feat(runtime,do): head + error-biased trace sampling - #166
Merged
Conversation
Add a configurable trace-sampling policy to the observability pipeline, mirroring Cloudflare Workers' head_sampling_rate. A deterministic per-trace head decision (derived from the traceId) keeps or drops a whole trace, biased to always keep traces that produced an error span (tail bias). Only trace spans are governed; metrics and logs are untouched. - shared/sampling.ts: zero-dep deterministic helpers (traceIdToUnitInterval, isTraceHeadSampled, resolveTraceSampling, shouldExportTrace). - shared/otlp.ts: buildTraceparent encodes the sampled flag; parseTraceparent returns it, so the runtime propagates its head decision to shards. - runtime: CreateWorkerOptions.sampling; emitRpcEvent gates the SERVER span; dispatchSingleShard sets the traceparent flag + x-lunora-sample-errors header. - do: ShardDO reads the propagated flags, holds sampled-out ctx.trace spans out of the live export, and re-decides at the dispatch finally (export on error). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_0197KjhmBDB3PXAc6TXiPqZ9
✅ Deploy Preview for lunorash ready!
To edit notification comments on pull requests, go to your Netlify project configuration. |
|
Caution Review failedThe pull request is closed. ℹ️ Recent review info⚙️ Run configurationConfiguration used: Path: .coderabbit.yaml Review profile: CHILL Plan: Pro Run ID: ⛔ Files ignored due to path filters (3)
📒 Files selected for processing (1)
Walkthrough
ChangesTrace sampling and tail-biased export
Estimated code review effort: 4 (Complex) | ~45 minutes Sequence Diagram(s)sequenceDiagram
participant Worker
participant ShardDO
participant ObservabilitySink
Worker->>Worker: Resolve trace sampling decision
Worker->>ShardDO: Dispatch with traceparent and error-retention header
ShardDO->>ShardDO: Buffer spans and track dispatch outcome
Worker->>ObservabilitySink: Emit sampled RPC event
ShardDO->>ObservabilitySink: Flush retained error-trace spans
Possibly related PRs
🚥 Pre-merge checks | ✅ 4 | ❌ 1❌ Failed checks (1 warning)
✅ Passed checks (4 passed)
✨ Finishing Touches📝 Generate docstrings
🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
Contributor
|
Thank you for following the naming conventions! 🙏 |
Contributor
|
Thank you for confirming the Contributor License Agreement! 🙏 |
Merging this PR will not alter performance
Comparing 
Footnotes
|
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 1
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@packages/do/src/shard-do.ts`:
- Around line 4847-4909: The sampled-out trace state in recordSpan and
flushSampledOutTrace is stored in shared per-instance fields, allowing
interleaved dispatches to overwrite sampling and sink decisions. Snapshot the
sampling decision, trace identity, and sink within each dispatch’s tracer/anchor
context (leveraging makeTracer’s per-call sink closure), or key deferred
held-trace state by traceId, so recordSpan and flushSampledOutTrace use the
originating dispatch’s state and preserve error flushing under concurrent
dispatches.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
Run ID: 650385eb-7039-4b03-8b4d-c06a98f348bb
⛔ Files ignored due to path filters (4)
packages/do/__tests__/shard-do.sampling.test.tsis excluded by!**/__tests__/**,!**/*.test.tsand included bypackages/**packages/runtime/__tests__/sampling.test.tsis excluded by!**/__tests__/**,!**/*.test.tsand included bypackages/**shared/otlp.tsis excluded by none and included by noneshared/sampling.tsis excluded by none and included by none
📒 Files selected for processing (4)
packages/do/src/shard-do.tspackages/runtime/src/create-worker.tspackages/runtime/src/index.tspackages/runtime/src/observability.ts
Records the new TraceSamplingConfig interface, the `sampling` worker option, and the emitRpcEvent sampling arg in the runtime/lunora api-snapshots. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_0197KjhmBDB3PXAc6TXiPqZ9
…ispatches CodeRabbit (#166, Major): the sampled-out / keep-errors / held-sink state was stored in flat per-instance fields (`currentTraceSampled`, `currentTraceKeepErrors`, `currentDispatchSink`) and read in recordSpan / flushSampledOutTrace, which run after a span body or the whole dispatch settles. A Durable Object interleaves dispatches across await points, so a sibling dispatch's `finally` could reset those fields first — leaking a sampled-out trace's spans live, or (worse) dropping a sampled-out+errored trace's held spans and defeating the always-sample-errors tail bias. The `currentRequestTrace` identity check in recordSpan had the same staleness. Replace the three fields with a single `traceSampling` Map keyed by traceId, registered at dispatch entry and deleted in the `finally`; recordSpan looks up by the span's own `traceId` and flushSampledOutTrace by the dispatch's trace, so two concurrent dispatches can't clobber each other's verdict or sink. Adds an interleaved-dispatch regression test (a slow sampled-out error trace whose flush runs after a sibling sampled-in dispatch's finally). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_0197KjhmBDB3PXAc6TXiPqZ9
This was referenced Jul 24, 2026
prisis
added a commit
that referenced
this pull request
Jul 28, 2026
From better-auth-ui #195, #193, #169, #166, #74 and #337. #195 — the organization limits are now published by uiConfig(). better-auth enforces organizationLimit / allowUserToCreateOrganization server-side and exposes no endpoint that reports them, so a UI can only find out by being refused. The create form is disabled with the reason instead. A function-form allowUserToCreateOrganization is a per-request decision the endpoint cannot evaluate, so it reports allowed and lets the server refuse — the alternative hides the button from everyone because one user might be denied. #166 — the 2FA card no longer vanishes for an OAuth-only account; it explains that a password is needed first. A setting that is simply absent reads as 'this app doesn't support 2FA'. Only a successful accounts read that found no credential row counts: while loading, or on error, the card behaves as before — the flow gate's rule, don't hide what you cannot reason about. #74 — the TOTP key is shown beside the otpauth:// URI. Most authenticators reject a pasted URI, so a desktop app or a broken camera had no way in. #193 — createActiveMemberController resolves the signed-in user's role in the active organization, which otherwise means fetching the org and finding yourself in members at every call site. A rendering hint, not authorization. #337 — organization.showSlug hides the slug field; create already derives one from the name. #169 was already supported by that same fallback. #165 (organization custom fields) is not done: it needs an additionalFields design that spans every card, not a flag. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Tier 0 — trace sampling (mirrors Cloudflare
head_sampling_rate)Configurable head sampling with an always-keep-errors tail bias, so high-traffic apps cut trace volume without losing error traces.
CreateWorkerOptions.sampling = { headRate?: 0..1 (default 1), alwaysSampleErrors?: boolean (default true) }.traceId(shared/sampling.ts, zero-dep) → a trace is wholly kept or dropped (no half traces).emitRpcEvent;ctx.traceINTERNAL spans gated in the DO — the head decision propagates via the W3Ctraceparentsampled flag + anx-lunora-sample-errorsheader (no DO codegen config needed). Spans always buffer locally (Studio panel unaffected); the export decision issampled || (alwaysSampleErrors && traceErrored), made at the flush choke point so the error-keep works even when the error is only known at the end.sampling⇒ everything exported as before.Verify: runtime 585 / do 1158 tests pass; tsc + eslint clean; container
otelconsumers still pass.🤖 Generated with Claude Code
Summary by CodeRabbit