fix(core): Allow non-recording spans to carry explicit sampling decisions by andreiborza · Pull Request #21406 · getsentry/sentry-javascript

andreiborza · 2026-06-09T14:22:35Z

What

Non-recording spans can now carry a sampled flag and parentSpanId on their span context, so spanToTraceHeader/spanToTraceparentHeader and spanToJSON reflect the real decision.

In Tracing without Performance mode, root non-recording spans keep the sampling decision deferred (no sentry-trace flag, no sentry-sampled/sample_rate in the DSC) instead of asserting a negative decision that would suppress downstream sampling.

Spans created for an unsampled trace (or ignored spans) carry an explicit sampled: false.

Placeholder/idle spans also inherit traceId/parentSpanId from the propagation context and capture their scopes.

Why

These changes are important for when we switch over to our own TracerProvider and Tracer because we will create native Sentry spans (including SentryNonRecordingSpan) and they need to be on par with the previously used OTel spans.

github-actions · 2026-06-09T14:32:41Z

size-limit report 📦

Path	Size	% Change	Change
@sentry/browser	27.41 kB	+0.02%	+4 B 🔺
@sentry/browser - with treeshaking flags	25.84 kB	+0.02%	+4 B 🔺
@sentry/browser (incl. Tracing)	46.23 kB	+1.16%	+530 B 🔺
@sentry/browser (incl. Tracing + Span Streaming)	48.48 kB	+1.14%	+543 B 🔺
@sentry/browser (incl. Tracing, Profiling)	51.01 kB	+1.01%	+507 B 🔺
@sentry/browser (incl. Tracing, Replay)	85.42 kB	+0.59%	+501 B 🔺
@sentry/browser (incl. Tracing, Replay) - with treeshaking flags	75.03 kB	+0.68%	+503 B 🔺
@sentry/browser (incl. Tracing, Replay with Canvas)	90.12 kB	+0.57%	+508 B 🔺
@sentry/browser (incl. Tracing, Replay, Feedback)	102.8 kB	+0.5%	+503 B 🔺
@sentry/browser (incl. Feedback)	44.57 kB	+0.02%	+5 B 🔺
@sentry/browser (incl. sendFeedback)	32.21 kB	+0.02%	+4 B 🔺
@sentry/browser (incl. FeedbackAsync)	37.32 kB	+0.03%	+8 B 🔺
@sentry/browser (incl. Metrics)	28.48 kB	+0.04%	+10 B 🔺
@sentry/browser (incl. Logs)	28.71 kB	+0.03%	+8 B 🔺
@sentry/browser (incl. Metrics & Logs)	29.42 kB	+0.06%	+15 B 🔺
@sentry/react	29.21 kB	+0.04%	+9 B 🔺
@sentry/react (incl. Tracing)	48.53 kB	+1.12%	+533 B 🔺
@sentry/vue	32.9 kB	+1.48%	+477 B 🔺
@sentry/vue (incl. Tracing)	48.13 kB	+1.14%	+542 B 🔺
@sentry/svelte	27.43 kB	+0.02%	+4 B 🔺
CDN Bundle	29.89 kB	+0.34%	+100 B 🔺
CDN Bundle (incl. Tracing)	48.74 kB	+1.14%	+547 B 🔺
CDN Bundle (incl. Logs, Metrics)	31.43 kB	+0.33%	+102 B 🔺
CDN Bundle (incl. Tracing, Logs, Metrics)	50.03 kB	+1.09%	+536 B 🔺
CDN Bundle (incl. Replay, Logs, Metrics)	70.73 kB	+0.15%	+103 B 🔺
CDN Bundle (incl. Tracing, Replay)	86.03 kB	+0.6%	+507 B 🔺
CDN Bundle (incl. Tracing, Replay, Logs, Metrics)	87.35 kB	+0.67%	+575 B 🔺
CDN Bundle (incl. Tracing, Replay, Feedback)	91.9 kB	+0.59%	+530 B 🔺
CDN Bundle (incl. Tracing, Replay, Feedback, Logs, Metrics)	93.17 kB	+0.6%	+554 B 🔺
CDN Bundle - uncompressed	88.85 kB	+0.3%	+261 B 🔺
CDN Bundle (incl. Tracing) - uncompressed	147.64 kB	+1.27%	+1.84 kB 🔺
CDN Bundle (incl. Logs, Metrics) - uncompressed	93.55 kB	+0.28%	+261 B 🔺
CDN Bundle (incl. Tracing, Logs, Metrics) - uncompressed	151.61 kB	+1.23%	+1.84 kB 🔺
CDN Bundle (incl. Replay, Logs, Metrics) - uncompressed	218.38 kB	+0.12%	+261 B 🔺
CDN Bundle (incl. Tracing, Replay) - uncompressed	266.52 kB	+0.7%	+1.85 kB 🔺
CDN Bundle (incl. Tracing, Replay, Logs, Metrics) - uncompressed	270.48 kB	+0.69%	+1.85 kB 🔺
CDN Bundle (incl. Tracing, Replay, Feedback) - uncompressed	280.22 kB	+0.67%	+1.85 kB 🔺
CDN Bundle (incl. Tracing, Replay, Feedback, Logs, Metrics) - uncompressed	284.17 kB	+0.66%	+1.85 kB 🔺
@sentry/nextjs (client)	50.93 kB	+0.95%	+479 B 🔺
@sentry/sveltekit (client)	46.64 kB	+1.13%	+521 B 🔺
@sentry/core/server	76.53 kB	+0.6%	+456 B 🔺
@sentry/core/browser	63.68 kB	+0.74%	+462 B 🔺
@sentry/node-core	61.98 kB	+0.42%	+256 B 🔺
@sentry/node	130.76 kB	+0.19%	+240 B 🔺
@sentry/node - without tracing	74.36 kB	+0.35%	+253 B 🔺
@sentry/aws-serverless	86.52 kB	+0.27%	+229 B 🔺
@sentry/cloudflare (withSentry) - minified	175.72 kB	+1.17%	+2.03 kB 🔺
@sentry/cloudflare (withSentry)	438.89 kB	+1.17%	+5.04 kB 🔺

