Agent Skills: Sentry Known Pitfalls

Sentry Known Pitfalls

Overview

Ten production-grade Sentry SDK anti-patterns that silently break error tracking, inflate costs, or leave teams blind to failures. Each pitfall includes the broken pattern, root cause, and production-ready fix.

For extended code samples and audit scripts, see configuration pitfalls, error capture pitfalls, SDK initialization pitfalls, integration pitfalls, and monitoring pitfalls.

Prerequisites

• Active Sentry project with @sentry/node >= 8.x or @sentry/browser >= 8.x
• Access to the codebase containing Sentry.init() configuration
• Environment variable management (.env, secrets manager, or CI/CD vars)

Instructions

Step 1: Scan for Existing Pitfalls

# Hardcoded DSNs (Pitfall 1)
grep -rn "ingest\.sentry\.io" --include="*.ts" --include="*.js" src/

# 100% sample rates (Pitfall 2)
grep -rn "sampleRate.*1\.0" --include="*.ts" --include="*.js" src/

# Missing flush calls (Pitfall 3)
grep -rn "Sentry\.flush\|Sentry\.close" --include="*.ts" --include="*.js" src/

# Wrong SDK imports (Pitfall 8)
grep -rn "@sentry/node" --include="*.tsx" --include="*.jsx" src/

Step 2: Pitfall 1 — Hardcoding DSN in Source Code

DSN in source ships in client bundles and cannot be rotated without a deploy. Attackers flood your project with garbage events.

// WRONG
Sentry.init({
  dsn: 'https://abc123@o123456.ingest.us.sentry.io/7890123',
});

// RIGHT — environment variable
Sentry.init({ dsn: process.env.SENTRY_DSN });

// RIGHT — browser apps: build-time injection (Vite)
// vite.config.ts: define: { __SENTRY_DSN__: JSON.stringify(process.env.SENTRY_DSN) }
// app.ts: Sentry.init({ dsn: __SENTRY_DSN__ });

Step 3: Pitfall 2 — sampleRate: 1.0 in Production

100% sampling sends every trace. At 500K requests/day, overage is ~$371/month.

// WRONG
Sentry.init({ tracesSampleRate: 1.0 });

// RIGHT — endpoint-specific sampling
Sentry.init({
  tracesSampler: ({ name, parentSampled }) => {
    if (typeof parentSampled === 'boolean') return parentSampled;
    if (name?.match(/\/(health|ping|ready)/)) return 0;
    if (name?.includes('/checkout')) return 0.25;
    return 0.01;  // 1% default
  },
  replaysSessionSampleRate: 0.1,
  replaysOnErrorSampleRate: 1.0,
});

Step 4: Pitfall 3 — Not Calling flush() in Serverless/CLI

Sentry queues events in memory. Serverless/CLI processes exit before the queue drains — events never reach Sentry.

// WRONG — Lambda exits, events lost
export const handler = async (event) => {
  try { return await processEvent(event); }
  catch (error) {
    Sentry.captureException(error);
    throw error;  // Queue never drains
  }
};

// RIGHT — flush before exit
export const handler = async (event) => {
  try { return await processEvent(event); }
  catch (error) {
    Sentry.captureException(error);
    await Sentry.flush(2000);
    throw error;
  }
};

// BEST — use @sentry/aws-serverless wrapper
import * as Sentry from '@sentry/aws-serverless';
export const handler = Sentry.wrapHandler(async (event) => {
  return await processEvent(event);
});

Step 5: Pitfall 4 — beforeSend Returning null for All Events

Missing return event causes JavaScript to return undefined, which Sentry treats as "drop." A single missing return kills all tracking.

// WRONG — non-error events silently vanish
Sentry.init({
  beforeSend(event) {
    if (event.level === 'error') return event;
    // Falls through — undefined — ALL non-errors dropped
  },
});

// RIGHT — always return event as the last line
Sentry.init({
  beforeSend(event, hint) {
    const error = hint?.originalException;
    if (error instanceof Error && error.message.match(/^NetworkError/)) {
      return null;  // Explicit drop
    }
    return event;  // Always the last line
  },
});

Step 6: Pitfall 5 — Release Version Mismatch (SDK vs Source Maps)

SDK release must exactly match sentry-cli releases new. A v prefix mismatch means source maps never apply.

// WRONG — "1.2.3" vs "v1.2.3"
Sentry.init({ release: process.env.npm_package_version });
// CLI: sentry-cli releases new "v1.2.3"

// RIGHT — single source of truth
const SENTRY_RELEASE = `myapp@${process.env.GIT_SHA || 'dev'}`;
Sentry.init({ release: SENTRY_RELEASE });

# CI — same variable feeds both SDK and CLI
export SENTRY_RELEASE="myapp@$(git rev-parse --short HEAD)"
npx sentry-cli releases new "$SENTRY_RELEASE"
npx sentry-cli sourcemaps upload --release="$SENTRY_RELEASE" \
  --url-prefix="~/static/js" ./dist/static/js/
npx sentry-cli releases finalize "$SENTRY_RELEASE"

Step 7: Pitfall 6 — Catching Errors Without Re-Throwing

Capturing to Sentry but not re-throwing means the function returns undefined. Downstream code breaks silently.

// WRONG — returns undefined
async function getUser(id: string) {
  try {
    return await fetch(`/api/users/${id}`).then(r => r.json());
  } catch (error) {
    Sentry.captureException(error);
    // Returns undefined — callers get TypeError
  }
}

// RIGHT — capture and re-throw
async function getUser(id: string) {
  try {
    return await fetch(`/api/users/${id}`).then(r => r.json());
  } catch (error) {
    Sentry.captureException(error);
    throw error;
  }
}

Step 8: Pitfall 7 — Missing environment Tag

Without environment, dev errors pollute prod dashboards. Alert rules fire on local noise. Issue counts are inflated.

