Sentry Knowledge Patch

Sentry JavaScript SDK Knowledge Patch

Use this skill when upgrading, configuring, or debugging Sentry JavaScript SDKs, especially when code crosses major SDK versions, framework wrappers, tracing, structured logs, or modern server runtimes.

Reference index

Upgrade triage

Before changing code:

1. Identify the installed Sentry SDK major and every framework-specific Sentry package.
2. Check the application runtime, framework, router, Prisma, OpenTelemetry, and bundler-plugin versions against the relevant compatibility notes.
3. Search for removed options, imports, integrations, wrappers, and low-level extension types before updating dependencies.
4. Review source-map generation, upload, deletion, and release naming as one build pipeline.
5. Verify sampling, session tracking, error filtering, and privacy behavior with focused tests after migration.

Major migration quick reference

Runtime floors

SDK packages may emit ES2020. Node-based packages require Node.js 18.0.0, while the ESM-only Astro, Nuxt, and SvelteKit packages require Node.js 18.19.1. Deno 2.0 and TypeScript 5.0.4 are the corresponding floors. Transpile for browser targets older than Chrome or Edge 80, Safari 14, iOS Safari 14.4, Firefox 74, Opera 67, or Samsung Internet 13.

Removed tracing switches and sampler fields

Replace enableTracing with tracesSampleRate. In tracesSampler, use normalizedRequest instead of request, and read transaction fields such as name directly from the sampling context instead of transactionContext. Node.js invokes the sampler for trace decisions rather than once per span.

Sentry.init({
  tracesSampler: ({ name, normalizedRequest, inheritOrSampleWith }) =>
    name === "/health-check" ? 0 : inheritOrSampleWith(0.5),
});

An explicitly undefined tracesSampleRate is treated as absent, allowing a downstream service to make its own sampling decision.

Span hooks cannot drop spans

beforeSendSpan receives root and child spans but cannot drop one by returning null. Configure recording or use the appropriate filtering feature instead. When stream mode is enabled, wrap the hook with Sentry.withStreamedSpan() or the SDK falls back to transaction mode.

Session tracking is integration-based

Remove autoSessionTracking. Browser sessions use browserSessionIntegration, server request sessions use httpIntegration, and Node.js process sessions use the default processSessionIntegration.

Sentry.init({
  integrations: [
    Sentry.httpIntegration({ trackIncomingRequestsAsSessions: false }),
  ],
});

Core package and client changes

Do not import from the unpublished @sentry/utils; move remaining imports from it, and deprecated @sentry/types imports, to @sentry/core. Removed hub and metrics APIs have no direct compatibility shim. Custom clients use Client directly; BaseClient is no longer available in the next major line.

OpenTelemetry boundary

Node-based v10 packages require the OpenTelemetry 2.x/0.20x generation. Stay on the prior Sentry major when OpenTelemetry 2 is unavailable, or use @sentry/node-core where its wider peer ranges fit. Pass custom instrumentation through openTelemetryInstrumentations; addOpenTelemetryInstrumentation() is removed.

Feedback and browser privacy

Replace captureUserFeedback() with captureFeedback() and rename payload comments to message. Browser SDKs no longer request backend IP inference by default; enable sendDefaultPii only when intended. Express user data is not copied from request.user; call Sentry.setUser() explicitly.

Framework wrapper selection

Use the explicit V6 or V7 React Router wrapper matching the installed router major. NestJS applications use @sentry/nestjs, SentryGlobalFilter, and @SentryExceptionCaptured. SolidStart uses withSentry; Vue tracing options belong under vueIntegration({ tracingOptions }).

Next.js configuration

Pass SDK options directly to withSentryConfig; do not use the discontinued sentry property in Next config. Set a deterministic release explicitly when needed. The removed hideSourceMaps option has no replacement.

export default withSentryConfig(nextConfig, {
  release: { name: "my-release" },
  sourcemaps: { deleteSourcemapsAfterUpload: false },
});

High-value current capabilities

Structured logs

Enable top-level enableLogs and beforeSendLog. Emit logs with Sentry.logger and attach searchable attributes as the second argument; use Sentry.logger.fmt for parameterized messages.

Sentry.init({ enableLogs: true, beforeSendLog: log => log });
Sentry.logger.info("Order created", { orderId: "order_456" });
Sentry.logger.info(Sentry.logger.fmt`User ${userId} purchased ${productName}`);

Use Sentry.setAttribute() or Sentry.setAttributes() for attributes shared by logs and metrics. Scope methods provide app-wide or operation-local placement.

Streamed spans

SDKs that support stream mode can send completed spans in periodic batches instead of retaining the full transaction. Configure traceLifecycle: "stream" on server SDKs, or install spanStreamingIntegration() in direct browser SDKs.

Sentry.init({ tracesSampleRate: 1, traceLifecycle: "stream" });

Use ignoreSpans for start-time dropping. Rules see only the initial name and attributes. A dropped service span drops descendants; dropping a child reparents its retained children. Record span metadata with attributes because scope tags do not propagate to spans in stream mode.

Framework data controls

Elysia, Hono, Nitro, and TanStack Start accept dataCollection controls for automatic request enrichment. Disable user data and request bodies explicitly where collection is not permitted.

Sentry.init({
  dataCollection: { userInfo: false, httpBodies: [] },
});

Error selection and trace continuation

Use strictTraceContinuation when stricter inbound trace continuation is required. Fastify and supported modern framework integrations expose shouldHandleError for application-specific capture policy.

Sentry.init({
  strictTraceContinuation: true,
  integrations: [
    Sentry.fastifyIntegration({
      shouldHandleError: error => shouldReport(error),
    }),
  ],
});

Verification checklist

• Build every client and server bundle and inspect source-map output and cleanup.
• Exercise one sampled and one unsampled distributed trace across service boundaries.
• Confirm error hooks, framework filters, and serverless flushes complete before responses terminate.
• Verify session counts after replacing legacy automatic tracking.
• Inspect an event for user, IP, request-body, stack-variable, and console handling according to the intended privacy policy.
• Query logs and spans using the new attribute names before changing dashboards.
• Test development, preview, and production entry points when instrumentation is preloaded with --import.