View base workflow run

andreiborza · 2026-06-10T08:13:30Z

   * @internal
   */
-  public recordException(_exception: unknown, _time?: number | undefined): void {
+  public recordException(_exception: unknown, _time?: SpanTimeInput | undefined): void {


This is a widening change to the more correct SpanTimeInput type, it contains number.

logaretm

Minor Qs, I noticed we added sampled flag to non recording spans but never set in the constructor opts.

Maybe we set it somewhere else?

isaacs

Found some small suggestions that might improve or tidy it up, defend against future mistakes, etc. But generally, this looks great :)

isaacs · 2026-06-10T20:30:24Z

+        dropUndefinedKeys({
+          ...getDynamicSamplingContextFromSpan(span),
+          transaction: source === 'url' ? undefined : spanArguments.name,
+        })) satisfies Partial<DynamicSamplingContext>;


This block is almost identical with the bit in packages/core/src/tracing/idleSpan.ts, looks like they only differ in the default name. Can that be abstracted out to a helper function, like a fancier version of freezeDscOnSpan for TwP root spans?

👍 extracted in 45a8a2f

isaacs · 2026-06-10T20:33:06Z


    return new SentryNonRecordingSpan({
      dropReason: 'ignored',
+      sampled: false,


The root span leaves sampled: undefined, and this one (intentionally) sets it false. It might be a good idea to add a comment so that we don't come along later and ""fix"" the inconsistency. (Also, below, on line 563.)

Updated in 07dfbf4

…ions Non-recording spans can now carry a `sampled` flag and `parentSpanId` on their span context, so `spanToTraceHeader`/`spanToTraceparentHeader` and `spanToJSON` reflect the real decision. In Tracing without Performance mode, root non-recording spans keep the sampling decision deferred (no `sentry-trace` flag, no `sentry-sampled`/`sample_rate` in the DSC) instead of asserting a negative decision that would suppress downstream sampling. Spans created for an unsampled trace (or ignored spans) carry an explicit `sampled: false`.

Lms24

Just some minor qs and a request for streamed spans. Good change overall (even in isolation without the traceprovider switch :) )

Lms24 · 2026-06-12T10:01:48Z

  event: Event,
-  { includeSampleRand = false, sdk = 'cloudflare' }: { includeSampleRand?: boolean; sdk?: 'cloudflare' | 'hono' } = {},
+  {
+    includeSamplingFields = false,


super-l: Not a fan of this tbh as it still limits what we're expecting in the specific values. Why not directly assert on the envelope headers?

This is fine for the PR though, so no need to change it :)

Agreed, it would be better to assert actual values.