// WRONG
Sentry.init({ dsn: process.env.SENTRY_DSN });

// RIGHT
Sentry.init({
  dsn: process.env.SENTRY_DSN,
  environment: process.env.NODE_ENV || 'development',
});

// For Vercel/Railway preview environments:
function getSentryEnvironment(): string {
  if (process.env.VERCEL_ENV) return process.env.VERCEL_ENV;
  if (process.env.RAILWAY_ENVIRONMENT) return process.env.RAILWAY_ENVIRONMENT;
  return process.env.NODE_ENV || 'development';
}

Step 9: Pitfall 8 — Importing @sentry/node in Browser Bundle

@sentry/node depends on Node.js built-ins (http, fs). Browser import causes build failures, 100KB+ polyfill bloat, or runtime crashes.

// WRONG
import * as Sentry from '@sentry/node';  // In React/Vue/browser code

// RIGHT — platform-specific SDK
import * as Sentry from '@sentry/react';     // React
import * as Sentry from '@sentry/vue';       // Vue
import * as Sentry from '@sentry/nextjs';    // Next.js (client + server)
import * as Sentry from '@sentry/node';      // Server-only
import * as Sentry from '@sentry/aws-serverless';  // AWS Lambda

Step 10: Pitfall 9 — Ignoring 429 Too Many Requests

When quota is exceeded, Sentry returns 429 and the SDK silently drops events. You lose data during peak traffic — exactly when you need it most.

Prevention:

1. Enable Spike Protection in Sentry Organization Settings
2. Set per-key rate limits in Project Settings > Client Keys
3. Monitor client reports: Project Settings > Client Keys > Stats

// Client-side circuit breaker for resilience
let sentryBackoff = 0;
Sentry.init({
  beforeSend(event) {
    if (Date.now() < sentryBackoff) return null;
    return event;
  },
});

Step 11: Pitfall 10 — No Alert Rules Configured

Sentry collects errors but does not notify anyone by default. Without alerts, critical bugs go unnoticed for hours.

Three-tier alert structure:

| Tier | Trigger | Channel | |------|---------|---------| | Immediate | New fatal/error issue in prod | PagerDuty | | Urgent | Error rate > 100 events in 5 min | Slack #alerts | | Awareness | Issue unresolved > 7 days | Email digest |

Set up in Sentry UI: Alerts > Create Alert > Issue Alert (Tier 1) or Metric Alert (Tier 2). See monitoring pitfalls for API-based alert creation.

Step 12: Run the Full Audit Checklist

See audit script for a bash script that checks all 10 pitfalls in one pass.

Output

• Audit report listing which of the 10 pitfalls were found
• Code changes applied for each identified pitfall
• Confirmation that beforeSend returns event on all paths
• environment and release properly configured
• Alert rules created or recommended (three tiers)

Error Handling

| Pitfall | Symptom | Fix | |---------|---------|-----| | Hardcoded DSN | Spam events from attackers | process.env.SENTRY_DSN or build-time injection | | sampleRate: 1.0 | 10-50x cost overrun | tracesSampler with per-endpoint rates | | No flush() | Zero events from Lambda/CLI | await Sentry.flush(2000) before exit | | beforeSend drops all | Events silently vanish | Always end with return event | | Release mismatch | Minified stack traces | Single SENTRY_RELEASE env var | | Swallowed catch | Cascading undefined errors | Re-throw after capture | | No environment | Dev noise in prod dashboard | environment: process.env.NODE_ENV | | Wrong SDK import | Build failure or bloat | Platform-specific SDK package | | Ignoring 429s | Data loss at peak traffic | Spike protection + circuit breaker | | No alerts | Bugs accumulate unnoticed | Three-tier alert rules |

Examples

Example 1: Full-stack audit of existing Sentry setup

Request: "Audit our Sentry integration for common mistakes"

Result: Found hardcoded DSN in config.ts (Pitfall 1), 100% tracesSampleRate (Pitfall 2), no environment tag (Pitfall 7), and zero alert rules (Pitfall 10). Applied fixes for all four, added CI gate for DSN detection, created three-tier alert config. See examples for more scenarios.

Example 2: Debugging missing Lambda errors

Request: "Sentry shows no errors from our Lambda functions but we know they're failing"

Result: Identified Pitfall 3 — no flush() call before Lambda return. Wrapped all handlers with Sentry.wrapHandler() from @sentry/aws-serverless. Events now appear within 5 seconds of invocation.

Example 3: Source map stack traces showing minified code

Request: "Sentry stack traces are all minified even though we upload source maps"

Result: SDK used release: "2.1.0" while CLI used "v2.1.0" (Pitfall 5). Unified both to $GIT_SHA via shared SENTRY_RELEASE env var. Stack traces now show original TypeScript source.

Resources

• Sentry JavaScript Troubleshooting
• Sentry Best Practices
• Sentry Configuration Options
• Sentry Source Maps
• Sentry Quota Management
• Sentry Alerts

Next Steps

• Run the scan commands from Step 1 against your codebase
• Fix pitfalls in priority order: Pitfall 1 (security) > Pitfall 3 (data loss) > Pitfall 10 (alerting)
• Add DSN CI gate to prevent regression
• Set up three-tier alert structure before further Sentry work