (dscOnTraceState) {
-    return dscOnTraceState;
+    return applyTraceSampleRateOverrideToDsc(dscOnTraceState);


why is this not already on the frozen dsc from the trace state here?

Because for example when the sample rate returned from a tracesSampler is different than on the incoming DSC. The sample rate returned from the tracesSampler wins.

mydea · 2025-01-20T12:54:48Z

packages/opentelemetry/src/sampler.ts

 }: {
  decision: SamplingDecision | undefined;
  context: Context;
  spanAttributes: SpanAttributes;
  sampleRand?: number;
+  sampleRateOverride?: number;


Maybe this is just the naming, but I am confused. This sounds to me as if we are overriding the used sample rate for this sampling decision? But if I read the comment below, I guess it means "propagate this sample rate to downstream spans"? Is this a correct interpretation? 😅

well here we are in wrapSamplingDecision, meaning that the sampling decision was already made. In this context, it kinda means what you wrote: Override the sampling decision on the DSC for everything downstream with this value, no matter what the DSC had as a sample_rate before. Do you have any suggestions?

Renamed it to shouldUpdateSampleRateOnDownstreamTrace for now.

packages/opentelemetry/src/sampler.ts

lforst · 2025-01-21T13:22:11Z

…ling-decision

lforst added 18 commits January 9, 2025 13:25

Remove unused function

ebcf7bf

Merge remote-tracking branch 'origin/develop' into lforst-sample-rand

a83a511

Write and use sample rand on propagation context

1646f0b

Implement propagating sampling decision properly

7fbba52

Inject sample rand

0b12615

Merge remote-tracking branch 'origin/develop' into lforst-sample-rand

1eac9af

format

38fe823

put onto initial dsc

4fb937f

Add tests for tracesSampleRate behavior

6c27931

mv

8d6984f

todo comment

9b394c3

unit tests

1d4a641

node integration tests

2d35988

browser integration tests

957bc09

Fix test

e1c531f

Merge remote-tracking branch 'origin/develop' into lforst-sample-rand

1fbbf82

test fix

a1d5dd3

feat: Pass parentSampleRate to tracesSampler

83ac885

lforst mentioned this pull request Jan 15, 2025

Implement Sampling Seed Propagation #14932

Closed

lforst added 3 commits January 15, 2025 10:10

always apply root span sample rate to dsc

045b41f

e l a b o r a t e

52ea3d2

Dont reinvent util we already have

631ed1b

lforst mentioned this pull request Jan 15, 2025

Update the DSC's sample-rate getsentry/team-sdks#117

Open

lforst added 5 commits January 15, 2025 11:36

beep boop actually do what I am supposed to beep boop

37b6ca2

don't leak attribute into actual data

5e3f706

Merge branch 'lforst-sample-rand' into lforst-parent-sampling-decision

8c83fb9

.

c4282b8

.

4a2a7ae

lforst requested review from chargome and andreiborza January 20, 2025 11:57

mydea reviewed Jan 20, 2025

View reviewed changes

packages/core/src/tracing/dynamicSamplingContext.ts Outdated Show resolved Hide resolved

mydea reviewed Jan 20, 2025

View reviewed changes

packages/opentelemetry/src/sampler.ts Outdated Show resolved Hide resolved

lforst added 2 commits January 20, 2025 13:12

fix and test 0% sample rate

063984b

apply before emit

8607cc8

lforst requested a review from mydea January 20, 2025 13:18

lforst added 10 commits January 20, 2025 15:56

Merge branch 'develop' into lforst-parent-sampling-decision

51526ec

fix merge mistake

08b6561

Refactor again

bceb9f3

lint

e2ffc52

update tests

feb43ee

.

2a127b2

fix crap

70c3fd9

update tests

0ca98e3

more tests

812e750

e2e tests

2882ca4

lforst added 3 commits January 21, 2025 13:35

.

f5a8220

Merge remote-tracking branch 'origin/develop' into lforst-parent-samp…

fbee088

…ling-decision

.

f683884

mydea approved these changes Jan 22, 2025

View reviewed changes

lforst merged commit 5021a58 into develop Jan 22, 2025
149 checks passed

lforst deleted the lforst-parent-sampling-decision branch January 22, 2025 08:23

krystofwoldrich mentioned this pull request Feb 14, 2025

Update the DSC's sample-rate getsentry/sentry-react-native#4540

Open

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

feat: Pass `parentSampleRate` to `tracesSampler` #15024

feat: Pass `parentSampleRate` to `tracesSampler` #15024

lforst commented Jan 15, 2025 •

edited

Loading

github-actions bot commented Jan 15, 2025 •

edited

Loading

codecov bot commented Jan 15, 2025 •

edited

Loading

mydea Jan 20, 2025

mydea Jan 20, 2025

lforst Jan 20, 2025 •

edited

Loading

lforst Jan 20, 2025

mydea Jan 20, 2025

lforst Jan 20, 2025 •

edited

Loading

mydea Jan 20, 2025

lforst Jan 20, 2025

mydea Jan 20, 2025

lforst Jan 20, 2025

mydea Jan 20, 2025 •

edited

Loading

lforst Jan 20, 2025 •

edited

Loading

lforst Jan 20, 2025

lforst commented Jan 21, 2025

feat: Pass parentSampleRate to tracesSampler #15024

feat: Pass parentSampleRate to tracesSampler #15024

Conversation

lforst commented Jan 15, 2025 • edited Loading

github-actions bot commented Jan 15, 2025 • edited Loading

size-limit report 📦

codecov bot commented Jan 15, 2025 • edited Loading

❌ 2 Tests Failed:

Choose a reason for hiding this comment

Choose a reason for hiding this comment

lforst Jan 20, 2025 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

lforst Jan 20, 2025 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

mydea Jan 20, 2025 • edited Loading

Choose a reason for hiding this comment

lforst Jan 20, 2025 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

lforst commented Jan 21, 2025

feat: Pass `parentSampleRate` to `tracesSampler` #15024

feat: Pass `parentSampleRate` to `tracesSampler` #15024

lforst commented Jan 15, 2025 •

edited

Loading

github-actions bot commented Jan 15, 2025 •

edited

Loading

codecov bot commented Jan 15, 2025 •

edited

Loading

lforst Jan 20, 2025 •

edited

Loading

lforst Jan 20, 2025 •

edited

Loading

mydea Jan 20, 2025 •

edited

Loading

lforst Jan 20, 2025 •

edited

Loading