I had a go at this but it snowballs quite a bit so I'll extract that out into a separate PR later.

Lms24 · 2026-06-12T10:08:45Z

+   * Sentry-specific sampling decision for this span context.
+   * `undefined` means no local sampling decision was made yet.
+   */
+  sampled?: boolean | undefined;


l: not opposed to this but just to double check: We're fine with diverging from OTel here? My understanding is we can do it because our tracer implementation will just create SentrySpans, so this should be fine.

Right, I walked back on this and go for the same _sampled + tracestate flag approach to be in line with SentrySpan.

Lms24 · 2026-06-12T10:10:44Z

  return {
    span_id,
    trace_id,
+    parent_span_id: (span as { parentSpanId?: string }).parentSpanId,


m: I think we can also update this on spanToStreamedSpanJSON, right?

Added in d5887e7

Lms24 · 2026-06-12T10:11:45Z

+      startSpan({ name: 'ignored-child' }, span => {
+        expect(span).toBeInstanceOf(SentryNonRecordingSpan);
+        // The ignored span still links to its parent so `spanToJSON` can surface it.
+        expect(spanToJSON(span).parent_span_id).toBe(rootSpan.spanContext().spanId);


m: spanToJSON does not change based on traceLifecycle: stream. i think we need to assert against spanToStreamedSpanJSON here.

Updated in d5887e7.

mydea · 2026-06-12T11:40:50Z

+    // In TwP mode, a new trace's sampling decision stays deferred (like `startSpan`) while a
+    // continued trace carries the upstream decision, so baggage and the `sentry-trace` header
+    // agree. Idle spans are always trace roots, so we freeze the DSC here.
+    freezeDscOnTwpRootSpan(span, {


I am not a fan of Twp as it is not super clear. Can we just call this e.g. freezeDscOnRootSpanWithoutSampling or something along these lines?

Right, renamed to freezeDscOnRootSpanWithoutSampling in ba51d24

mydea · 2026-06-12T11:43:01Z

      spanId: this._spanId,
      traceId: this._traceId,
-      traceFlags: TRACE_FLAG_NONE,
+      traceFlags: this._sampled ? TRACE_FLAG_SAMPLED : TRACE_FLAG_NONE,


This is weird and Imho not ideal, that a non recording span can be sampled = true? We should enforce that this is either false or undefined. If it is true we should not create a non recording span I suppose...

I reworked this to be in line with SentrySpan, using _sampled + a flag in trace state and using the same helpers across both span types in ba51d24

mydea · 2026-06-12T11:44:01Z

+      parentSpanId?: string;
+      sampled?: boolean;
+      dsc?: Partial<DynamicSamplingContext>;
+    } = parentSpan


Can we not use an existing utility like spantodsc or similar here?

Updated in ba51d24, if I understand your comment correctly 😅

mydea · 2026-06-12T11:44:46Z

-        ...getDynamicSamplingContextFromSpan(span),
-      } satisfies Partial<DynamicSamplingContext>;
-      freezeDscOnSpan(span, dsc);
+      freezeDscOnTwpRootSpan(span, {


This is not a twp span, is it? Just a forced transaction?

Updated in ba51d24

Just verify correct span nesting for redis

andreiborza · 2026-06-14T15:41:07Z

bugbot run

cursor

Cursor Bugbot has reviewed your changes and found 1 potential issue.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit d5887e7. Configure here.}

andreiborza · 2026-06-15T19:28:36Z

Closing in favor of #21549 which keeps SentryNonRecordingSpan a thin container and uses the scope to handle the sampling decision in cases where it's needed.

…21549) In Tracing-without-Performance (spans disabled), a root placeholder previously froze a negative sampling decision in the DSC, which suppressed downstream sampling instead of leaving the decision to a performance-enabled service further along the trace. The scope is the source of truth for a TwP placeholder's trace state: - `getTraceData` reads the sampling decision from the scope (deferred for a new trace, the upstream decision for a continued trace), so the outgoing `sentry-trace` header omits the flag instead of asserting `-0`. The span id comes from the scope's `propagationSpanId` (a fresh id is generated when the scope has none). - `getDynamicSamplingContextFromSpan` resolves a placeholder's DSC from its captured scope (continued traces keep the incoming DSC; new traces derive it from the client). The scope is only consulted for genuine TwP placeholders. A non-recording span in tracing mode, the child of an unsampled span, or an ignored span carries an explicit negative decision and keeps propagating `-0` via `spanToTraceHeader`. A new (head-of-trace) TwP trace does not stamp a local `transaction` in its DSC; continued traces still propagate the upstream decision and DSC. No DSC is written to the scope at span start, preserving the browser's "scope stays DSC-free between navigations" behavior. This is an alternative to #21406

cursor Bot reviewed Jun 9, 2026

View reviewed changes

Comment thread packages/core/src/utils/spanUtils.ts Outdated

cursor Bot reviewed Jun 9, 2026

View reviewed changes

Comment thread packages/core/src/tracing/trace.ts Outdated

cursor Bot reviewed Jun 9, 2026

View reviewed changes

Comment thread packages/core/src/tracing/trace.ts

Comment thread packages/core/src/tracing/trace.ts Outdated

andreiborza force-pushed the ab/nonrecording-span-sampling branch from 9a455d0 to 69d87c1 Compare June 9, 2026 17:27

andreiborza commented Jun 10, 2026

View reviewed changes

andreiborza marked this pull request as ready for review June 10, 2026 08:22

andreiborza requested a review from a team as a code owner June 10, 2026 08:22

andreiborza requested review from JPeer264, Lms24, chargome, logaretm, mydea and nicohrubec and removed request for a team and JPeer264 June 10, 2026 08:22

logaretm reviewed Jun 10, 2026

View reviewed changes

Comment thread packages/core/src/tracing/idleSpan.ts

Comment thread packages/core/src/tracing/trace.ts

logaretm approved these changes Jun 10, 2026

View reviewed changes

isaacs approved these changes Jun 10, 2026

View reviewed changes

cursor Bot reviewed Jun 11, 2026

View reviewed changes

Comment thread packages/core/src/tracing/trace.ts Outdated

sentry Bot reviewed Jun 11, 2026

View reviewed changes

Comment thread packages/core/src/utils/spanUtils.ts Outdated

andreiborza added 9 commits June 12, 2026 09:39

Defer idle span sampling decision in TwP

bc7f4f9

Freeze continued DSC as is

0a77a65

Capture scopes on the non-recording idle span placeholder

4882b00

Ensure TwP spans store their scope

cf23c31

Ensure parent span id is passed along

e03935a

Don't add span names to TwP DSC when source is url

9a54d7a

Fix cf tests

c117ab2

Ensure non-recording spans propagate incoming sampling decision

26fea7e

andreiborza added 3 commits June 12, 2026 09:39

Ensure DSC stays frozen on nested TwP placeholder spans

5d3f186

Extract functionality into freezeDscOnTwpRootSpan helper

fd15748

Clarify sampling intent

07dfbf4

andreiborza force-pushed the ab/nonrecording-span-sampling branch from b539317 to 07dfbf4 Compare June 12, 2026 07:39

Lms24 approved these changes Jun 12, 2026

View reviewed changes

mydea reviewed Jun 12, 2026

View reviewed changes

chargome and others added 2 commits June 14, 2026 17:07

test(react-router): Fix e2e tests (#21485)

6a476c0

Just verify correct span nesting for redis

Encode span sampling decisions via trace flags and trace state

ba51d24

andreiborza force-pushed the ab/nonrecording-span-sampling branch from ca1ccfb to ba51d24 Compare June 14, 2026 15:07

Surface parent_span_id on spanToStreamedSpanJSON fallback

d5887e7

cursor Bot reviewed Jun 14, 2026

View reviewed changes

Comment thread packages/core/src/tracing/trace.ts

andreiborza mentioned this pull request Jun 15, 2026

fix(core): Defer TwP sampling by reading trace state from the scope #21549

Merged

andreiborza marked this pull request as draft June 15, 2026 16:58

andreiborza closed this Jun 15, 2026

JPeer264 deleted the ab/nonrecording-span-sampling branch June 16, 2026 05:26

Uh oh!

Conversation

andreiborza commented Jun 9, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

What

Why

Uh oh!

Uh oh!

github-actions Bot commented Jun 9, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

size-limit report 📦

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

logaretm left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

isaacs left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Lms24 left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

andreiborza commented Jun 14, 2026

Uh oh!

cursor Bot left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

andreiborza commented Jun 15, 2026

Uh oh!

Reviewers

andreiborza commented Jun 9, 2026 •

edited

Loading

github-actions Bot commented Jun 9, 2026 •

edited

Loading

logaretm left a comment •

edited

Loading